cephi/ampere_research
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

datarm ../pytorch-xeon_4216.Containerfile !

Browse Source
This commit is contained in:
cephi 2024-12-17 17:15:38 -05:00
parent 17b9f75aee
commit a9a2b170fc
193 changed files with 8613 additions and 8109 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1755, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.253487825393677, "TIME_S_1KI": 5.842443205352523, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 467.2838370990753, "W": 31.934101368331916, "J_1KI": 266.2585966376497, "W_1KI": 18.196069155744684, "W_D": 16.879101368331916, "J_D": 246.98773149132728, "W_D_1KI": 9.617721577397104, "J_D_1KI": 5.480183234984104}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1748, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.143786191940308, "TIME_S_1KI": 5.8030813455036085, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 530.3414198303223, "W": 36.295146632325824, "J_1KI": 303.3989815962942, "W_1KI": 20.763813862886625, "W_D": 17.602146632325823, "J_D": 257.20098424220083, "W_D_1KI": 10.069877936113171, "J_D_1KI": 5.760799734618519}

58
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_0.0001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 100000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 5.980836629867554}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 100000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.6005632877349854}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 11, 22, ..., 999980,
tensor(crow_indices=tensor([ 0, 7, 17, ..., 999970,
999989, 1000000]),
col_indices=tensor([ 6100, 13265, 27848, ..., 84407, 91090, 94721]),
values=tensor([0.4400, 0.3445, 0.5606, ..., 0.5861, 0.7102, 0.2795]),
col_indices=tensor([27708, 32922, 35240, ..., 82805, 88487, 98517]),
values=tensor([0.0088, 0.7733, 0.0012, ..., 0.6420, 0.7382, 0.2177]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.6757, 0.5029, 0.1898, ..., 0.2612, 0.6123, 0.0844])
tensor([0.1129, 0.5965, 0.7496, ..., 0.0902, 0.9107, 0.7724])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,19 +16,19 @@ Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 5.980836629867554 seconds
Time: 0.6005632877349854 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1755 -ss 100000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.253487825393677}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1748 -ss 100000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.143786191940308}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 5, 18, ..., 999983,
999994, 1000000]),
col_indices=tensor([22305, 51740, 53616, ..., 72974, 76091, 88145]),
values=tensor([0.7756, 0.0657, 0.7358, ..., 0.9841, 0.0331, 0.5251]),
tensor(crow_indices=tensor([ 0, 10, 23, ..., 999980,
999992, 1000000]),
col_indices=tensor([ 8017, 17251, 18992, ..., 72823, 91334, 91663]),
values=tensor([0.4596, 0.1797, 0.9797, ..., 0.0499, 0.7967, 0.0183]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.4750, 0.6821, 0.2847, ..., 0.3502, 0.4038, 0.5877])
tensor([0.8873, 0.5523, 0.3791, ..., 0.8812, 0.4027, 0.2259])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -36,16 +36,16 @@ Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 10.253487825393677 seconds
Time: 10.143786191940308 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 5, 18, ..., 999983,
999994, 1000000]),
col_indices=tensor([22305, 51740, 53616, ..., 72974, 76091, 88145]),
values=tensor([0.7756, 0.0657, 0.7358, ..., 0.9841, 0.0331, 0.5251]),
tensor(crow_indices=tensor([ 0, 10, 23, ..., 999980,
999992, 1000000]),
col_indices=tensor([ 8017, 17251, 18992, ..., 72823, 91334, 91663]),
values=tensor([0.4596, 0.1797, 0.9797, ..., 0.0499, 0.7967, 0.0183]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.4750, 0.6821, 0.2847, ..., 0.3502, 0.4038, 0.5877])
tensor([0.8873, 0.5523, 0.3791, ..., 0.8812, 0.4027, 0.2259])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -53,13 +53,13 @@ Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 10.253487825393677 seconds
Time: 10.143786191940308 seconds
[16.48, 16.16, 16.52, 16.48, 16.48, 16.48, 16.72, 16.72, 16.84, 17.12]
[17.08, 17.16, 17.96, 19.72, 22.2, 27.16, 34.52, 39.08, 43.72, 45.96, 46.32, 46.32, 46.2, 46.28]
14.632753610610962
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1755, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.253487825393677, 'TIME_S_1KI': 5.842443205352523, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 467.2838370990753, 'W': 31.934101368331916}
[16.48, 16.16, 16.52, 16.48, 16.48, 16.48, 16.72, 16.72, 16.84, 17.12, 16.72, 16.8, 16.96, 17.0, 17.04, 16.88, 16.8, 16.88, 16.76, 16.84]
301.1
15.055000000000001
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1755, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.253487825393677, 'TIME_S_1KI': 5.842443205352523, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 467.2838370990753, 'W': 31.934101368331916, 'J_1KI': 266.2585966376497, 'W_1KI': 18.196069155744684, 'W_D': 16.879101368331916, 'J_D': 246.98773149132728, 'W_D_1KI': 9.617721577397104, 'J_D_1KI': 5.480183234984104}
[20.48, 20.6, 20.64, 20.68, 21.12, 21.12, 21.24, 21.12, 21.04, 21.04]
[20.76, 20.76, 21.04, 22.36, 24.68, 32.36, 38.56, 45.04, 50.12, 51.88, 51.52, 52.0, 52.0, 51.84]
14.611910104751587
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1748, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.143786191940308, 'TIME_S_1KI': 5.8030813455036085, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 530.3414198303223, 'W': 36.295146632325824}
[20.48, 20.6, 20.64, 20.68, 21.12, 21.12, 21.24, 21.12, 21.04, 21.04, 20.36, 20.64, 20.56, 20.72, 20.68, 20.52, 20.72, 20.76, 20.52, 20.48]
373.86
18.693
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1748, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.143786191940308, 'TIME_S_1KI': 5.8030813455036085, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 530.3414198303223, 'W': 36.295146632325824, 'J_1KI': 303.3989815962942, 'W_1KI': 20.763813862886625, 'W_D': 17.602146632325823, 'J_D': 257.20098424220083, 'W_D_1KI': 10.069877936113171, 'J_D_1KI': 5.760799734618519}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_0.001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 57.53653693199158, "TIME_S_1KI": 57.53653693199158, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2684.526071929932, "W": 41.311972802980506, "J_1KI": 2684.526071929932, "W_1KI": 41.311972802980506, "W_D": 26.003972802980506, "J_D": 1689.784782156945, "W_D_1KI": 26.003972802980506, "J_D_1KI": 26.003972802980506}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 175, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 11.183927774429321, "TIME_S_1KI": 63.90815871102469, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 628.162894821167, "W": 35.42748603907242, "J_1KI": 3589.5022561209544, "W_1KI": 202.44277736612813, "W_D": 16.71548603907242, "J_D": 296.38140530395515, "W_D_1KI": 95.51706308041383, "J_D_1KI": 545.8117890309362}

64
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_0.001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 100000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 57.53653693199158}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 100000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 5.9738054275512695}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 92, 190, ..., 9999802,
9999900, 10000000]),
col_indices=tensor([ 766, 3080, 3658, ..., 98863, 99077, 99078]),
values=tensor([0.0329, 0.7493, 0.2063, ..., 0.1215, 0.3807, 0.7288]),
tensor(crow_indices=tensor([ 0, 100, 229, ..., 9999825,
9999913, 10000000]),
col_indices=tensor([ 2839, 3131, 5153, ..., 92533, 94576, 98932]),
values=tensor([0.4697, 0.9996, 0.7875, ..., 0.5192, 0.5202, 0.9540]),
size=(100000, 100000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.3516, 0.3347, 0.9443, ..., 0.8917, 0.2195, 0.8723])
tensor([0.9598, 0.0952, 0.8851, ..., 0.3844, 0.8104, 0.5939])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,16 +16,19 @@ Rows: 100000
Size: 10000000000
NNZ: 10000000
Density: 0.001
Time: 57.53653693199158 seconds
Time: 5.9738054275512695 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 175 -ss 100000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 11.183927774429321}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 92, 190, ..., 9999802,
9999900, 10000000]),
col_indices=tensor([ 766, 3080, 3658, ..., 98863, 99077, 99078]),
values=tensor([0.0329, 0.7493, 0.2063, ..., 0.1215, 0.3807, 0.7288]),
tensor(crow_indices=tensor([ 0, 95, 193, ..., 9999801,
9999901, 10000000]),
col_indices=tensor([ 2869, 4015, 6080, ..., 94953, 95635, 98117]),
values=tensor([0.0857, 0.9758, 0.7363, ..., 0.8151, 0.8595, 0.7723]),
size=(100000, 100000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.3516, 0.3347, 0.9443, ..., 0.8917, 0.2195, 0.8723])
tensor([0.7586, 0.5970, 0.0221, ..., 0.6721, 0.2659, 0.4588])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -33,13 +36,30 @@ Rows: 100000
Size: 10000000000
NNZ: 10000000
Density: 0.001
Time: 57.53653693199158 seconds
Time: 11.183927774429321 seconds
[17.2, 17.2, 16.8, 16.8, 17.0, 16.88, 16.88, 17.0, 17.0, 16.6]
[16.88, 16.84, 17.36, 19.48, 20.16, 21.92, 24.16, 24.88, 27.6, 33.04, 37.56, 43.08, 47.0, 46.96, 47.68, 47.76, 47.76, 47.48, 47.4, 46.92, 47.28, 47.04, 47.56, 48.36, 48.0, 47.68, 47.44, 46.16, 45.68, 46.04, 46.32, 47.44, 47.76, 47.84, 47.64, 47.36, 47.08, 46.96, 46.96, 47.16, 46.68, 46.24, 46.2, 46.44, 46.56, 47.0, 48.08, 48.0, 48.12, 48.44, 48.48, 48.2, 47.64, 47.32, 47.2, 47.2, 47.56, 47.52, 47.68, 47.8, 47.8, 48.0]
64.98179316520691
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 57.53653693199158, 'TIME_S_1KI': 57.53653693199158, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2684.526071929932, 'W': 41.311972802980506}
[17.2, 17.2, 16.8, 16.8, 17.0, 16.88, 16.88, 17.0, 17.0, 16.6, 17.12, 17.12, 17.04, 17.0, 17.0, 16.96, 16.88, 16.92, 17.32, 17.8]
306.16
15.308000000000002
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 57.53653693199158, 'TIME_S_1KI': 57.53653693199158, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2684.526071929932, 'W': 41.311972802980506, 'J_1KI': 2684.526071929932, 'W_1KI': 41.311972802980506, 'W_D': 26.003972802980506, 'J_D': 1689.784782156945, 'W_D_1KI': 26.003972802980506, 'J_D_1KI': 26.003972802980506}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 95, 193, ..., 9999801,
9999901, 10000000]),
col_indices=tensor([ 2869, 4015, 6080, ..., 94953, 95635, 98117]),
values=tensor([0.0857, 0.9758, 0.7363, ..., 0.8151, 0.8595, 0.7723]),
size=(100000, 100000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.7586, 0.5970, 0.0221, ..., 0.6721, 0.2659, 0.4588])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 10000000
Density: 0.001
Time: 11.183927774429321 seconds
[20.52, 20.84, 20.72, 20.68, 20.8, 20.68, 20.68, 20.72, 20.64, 20.64]
[20.64, 20.76, 20.84, 24.4, 27.0, 28.6, 31.96, 32.96, 34.0, 38.76, 44.2, 47.96, 51.6, 50.84, 50.88, 50.6, 50.76]
17.730947494506836
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 175, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 11.183927774429321, 'TIME_S_1KI': 63.90815871102469, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 628.162894821167, 'W': 35.42748603907242}
[20.52, 20.84, 20.72, 20.68, 20.8, 20.68, 20.68, 20.72, 20.64, 20.64, 20.64, 20.64, 20.64, 20.72, 20.76, 20.76, 21.0, 20.92, 21.36, 21.56]
374.24
18.712
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 175, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 11.183927774429321, 'TIME_S_1KI': 63.90815871102469, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 628.162894821167, 'W': 35.42748603907242, 'J_1KI': 3589.5022561209544, 'W_1KI': 202.44277736612813, 'W_D': 16.71548603907242, 'J_D': 296.38140530395515, 'W_D_1KI': 95.51706308041383, 'J_D_1KI': 545.8117890309362}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 11928, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.779903888702393, "TIME_S_1KI": 0.9037478109240772, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 428.71311614990236, "W": 29.263266081724595, "J_1KI": 35.94174347333186, "W_1KI": 2.4533254595677896, "W_D": 14.015266081724594, "J_D": 205.326650100708, "W_D_1KI": 1.1749887727803985, "J_D_1KI": 0.09850677169520444}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 11597, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.436057329177856, "TIME_S_1KI": 0.8998928454926151, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 483.6025880336762, "W": 33.00177532885384, "J_1KI": 41.700662932971994, "W_1KI": 2.845716592985586, "W_D": 14.225775328853839, "J_D": 208.462172027588, "W_D_1KI": 1.2266771862424626, "J_D_1KI": 0.10577538900081596}

98
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_1e-05.output
View File

@ -1,14 +1,54 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 100000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.8802759647369385}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 100000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.12187767028808594}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 2, ..., 99998, 99999,
100000]),
col_indices=tensor([99237, 81965, 52149, ..., 94819, 50598, 82628]),
values=tensor([0.3300, 0.8237, 0.5005, ..., 0.6469, 0.1010, 0.4687]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.0038, 0.2456, 0.3182, ..., 0.7163, 0.7510, 0.9775])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 0.12187767028808594 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 8615 -ss 100000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 7.799410104751587}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 99998, 99999,
100000]),
col_indices=tensor([88588, 42232, 90125, ..., 27244, 80106, 39636]),
values=tensor([0.8018, 0.8315, 0.5597, ..., 0.5532, 0.0030, 0.5793]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.1929, 0.1411, 0.4568, ..., 0.6294, 0.2188, 0.4350])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 7.799410104751587 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 11597 -ss 100000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.436057329177856}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 99998, 99998,
100000]),
col_indices=tensor([50190, 32056, 73796, ..., 55938, 31334, 37461]),
values=tensor([0.0722, 0.7116, 0.8310, ..., 0.7930, 0.8115, 0.4149]),
col_indices=tensor([24172, 83350, 29274, ..., 76990, 53592, 71081]),
values=tensor([0.7302, 0.8346, 0.3553, ..., 0.4222, 0.8183, 0.0288]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.5168, 0.3496, 0.0063, ..., 0.9888, 0.0960, 0.5324])
tensor([0.8537, 0.9431, 0.0277, ..., 0.1357, 0.7019, 0.9196])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,19 +56,16 @@ Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 0.8802759647369385 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 11928 -ss 100000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.779903888702393}
Time: 10.436057329177856 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2, 3, ..., 99995, 99996,
tensor(crow_indices=tensor([ 0, 1, 1, ..., 99998, 99998,
100000]),
col_indices=tensor([15079, 22431, 71484, ..., 38240, 57604, 63673]),
values=tensor([0.6856, 0.2309, 0.0261, ..., 0.6883, 0.7108, 0.1151]),
col_indices=tensor([24172, 83350, 29274, ..., 76990, 53592, 71081]),
values=tensor([0.7302, 0.8346, 0.3553, ..., 0.4222, 0.8183, 0.0288]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.6131, 0.6051, 0.4027, ..., 0.3545, 0.9505, 0.4978])
tensor([0.8537, 0.9431, 0.0277, ..., 0.1357, 0.7019, 0.9196])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -36,30 +73,13 @@ Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 10.779903888702393 seconds
Time: 10.436057329177856 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2, 3, ..., 99995, 99996,
100000]),
col_indices=tensor([15079, 22431, 71484, ..., 38240, 57604, 63673]),
values=tensor([0.6856, 0.2309, 0.0261, ..., 0.6883, 0.7108, 0.1151]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.6131, 0.6051, 0.4027, ..., 0.3545, 0.9505, 0.4978])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 10.779903888702393 seconds
[17.04, 17.24, 17.0, 17.0, 16.8, 16.64, 16.64, 16.68, 16.92, 16.84]
[16.64, 16.52, 16.72, 17.8, 20.04, 25.4, 30.88, 34.96, 39.88, 41.96, 42.28, 42.44, 42.56, 42.56]
14.650214195251465
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 11928, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.779903888702393, 'TIME_S_1KI': 0.9037478109240772, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 428.71311614990236, 'W': 29.263266081724595}
[17.04, 17.24, 17.0, 17.0, 16.8, 16.64, 16.64, 16.68, 16.92, 16.84, 16.84, 16.84, 17.0, 17.0, 16.92, 17.0, 17.16, 17.0, 17.16, 17.2]
304.96000000000004
15.248000000000001
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 11928, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.779903888702393, 'TIME_S_1KI': 0.9037478109240772, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 428.71311614990236, 'W': 29.263266081724595, 'J_1KI': 35.94174347333186, 'W_1KI': 2.4533254595677896, 'W_D': 14.015266081724594, 'J_D': 205.326650100708, 'W_D_1KI': 1.1749887727803985, 'J_D_1KI': 0.09850677169520444}
[21.08, 20.64, 20.76, 20.76, 20.92, 21.32, 21.32, 21.36, 21.0, 20.8]
[20.64, 20.36, 20.72, 23.08, 24.72, 29.44, 35.88, 40.08, 43.56, 45.4, 45.96, 45.6, 45.36, 46.04]
14.653835535049438
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 11597, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.436057329177856, 'TIME_S_1KI': 0.8998928454926151, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 483.6025880336762, 'W': 33.00177532885384}
[21.08, 20.64, 20.76, 20.76, 20.92, 21.32, 21.32, 21.36, 21.0, 20.8, 20.36, 20.32, 20.32, 20.44, 20.64, 20.8, 21.0, 21.16, 21.12, 21.04]
375.52
18.776
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 11597, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.436057329177856, 'TIME_S_1KI': 0.8998928454926151, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 483.6025880336762, 'W': 33.00177532885384, 'J_1KI': 41.700662932971994, 'W_1KI': 2.845716592985586, 'W_D': 14.225775328853839, 'J_D': 208.462172027588, 'W_D_1KI': 1.2266771862424626, 'J_D_1KI': 0.10577538900081596}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 3268, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.551287174224854, "TIME_S_1KI": 3.2286680459684374, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 514.9588033294677, "W": 35.14704993762719, "J_1KI": 157.57613320975145, "W_1KI": 10.754911241623986, "W_D": 16.333049937627187, "J_D": 239.30451817512508, "W_D_1KI": 4.997873297927536, "J_D_1KI": 1.5293369944698703}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 3297, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.891435861587524, "TIME_S_1KI": 3.3034382352403777, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 528.672490940094, "W": 36.05478479295784, "J_1KI": 160.34955745832394, "W_1KI": 10.935633846817664, "W_D": 17.45578479295784, "J_D": 255.95474444794658, "W_D_1KI": 5.2944448871573675, "J_D_1KI": 1.605837090432929}

74
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_100000_5e-05.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 100000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 3.21274733543396}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 100000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.4573814868927002}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 5, 10, ..., 499991, 499996,
tensor(crow_indices=tensor([ 0, 10, 16, ..., 499990, 499995,
500000]),
col_indices=tensor([ 6819, 16249, 65142, ..., 35181, 90238, 95591]),
values=tensor([0.9907, 0.7784, 0.8470, ..., 0.0401, 0.4552, 0.5172]),
col_indices=tensor([ 5164, 6869, 8448, ..., 29154, 68140, 97893]),
values=tensor([0.8386, 0.0921, 0.7067, ..., 0.9232, 0.1449, 0.6848]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.1211, 0.3699, 0.8120, ..., 0.3387, 0.3308, 0.0427])
tensor([0.0246, 0.8160, 0.3295, ..., 0.0588, 0.6998, 0.9868])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,19 +16,19 @@ Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 3.21274733543396 seconds
Time: 0.4573814868927002 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 3268 -ss 100000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.551287174224854}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 2295 -ss 100000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 7.306779146194458}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 10, ..., 499988, 499994,
tensor(crow_indices=tensor([ 0, 4, 8, ..., 499989, 499995,
500000]),
col_indices=tensor([ 4698, 29712, 45324, ..., 47109, 54294, 79244]),
values=tensor([0.0467, 0.0395, 0.2018, ..., 0.4601, 0.1623, 0.8954]),
col_indices=tensor([ 2059, 19971, 54406, ..., 65065, 65922, 83323]),
values=tensor([0.5530, 0.6181, 0.7781, ..., 0.5380, 0.6243, 0.8378]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.1366, 0.5632, 0.5706, ..., 0.2593, 0.9938, 0.0917])
tensor([0.4055, 0.5945, 0.9428, ..., 0.6446, 0.1456, 0.3700])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -36,16 +36,19 @@ Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 10.551287174224854 seconds
Time: 7.306779146194458 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 3297 -ss 100000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.891435861587524}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 10, ..., 499988, 499994,
tensor(crow_indices=tensor([ 0, 5, 11, ..., 499995, 499995,
500000]),
col_indices=tensor([ 4698, 29712, 45324, ..., 47109, 54294, 79244]),
values=tensor([0.0467, 0.0395, 0.2018, ..., 0.4601, 0.1623, 0.8954]),
col_indices=tensor([ 8913, 22689, 49331, ..., 65321, 72756, 72788]),
values=tensor([0.7511, 0.0782, 0.7533, ..., 0.6341, 0.1803, 0.2288]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.1366, 0.5632, 0.5706, ..., 0.2593, 0.9938, 0.0917])
tensor([0.5601, 0.4293, 0.2285, ..., 0.5137, 0.5400, 0.5797])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -53,13 +56,30 @@ Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 10.551287174224854 seconds
Time: 10.891435861587524 seconds
[20.52, 20.48, 20.56, 20.6, 20.6, 20.88, 21.08, 21.04, 21.04, 20.92]
[20.68, 20.64, 20.84, 22.28, 23.8, 29.96, 35.72, 42.6, 47.2, 50.88, 50.56, 50.96, 50.68, 50.8]
14.651551246643066
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3268, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.551287174224854, 'TIME_S_1KI': 3.2286680459684374, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 514.9588033294677, 'W': 35.14704993762719}
[20.52, 20.48, 20.56, 20.6, 20.6, 20.88, 21.08, 21.04, 21.04, 20.92, 20.72, 21.08, 21.16, 21.16, 21.24, 21.28, 21.16, 20.88, 20.68, 20.56]
376.28
18.814
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3268, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.551287174224854, 'TIME_S_1KI': 3.2286680459684374, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 514.9588033294677, 'W': 35.14704993762719, 'J_1KI': 157.57613320975145, 'W_1KI': 10.754911241623986, 'W_D': 16.333049937627187, 'J_D': 239.30451817512508, 'W_D_1KI': 4.997873297927536, 'J_D_1KI': 1.5293369944698703}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 5, 11, ..., 499995, 499995,
500000]),
col_indices=tensor([ 8913, 22689, 49331, ..., 65321, 72756, 72788]),
values=tensor([0.7511, 0.0782, 0.7533, ..., 0.6341, 0.1803, 0.2288]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.5601, 0.4293, 0.2285, ..., 0.5137, 0.5400, 0.5797])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 10.891435861587524 seconds
[20.84, 20.48, 20.44, 20.64, 20.64, 20.68, 20.72, 20.64, 20.64, 20.92]
[20.6, 20.48, 20.72, 25.12, 26.88, 32.4, 38.84, 42.28, 46.96, 50.12, 51.2, 51.8, 51.56, 51.6]
14.66303277015686
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3297, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.891435861587524, 'TIME_S_1KI': 3.3034382352403777, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 528.672490940094, 'W': 36.05478479295784}
[20.84, 20.48, 20.44, 20.64, 20.64, 20.68, 20.72, 20.64, 20.64, 20.92, 20.84, 21.04, 20.84, 20.84, 20.76, 20.8, 20.68, 20.44, 20.28, 20.24]
371.98
18.599
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3297, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.891435861587524, 'TIME_S_1KI': 3.3034382352403777, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 528.672490940094, 'W': 36.05478479295784, 'J_1KI': 160.34955745832394, 'W_1KI': 10.935633846817664, 'W_D': 17.45578479295784, 'J_D': 255.95474444794658, 'W_D_1KI': 5.2944448871573675, 'J_D_1KI': 1.605837090432929}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 32824, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.612937211990356, "TIME_S_1KI": 0.3233285770165232, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 275.6484677696228, "W": 19.42651909848855, "J_1KI": 8.397771989081855, "W_1KI": 0.5918388709020398, "W_D": 4.498519098488551, "J_D": 63.83078154373167, "W_D_1KI": 0.13704969225227123, "J_D_1KI": 0.004175289186335341}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 32636, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.47010850906372, "TIME_S_1KI": 0.32081469877018387, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 319.63640854835506, "W": 22.48260624860275, "J_1KI": 9.793982367580435, "W_1KI": 0.6888897612637195, "W_D": 3.984606248602752, "J_D": 56.64935891771315, "W_D_1KI": 0.12209235962136145, "J_D_1KI": 0.003741033203252894}

68
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.0001.output
View File

@ -1,13 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.3622722625732422}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.050879478454589844}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2, 4, ..., 9997, 10000, 10000]),
col_indices=tensor([2430, 5032, 1477, ..., 758, 3153, 4599]),
values=tensor([0.8038, 0.4543, 0.3152, ..., 0.6785, 0.4391, 0.0535]),
tensor(crow_indices=tensor([ 0, 0, 1, ..., 9998, 9999, 10000]),
col_indices=tensor([5382, 2827, 5658, ..., 9195, 8647, 1137]),
values=tensor([0.6423, 0.5656, 0.8194, ..., 0.3825, 0.7281, 0.0248]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.9594, 0.1900, 0.3074, ..., 0.8950, 0.9459, 0.6732])
tensor([0.6609, 0.7541, 0.4159, ..., 0.2180, 0.3481, 0.0053])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -15,18 +15,18 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 0.3622722625732422 seconds
Time: 0.050879478454589844 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 28983 -ss 10000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 9.27123761177063}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 20637 -ss 10000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 6.639445781707764}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 2, ..., 10000, 10000, 10000]),
col_indices=tensor([1532, 2817, 884, ..., 2356, 6175, 1948]),
values=tensor([0.3809, 0.2852, 0.7235, ..., 0.6592, 0.2563, 0.7726]),
tensor(crow_indices=tensor([ 0, 0, 1, ..., 9998, 10000, 10000]),
col_indices=tensor([1538, 6690, 5733, ..., 9607, 7438, 7782]),
values=tensor([0.7222, 0.1089, 0.5631, ..., 0.3116, 0.0243, 0.6999]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.6771, 0.1497, 0.5070, ..., 0.8092, 0.9643, 0.7887])
tensor([0.2878, 0.8940, 0.0961, ..., 0.0631, 0.2895, 0.2219])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -34,18 +34,18 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 9.27123761177063 seconds
Time: 6.639445781707764 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 32824 -ss 10000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.612937211990356}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 32636 -ss 10000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.47010850906372}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 9999, 9999, 10000]),
col_indices=tensor([3350, 4490, 6839, ..., 6784, 8596, 8737]),
values=tensor([0.3991, 0.5600, 0.2439, ..., 0.8859, 0.9485, 0.6345]),
tensor(crow_indices=tensor([ 0, 1, 2, ..., 9999, 10000, 10000]),
col_indices=tensor([7014, 8766, 3433, ..., 9466, 1431, 7728]),
values=tensor([0.0370, 0.3747, 0.2051, ..., 0.2901, 0.3737, 0.7201]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.2861, 0.2741, 0.4038, ..., 0.8389, 0.9796, 0.7969])
tensor([0.8451, 0.4833, 0.4298, ..., 0.9015, 0.0937, 0.6764])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -53,15 +53,15 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 10.612937211990356 seconds
Time: 10.47010850906372 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 9999, 9999, 10000]),
col_indices=tensor([3350, 4490, 6839, ..., 6784, 8596, 8737]),
values=tensor([0.3991, 0.5600, 0.2439, ..., 0.8859, 0.9485, 0.6345]),
tensor(crow_indices=tensor([ 0, 1, 2, ..., 9999, 10000, 10000]),
col_indices=tensor([7014, 8766, 3433, ..., 9466, 1431, 7728]),
values=tensor([0.0370, 0.3747, 0.2051, ..., 0.2901, 0.3737, 0.7201]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.2861, 0.2741, 0.4038, ..., 0.8389, 0.9796, 0.7969])
tensor([0.8451, 0.4833, 0.4298, ..., 0.9015, 0.0937, 0.6764])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -69,13 +69,13 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 10.612937211990356 seconds
Time: 10.47010850906372 seconds
[16.64, 16.84, 16.96, 16.88, 16.84, 16.72, 17.08, 16.96, 16.92, 16.92]
[17.08, 16.72, 16.76, 21.08, 22.52, 24.76, 25.6, 23.4, 22.04, 20.32, 20.04, 20.0, 20.0, 20.12]
14.189287662506104
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 32824, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.612937211990356, 'TIME_S_1KI': 0.3233285770165232, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 275.6484677696228, 'W': 19.42651909848855}
[16.64, 16.84, 16.96, 16.88, 16.84, 16.72, 17.08, 16.96, 16.92, 16.92, 16.36, 16.04, 15.84, 15.92, 16.12, 16.28, 16.36, 16.68, 16.72, 16.88]
298.56
14.928
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 32824, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.612937211990356, 'TIME_S_1KI': 0.3233285770165232, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 275.6484677696228, 'W': 19.42651909848855, 'J_1KI': 8.397771989081855, 'W_1KI': 0.5918388709020398, 'W_D': 4.498519098488551, 'J_D': 63.83078154373167, 'W_D_1KI': 0.13704969225227123, 'J_D_1KI': 0.004175289186335341}
[20.64, 20.64, 20.48, 20.48, 20.48, 20.4, 20.32, 20.16, 20.24, 20.4]
[20.72, 20.64, 21.24, 23.32, 25.32, 26.04, 26.72, 26.72, 26.48, 25.08, 23.72, 23.6, 23.56, 23.48]
14.217053174972534
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 32636, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.47010850906372, 'TIME_S_1KI': 0.32081469877018387, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 319.63640854835506, 'W': 22.48260624860275}
[20.64, 20.64, 20.48, 20.48, 20.48, 20.4, 20.32, 20.16, 20.24, 20.4, 20.64, 20.56, 20.48, 20.52, 20.68, 20.84, 20.88, 20.88, 20.84, 20.48]
369.96
18.497999999999998
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 32636, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.47010850906372, 'TIME_S_1KI': 0.32081469877018387, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 319.63640854835506, 'W': 22.48260624860275, 'J_1KI': 9.793982367580435, 'W_1KI': 0.6888897612637195, 'W_D': 3.984606248602752, 'J_D': 56.64935891771315, 'W_D_1KI': 0.12209235962136145, 'J_D_1KI': 0.003741033203252894}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 4599, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.21649432182312, "TIME_S_1KI": 2.2214599525599303, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 276.3690100479126, "W": 19.391688491473598, "J_1KI": 60.09328333287945, "W_1KI": 4.2165010853389, "W_D": 4.4646884914736, "J_D": 63.630433167457575, "W_D_1KI": 0.9707954971675582, "J_D_1KI": 0.21108838816428752}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 4519, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.300970077514648, "TIME_S_1KI": 2.279479990598506, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 332.2094431686402, "W": 23.392743273719635, "J_1KI": 73.51392856132777, "W_1KI": 5.176530930232272, "W_D": 4.8837432737196345, "J_D": 69.35593720483783, "W_D_1KI": 1.0807132714582062, "J_D_1KI": 0.2391487655362262}

74
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 2.282747268676758}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.2727935314178467}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 15, 26, ..., 99983, 99992,
tensor(crow_indices=tensor([ 0, 6, 18, ..., 99975, 99993,
100000]),
col_indices=tensor([ 746, 1254, 2691, ..., 5665, 9904, 9986]),
values=tensor([0.7024, 0.2927, 0.8116, ..., 0.2675, 0.5863, 0.1724]),
col_indices=tensor([2872, 4034, 5620, ..., 6357, 6556, 9590]),
values=tensor([0.7995, 0.0045, 0.2448, ..., 0.5761, 0.7842, 0.1546]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.2042, 0.3555, 0.3767, ..., 0.6038, 0.4952, 0.0036])
tensor([0.8077, 0.7130, 0.7281, ..., 0.3829, 0.9486, 0.9162])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,19 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 2.282747268676758 seconds
Time: 0.2727935314178467 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 4599 -ss 10000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.21649432182312}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 3849 -ss 10000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 8.942286252975464}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 18, ..., 99975, 99989,
tensor(crow_indices=tensor([ 0, 12, 21, ..., 99991, 99998,
100000]),
col_indices=tensor([5193, 5456, 6247, ..., 5100, 5946, 8330]),
values=tensor([0.7086, 0.0012, 0.4180, ..., 0.5448, 0.8405, 0.8114]),
col_indices=tensor([ 425, 574, 695, ..., 9570, 6024, 9715]),
values=tensor([0.7410, 0.8879, 0.5840, ..., 0.6995, 0.9280, 0.9465]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.0495, 0.0946, 0.7654, ..., 0.8976, 0.3544, 0.9283])
tensor([0.2929, 0.5164, 0.5482, ..., 0.5103, 0.5008, 0.9557])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -36,16 +36,19 @@ Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 10.21649432182312 seconds
Time: 8.942286252975464 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 4519 -ss 10000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.300970077514648}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 18, ..., 99975, 99989,
tensor(crow_indices=tensor([ 0, 7, 16, ..., 99974, 99990,
100000]),
col_indices=tensor([5193, 5456, 6247, ..., 5100, 5946, 8330]),
values=tensor([0.7086, 0.0012, 0.4180, ..., 0.5448, 0.8405, 0.8114]),
col_indices=tensor([ 582, 1691, 2515, ..., 7345, 7996, 8295]),
values=tensor([0.8177, 0.9283, 0.6030, ..., 0.2647, 0.3717, 0.8633]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.0495, 0.0946, 0.7654, ..., 0.8976, 0.3544, 0.9283])
tensor([0.2209, 0.7260, 0.0429, ..., 0.4887, 0.7834, 0.0043])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -53,13 +56,30 @@ Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 10.21649432182312 seconds
Time: 10.300970077514648 seconds
[16.44, 16.28, 16.32, 16.56, 16.56, 16.6, 16.76, 16.76, 16.72, 16.68]
[16.52, 16.48, 16.6, 20.0, 22.08, 24.8, 25.56, 23.6, 23.04, 20.28, 20.28, 20.04, 20.16, 20.2]
14.251931190490723
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 4599, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.21649432182312, 'TIME_S_1KI': 2.2214599525599303, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 276.3690100479126, 'W': 19.391688491473598}
[16.44, 16.28, 16.32, 16.56, 16.56, 16.6, 16.76, 16.76, 16.72, 16.68, 16.4, 16.48, 16.68, 16.36, 16.44, 16.64, 16.64, 16.8, 16.8, 16.76]
298.53999999999996
14.926999999999998
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 4599, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.21649432182312, 'TIME_S_1KI': 2.2214599525599303, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 276.3690100479126, 'W': 19.391688491473598, 'J_1KI': 60.09328333287945, 'W_1KI': 4.2165010853389, 'W_D': 4.4646884914736, 'J_D': 63.630433167457575, 'W_D_1KI': 0.9707954971675582, 'J_D_1KI': 0.21108838816428752}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 7, 16, ..., 99974, 99990,
100000]),
col_indices=tensor([ 582, 1691, 2515, ..., 7345, 7996, 8295]),
values=tensor([0.8177, 0.9283, 0.6030, ..., 0.2647, 0.3717, 0.8633]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.2209, 0.7260, 0.0429, ..., 0.4887, 0.7834, 0.0043])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 10.300970077514648 seconds
[20.48, 20.64, 20.44, 20.28, 20.28, 20.6, 20.72, 20.68, 20.68, 20.48]
[20.4, 20.36, 23.16, 24.08, 27.0, 27.6, 28.72, 28.72, 26.24, 26.28, 24.44, 24.28, 24.24, 24.08]
14.201388835906982
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 4519, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.300970077514648, 'TIME_S_1KI': 2.279479990598506, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 332.2094431686402, 'W': 23.392743273719635}
[20.48, 20.64, 20.44, 20.28, 20.28, 20.6, 20.72, 20.68, 20.68, 20.48, 20.52, 20.6, 20.68, 20.64, 20.64, 20.52, 20.52, 20.48, 20.68, 20.72]
370.18
18.509
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 4519, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.300970077514648, 'TIME_S_1KI': 2.279479990598506, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 332.2094431686402, 'W': 23.392743273719635, 'J_1KI': 73.51392856132777, 'W_1KI': 5.176530930232272, 'W_D': 4.8837432737196345, 'J_D': 69.35593720483783, 'W_D_1KI': 1.0807132714582062, 'J_D_1KI': 0.2391487655362262}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.01.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 21.366477489471436, "TIME_S_1KI": 21.366477489471436, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 489.4337509441374, "W": 19.31282985940674, "J_1KI": 489.4337509441374, "W_1KI": 19.31282985940674, "W_D": 4.539829859406739, "J_D": 115.05025275492645, "W_D_1KI": 4.539829859406739, "J_D_1KI": 4.539829859406739}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 490, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.49430251121521, "TIME_S_1KI": 21.416943900439204, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 322.6961988353729, "W": 22.678695903983957, "J_1KI": 658.5636710925977, "W_1KI": 46.283052865273376, "W_D": 4.3196959039839555, "J_D": 61.46515012335775, "W_D_1KI": 8.815705926497868, "J_D_1KI": 17.991236584689528}

64
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.01.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 21.366477489471436}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 2.140977382659912}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 93, 201, ..., 999801,
999899, 1000000]),
col_indices=tensor([ 106, 113, 159, ..., 9934, 9937, 9966]),
values=tensor([0.1214, 0.4144, 0.1866, ..., 0.5194, 0.7412, 0.0565]),
tensor(crow_indices=tensor([ 0, 103, 212, ..., 999798,
999901, 1000000]),
col_indices=tensor([ 63, 140, 146, ..., 9691, 9771, 9918]),
values=tensor([0.8748, 0.2571, 0.8906, ..., 0.1504, 0.2890, 0.7825]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.4749, 0.5757, 0.5717, ..., 0.5026, 0.5396, 0.1085])
tensor([0.5882, 0.3416, 0.1892, ..., 0.3016, 0.5220, 0.0626])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,16 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 21.366477489471436 seconds
Time: 2.140977382659912 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 490 -ss 10000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.49430251121521}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 93, 201, ..., 999801,
999899, 1000000]),
col_indices=tensor([ 106, 113, 159, ..., 9934, 9937, 9966]),
values=tensor([0.1214, 0.4144, 0.1866, ..., 0.5194, 0.7412, 0.0565]),
tensor(crow_indices=tensor([ 0, 113, 202, ..., 999820,
999916, 1000000]),
col_indices=tensor([ 2, 35, 39, ..., 9519, 9605, 9656]),
values=tensor([0.0992, 0.7724, 0.9238, ..., 0.3639, 0.9758, 0.0697]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.4749, 0.5757, 0.5717, ..., 0.5026, 0.5396, 0.1085])
tensor([0.2199, 0.1288, 0.7757, ..., 0.1449, 0.2950, 0.2928])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -33,13 +36,30 @@ Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 21.366477489471436 seconds
Time: 10.49430251121521 seconds
[16.76, 16.8, 16.72, 16.6, 16.36, 16.12, 16.0, 16.04, 16.28, 16.36]
[16.48, 16.48, 16.36, 17.76, 18.4, 22.04, 22.96, 22.96, 22.68, 21.88, 20.28, 20.28, 20.36, 20.0, 20.0, 19.8, 19.72, 19.84, 19.96, 20.12, 20.32, 20.36, 20.56, 20.72, 20.6]
25.34241509437561
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 21.366477489471436, 'TIME_S_1KI': 21.366477489471436, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 489.4337509441374, 'W': 19.31282985940674}
[16.76, 16.8, 16.72, 16.6, 16.36, 16.12, 16.0, 16.04, 16.28, 16.36, 16.6, 16.56, 16.28, 16.28, 16.28, 16.24, 16.56, 16.68, 16.52, 16.56]
295.46000000000004
14.773000000000001
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 21.366477489471436, 'TIME_S_1KI': 21.366477489471436, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 489.4337509441374, 'W': 19.31282985940674, 'J_1KI': 489.4337509441374, 'W_1KI': 19.31282985940674, 'W_D': 4.539829859406739, 'J_D': 115.05025275492645, 'W_D_1KI': 4.539829859406739, 'J_D_1KI': 4.539829859406739}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 113, 202, ..., 999820,
999916, 1000000]),
col_indices=tensor([ 2, 35, 39, ..., 9519, 9605, 9656]),
values=tensor([0.0992, 0.7724, 0.9238, ..., 0.3639, 0.9758, 0.0697]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.2199, 0.1288, 0.7757, ..., 0.1449, 0.2950, 0.2928])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 10.49430251121521 seconds
[20.68, 20.72, 20.56, 20.52, 20.48, 20.2, 19.92, 19.96, 19.92, 19.92]
[19.96, 20.36, 20.52, 22.04, 24.08, 26.92, 27.96, 27.96, 26.64, 25.04, 24.52, 24.4, 24.4, 24.52]
14.229045629501343
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 490, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.49430251121521, 'TIME_S_1KI': 21.416943900439204, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 322.6961988353729, 'W': 22.678695903983957}
[20.68, 20.72, 20.56, 20.52, 20.48, 20.2, 19.92, 19.96, 19.92, 19.92, 20.44, 20.36, 20.4, 20.52, 20.4, 20.64, 20.76, 20.52, 20.52, 20.52]
367.18
18.359
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 490, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.49430251121521, 'TIME_S_1KI': 21.416943900439204, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 322.6961988353729, 'W': 22.678695903983957, 'J_1KI': 658.5636710925977, 'W_1KI': 46.283052865273376, 'W_D': 4.3196959039839555, 'J_D': 61.46515012335775, 'W_D_1KI': 8.815705926497868, 'J_D_1KI': 17.991236584689528}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 106.56549715995789, "TIME_S_1KI": 106.56549715995789, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2685.1749729537964, "W": 24.108298502680796, "J_1KI": 2685.1749729537964, "W_1KI": 24.108298502680796, "W_D": 5.534298502680798, "J_D": 616.4084881644253, "W_D_1KI": 5.534298502680798, "J_D_1KI": 5.534298502680798}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.575754880905151, "TIME_S_1KI": 105.75754880905151, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 357.48722974777223, "W": 23.357767084816274, "J_1KI": 3574.8722974777224, "W_1KI": 233.57767084816274, "W_D": 4.891767084816273, "J_D": 74.86778412389756, "W_D_1KI": 48.91767084816273, "J_D_1KI": 489.1767084816273}

44
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.05.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 0.05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 106.56549715995789}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 0.05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.575754880905151}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 506, 1012, ..., 4998963,
4999486, 5000000]),
col_indices=tensor([ 12, 18, 20, ..., 9951, 9962, 9995]),
values=tensor([0.8717, 0.8420, 0.2460, ..., 0.1141, 0.5771, 0.8192]),
tensor(crow_indices=tensor([ 0, 505, 1011, ..., 4998999,
4999505, 5000000]),
col_indices=tensor([ 17, 34, 93, ..., 9927, 9945, 9977]),
values=tensor([0.3942, 0.9668, 0.2842, ..., 0.3748, 0.5474, 0.6270]),
size=(10000, 10000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.3519, 0.6034, 0.3549, ..., 0.4924, 0.0253, 0.7056])
tensor([0.3102, 0.7326, 0.1847, ..., 0.2267, 0.2009, 0.0941])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,16 +16,16 @@ Rows: 10000
Size: 100000000
NNZ: 5000000
Density: 0.05
Time: 106.56549715995789 seconds
Time: 10.575754880905151 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 506, 1012, ..., 4998963,
4999486, 5000000]),
col_indices=tensor([ 12, 18, 20, ..., 9951, 9962, 9995]),
values=tensor([0.8717, 0.8420, 0.2460, ..., 0.1141, 0.5771, 0.8192]),
tensor(crow_indices=tensor([ 0, 505, 1011, ..., 4998999,
4999505, 5000000]),
col_indices=tensor([ 17, 34, 93, ..., 9927, 9945, 9977]),
values=tensor([0.3942, 0.9668, 0.2842, ..., 0.3748, 0.5474, 0.6270]),
size=(10000, 10000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.3519, 0.6034, 0.3549, ..., 0.4924, 0.0253, 0.7056])
tensor([0.3102, 0.7326, 0.1847, ..., 0.2267, 0.2009, 0.0941])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -33,13 +33,13 @@ Rows: 10000
Size: 100000000
NNZ: 5000000
Density: 0.05
Time: 106.56549715995789 seconds
Time: 10.575754880905151 seconds
[20.96, 20.76, 20.72, 20.52, 20.44, 20.56, 20.84, 20.84, 20.6, 20.6]
[20.64, 20.64, 21.08, 23.16, 24.52, 27.56, 29.68, 29.72, 28.4, 27.2, 27.2, 24.36, 24.36, 24.36, 24.84, 24.88, 24.76, 24.96, 24.72, 24.4, 24.36, 24.16, 24.12, 24.12, 24.24, 24.64, 24.72, 24.76, 24.88, 24.76, 24.4, 24.4, 24.32, 24.32, 24.24, 24.24, 24.44, 24.36, 24.16, 24.16, 24.2, 24.24, 24.4, 24.68, 24.56, 24.6, 24.52, 24.48, 24.48, 24.24, 24.36, 24.36, 24.44, 24.52, 24.52, 24.32, 24.44, 24.28, 24.04, 23.96, 23.96, 24.04, 24.16, 24.28, 24.36, 24.44, 24.48, 24.56, 24.64, 24.56, 24.48, 24.24, 24.24, 24.24, 24.2, 24.28, 24.36, 24.2, 24.4, 24.12, 24.16, 24.24, 24.48, 24.48, 24.8, 24.8, 24.92, 24.92, 24.96, 24.84, 24.72, 24.88, 24.72, 24.68, 24.48, 24.28, 24.0, 24.04, 24.04, 24.16, 24.32, 24.28, 24.32, 24.32, 24.48, 24.4, 24.68, 24.8, 24.64, 24.48]
111.37969660758972
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 106.56549715995789, 'TIME_S_1KI': 106.56549715995789, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2685.1749729537964, 'W': 24.108298502680796}
[20.96, 20.76, 20.72, 20.52, 20.44, 20.56, 20.84, 20.84, 20.6, 20.6, 20.64, 20.6, 20.6, 20.72, 20.8, 20.6, 20.52, 20.28, 20.68, 20.6]
371.47999999999996
18.573999999999998
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 106.56549715995789, 'TIME_S_1KI': 106.56549715995789, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2685.1749729537964, 'W': 24.108298502680796, 'J_1KI': 2685.1749729537964, 'W_1KI': 24.108298502680796, 'W_D': 5.534298502680798, 'J_D': 616.4084881644253, 'W_D_1KI': 5.534298502680798, 'J_D_1KI': 5.534298502680798}
[20.44, 20.44, 20.44, 20.52, 20.44, 20.4, 20.32, 20.2, 20.08, 20.32]
[20.32, 20.16, 21.08, 22.56, 24.44, 27.6, 27.6, 29.28, 28.88, 28.12, 25.16, 24.16, 24.12, 24.4, 24.56]
15.30485463142395
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.575754880905151, 'TIME_S_1KI': 105.75754880905151, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 357.48722974777223, 'W': 23.357767084816274}
[20.44, 20.44, 20.44, 20.52, 20.44, 20.4, 20.32, 20.2, 20.08, 20.32, 20.52, 20.56, 20.76, 20.76, 20.84, 20.84, 20.72, 20.56, 20.56, 20.48]
369.32
18.466
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.575754880905151, 'TIME_S_1KI': 105.75754880905151, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 357.48722974777223, 'W': 23.357767084816274, 'J_1KI': 3574.8722974777224, 'W_1KI': 233.57767084816274, 'W_D': 4.891767084816273, 'J_D': 74.86778412389756, 'W_D_1KI': 48.91767084816273, 'J_D_1KI': 489.1767084816273}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.1.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 213.33555269241333, "TIME_S_1KI": 213.33555269241333, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 5226.200544624329, "W": 24.133137838334523, "J_1KI": 5226.200544624329, "W_1KI": 24.133137838334523, "W_D": 5.8551378383345245, "J_D": 1267.97123376751, "W_D_1KI": 5.8551378383345245, "J_D_1KI": 5.8551378383345245}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 21.58586049079895, "TIME_S_1KI": 215.8586049079895, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 695.2596657943726, "W": 24.42843676698333, "J_1KI": 6952.596657943726, "W_1KI": 244.2843676698333, "W_D": 6.1854367669833294, "J_D": 176.04420374608034, "W_D_1KI": 61.854367669833294, "J_D_1KI": 618.543676698333}

44
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.1.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 0.1 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 213.33555269241333}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 0.1 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 21.58586049079895}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1053, 2094, ..., 9997974,
9998964, 10000000]),
col_indices=tensor([ 7, 12, 13, ..., 9961, 9986, 9993]),
values=tensor([0.1704, 0.0678, 0.1172, ..., 0.2208, 0.7370, 0.2820]),
tensor(crow_indices=tensor([ 0, 983, 1945, ..., 9997995,
9998975, 10000000]),
col_indices=tensor([ 7, 29, 32, ..., 9986, 9994, 9999]),
values=tensor([0.5805, 0.0545, 0.7779, ..., 0.8799, 0.6314, 0.5149]),
size=(10000, 10000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.7446, 0.0915, 0.8679, ..., 0.5982, 0.0596, 0.6566])
tensor([0.1246, 0.2739, 0.0084, ..., 0.7975, 0.3318, 0.0977])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,16 +16,16 @@ Rows: 10000
Size: 100000000
NNZ: 10000000
Density: 0.1
Time: 213.33555269241333 seconds
Time: 21.58586049079895 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1053, 2094, ..., 9997974,
9998964, 10000000]),
col_indices=tensor([ 7, 12, 13, ..., 9961, 9986, 9993]),
values=tensor([0.1704, 0.0678, 0.1172, ..., 0.2208, 0.7370, 0.2820]),
tensor(crow_indices=tensor([ 0, 983, 1945, ..., 9997995,
9998975, 10000000]),
col_indices=tensor([ 7, 29, 32, ..., 9986, 9994, 9999]),
values=tensor([0.5805, 0.0545, 0.7779, ..., 0.8799, 0.6314, 0.5149]),
size=(10000, 10000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.7446, 0.0915, 0.8679, ..., 0.5982, 0.0596, 0.6566])
tensor([0.1246, 0.2739, 0.0084, ..., 0.7975, 0.3318, 0.0977])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -33,13 +33,13 @@ Rows: 10000
Size: 100000000
NNZ: 10000000
Density: 0.1
Time: 213.33555269241333 seconds
Time: 21.58586049079895 seconds
[20.4, 20.4, 20.4, 20.36, 20.36, 20.16, 20.32, 20.2, 20.2, 20.28]
[20.32, 20.48, 21.4, 22.68, 24.52, 24.52, 27.12, 28.4, 28.88, 28.48, 27.56, 25.92, 24.36, 24.48, 24.52, 24.56, 24.56, 24.24, 24.24, 24.2, 23.96, 23.92, 24.16, 24.36, 24.48, 24.52, 24.56, 24.56, 24.36, 24.28, 24.28, 24.48, 24.36, 24.36, 24.6, 24.36, 24.24, 24.2, 24.24, 24.12, 24.16, 24.12, 24.24, 24.24, 24.36, 24.48, 24.48, 24.56, 24.4, 24.4, 24.24, 24.56, 24.84, 24.76, 24.76, 24.76, 24.84, 24.52, 24.64, 24.64, 24.64, 24.64, 24.52, 24.52, 24.52, 24.4, 24.36, 24.36, 24.36, 24.32, 24.24, 24.44, 24.2, 24.08, 24.2, 24.0, 23.92, 23.88, 23.84, 24.08, 24.08, 24.4, 24.6, 24.72, 24.8, 24.68, 24.4, 24.36, 24.24, 24.12, 24.16, 24.2, 24.04, 24.04, 24.0, 24.12, 24.32, 24.36, 24.32, 24.4, 24.16, 24.08, 24.12, 24.32, 24.32, 24.32, 24.44, 24.56, 24.32, 24.36, 24.56, 24.68, 24.44, 24.48, 24.52, 24.44, 24.56, 24.56, 24.68, 24.6, 24.2, 24.6, 24.16, 24.24, 24.32, 24.56, 24.28, 24.48, 24.8, 24.68, 24.68, 24.84, 24.84, 24.84, 24.88, 24.56, 24.84, 24.68, 24.48, 24.76, 24.64, 24.64, 24.64, 24.52, 24.8, 24.68, 24.52, 24.68, 24.6, 24.44, 24.68, 24.76, 24.76, 24.64, 24.6, 24.6, 24.56, 24.44, 24.4, 24.52, 24.48, 24.44, 24.44, 24.32, 24.44, 24.28, 24.44, 24.44, 24.36, 24.16, 24.04, 24.0, 24.12, 24.0, 24.12, 24.4, 24.4, 24.32, 24.28, 24.28, 24.28, 24.2, 24.32, 24.32, 24.4, 24.6, 24.44, 24.56, 24.76, 24.84, 24.72, 24.72, 24.72, 24.56, 24.52, 24.56, 24.64, 24.6, 24.36, 24.44, 24.4, 24.4, 24.56, 24.72, 24.72, 24.6, 24.48, 24.36, 24.24, 24.24, 24.28, 24.44, 24.56, 24.56]
216.55702543258667
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 213.33555269241333, 'TIME_S_1KI': 213.33555269241333, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 5226.200544624329, 'W': 24.133137838334523}
[20.4, 20.4, 20.4, 20.36, 20.36, 20.16, 20.32, 20.2, 20.2, 20.28, 20.2, 19.96, 20.08, 20.24, 20.24, 20.28, 20.48, 20.6, 20.64, 20.4]
365.55999999999995
18.278
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 213.33555269241333, 'TIME_S_1KI': 213.33555269241333, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 5226.200544624329, 'W': 24.133137838334523, 'J_1KI': 5226.200544624329, 'W_1KI': 24.133137838334523, 'W_D': 5.8551378383345245, 'J_D': 1267.97123376751, 'W_D_1KI': 5.8551378383345245, 'J_D_1KI': 5.8551378383345245}
[20.44, 20.24, 20.32, 20.24, 20.16, 20.52, 20.56, 20.68, 20.88, 21.04]
[20.8, 20.6, 23.72, 23.72, 26.04, 27.56, 30.88, 32.68, 29.8, 29.08, 27.52, 26.28, 24.44, 24.48, 24.48, 24.2, 24.2, 24.2, 24.12, 24.12, 24.28, 24.28, 24.4, 24.32, 24.16, 24.0, 24.08, 24.16]
28.461078882217407
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 21.58586049079895, 'TIME_S_1KI': 215.8586049079895, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 695.2596657943726, 'W': 24.42843676698333}
[20.44, 20.24, 20.32, 20.24, 20.16, 20.52, 20.56, 20.68, 20.88, 21.04, 19.8, 19.8, 19.8, 19.96, 20.0, 20.08, 20.12, 20.36, 20.32, 20.36]
364.86
18.243000000000002
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 21.58586049079895, 'TIME_S_1KI': 215.8586049079895, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 695.2596657943726, 'W': 24.42843676698333, 'J_1KI': 6952.596657943726, 'W_1KI': 244.2843676698333, 'W_D': 6.1854367669833294, 'J_D': 176.04420374608034, 'W_D_1KI': 61.854367669833294, 'J_D_1KI': 618.543676698333}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.2.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 424.4943735599518, "TIME_S_1KI": 424.4943735599518, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 10473.538437499996, "W": 24.301530579701517, "J_1KI": 10473.538437499996, "W_1KI": 24.301530579701517, "W_D": 5.865530579701517, "J_D": 2527.942006835934, "W_D_1KI": 5.865530579701517, "J_D_1KI": 5.865530579701517}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 42.006776571273804, "TIME_S_1KI": 420.06776571273804, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1323.3857918548583, "W": 24.65839535462019, "J_1KI": 13233.857918548583, "W_1KI": 246.58395354620188, "W_D": 6.256395354620192, "J_D": 335.77305422592167, "W_D_1KI": 62.56395354620192, "J_D_1KI": 625.6395354620192}

44
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.2.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 0.2 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 424.4943735599518}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 0.2 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 42.006776571273804}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1949, 3966, ..., 19996026,
19998069, 20000000]),
col_indices=tensor([ 7, 21, 26, ..., 9986, 9990, 9999]),
values=tensor([0.4961, 0.8613, 0.9281, ..., 0.1556, 0.7430, 0.8560]),
tensor(crow_indices=tensor([ 0, 1950, 3929, ..., 19995954,
19997973, 20000000]),
col_indices=tensor([ 0, 10, 17, ..., 9977, 9980, 9990]),
values=tensor([0.1470, 0.8510, 0.9446, ..., 0.3735, 0.6466, 0.3885]),
size=(10000, 10000), nnz=20000000, layout=torch.sparse_csr)
tensor([0.5082, 0.3131, 0.5448, ..., 0.5922, 0.3726, 0.5476])
tensor([0.1499, 0.7404, 0.4886, ..., 0.1182, 0.4158, 0.3615])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,16 +16,16 @@ Rows: 10000
Size: 100000000
NNZ: 20000000
Density: 0.2
Time: 424.4943735599518 seconds
Time: 42.006776571273804 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1949, 3966, ..., 19996026,
19998069, 20000000]),
col_indices=tensor([ 7, 21, 26, ..., 9986, 9990, 9999]),
values=tensor([0.4961, 0.8613, 0.9281, ..., 0.1556, 0.7430, 0.8560]),
tensor(crow_indices=tensor([ 0, 1950, 3929, ..., 19995954,
19997973, 20000000]),
col_indices=tensor([ 0, 10, 17, ..., 9977, 9980, 9990]),
values=tensor([0.1470, 0.8510, 0.9446, ..., 0.3735, 0.6466, 0.3885]),
size=(10000, 10000), nnz=20000000, layout=torch.sparse_csr)
tensor([0.5082, 0.3131, 0.5448, ..., 0.5922, 0.3726, 0.5476])
tensor([0.1499, 0.7404, 0.4886, ..., 0.1182, 0.4158, 0.3615])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -33,13 +33,13 @@ Rows: 10000
Size: 100000000
NNZ: 20000000
Density: 0.2
Time: 424.4943735599518 seconds
Time: 42.006776571273804 seconds
[20.56, 20.4, 20.48, 20.48, 20.56, 20.8, 20.96, 20.68, 20.68, 20.52]
[20.4, 20.4, 20.84, 23.24, 23.96, 25.96, 29.08, 30.08, 29.8, 30.04, 30.84, 27.8, 27.8, 27.2, 26.24, 24.6, 24.44, 24.32, 24.24, 24.32, 24.2, 24.44, 24.6, 24.6, 24.72, 24.72, 24.84, 24.88, 24.92, 24.96, 24.6, 24.48, 24.36, 24.16, 24.36, 24.64, 24.68, 24.68, 24.72, 24.68, 24.52, 24.52, 24.4, 24.44, 24.24, 24.36, 24.4, 24.48, 24.44, 24.44, 24.44, 24.24, 24.28, 24.24, 24.44, 24.4, 24.28, 24.24, 24.28, 24.24, 24.48, 24.84, 24.84, 24.56, 24.64, 24.6, 24.88, 24.68, 24.56, 24.56, 24.4, 24.4, 24.32, 24.36, 24.44, 24.44, 24.36, 24.48, 24.64, 24.8, 24.68, 24.92, 24.88, 24.76, 24.88, 24.68, 24.36, 24.36, 24.4, 24.44, 24.68, 24.76, 24.56, 24.64, 24.24, 24.32, 24.72, 24.6, 24.96, 24.96, 25.12, 25.04, 24.44, 24.36, 24.28, 24.16, 24.4, 24.44, 24.56, 24.68, 24.72, 24.48, 24.48, 24.4, 24.36, 24.24, 24.2, 24.52, 24.64, 24.84, 24.8, 24.56, 24.28, 24.16, 24.16, 24.16, 24.28, 24.4, 24.6, 24.64, 24.52, 24.56, 24.44, 24.36, 24.56, 24.84, 24.64, 24.64, 24.84, 24.72, 24.48, 24.6, 24.64, 24.68, 24.6, 24.4, 24.4, 24.16, 24.12, 24.12, 24.28, 24.4, 24.48, 24.76, 24.6, 24.44, 24.28, 24.32, 24.24, 24.32, 24.4, 24.4, 24.56, 24.48, 24.48, 24.28, 24.68, 24.52, 24.32, 24.4, 24.56, 24.28, 24.4, 24.72, 24.72, 24.76, 24.76, 24.72, 24.64, 24.56, 24.6, 24.56, 24.64, 24.48, 24.44, 24.4, 24.4, 24.36, 24.32, 24.48, 24.48, 24.48, 24.32, 24.4, 24.08, 24.16, 24.28, 24.44, 24.48, 24.48, 24.6, 24.56, 24.44, 24.4, 24.44, 24.44, 24.64, 24.72, 24.44, 24.48, 24.32, 24.32, 24.4, 24.28, 24.52, 24.56, 24.68, 24.68, 24.72, 24.88, 24.96, 25.04, 24.84, 24.84, 24.84, 24.6, 24.64, 24.6, 24.6, 24.6, 24.36, 24.28, 24.32, 24.48, 24.48, 24.44, 24.44, 24.56, 24.56, 24.84, 24.72, 24.8, 24.72, 24.6, 24.56, 24.68, 24.88, 24.88, 24.88, 24.88, 24.48, 24.48, 24.32, 24.32, 24.36, 24.36, 24.44, 24.32, 24.36, 24.44, 24.32, 24.32, 24.28, 24.64, 24.8, 24.8, 24.96, 24.76, 24.8, 24.68, 24.52, 24.64, 24.48, 24.48, 24.56, 24.48, 24.32, 24.28, 24.32, 24.36, 24.28, 24.36, 24.4, 24.28, 24.36, 24.36, 24.32, 24.16, 24.04, 24.2, 24.2, 24.32, 24.36, 24.52, 24.36, 24.44, 24.56, 24.64, 24.64, 24.84, 24.8, 24.8, 24.64, 24.48, 24.52, 24.4, 24.6, 24.32, 24.24, 24.2, 24.2, 24.16, 24.2, 24.56, 24.56, 24.68, 24.8, 24.8, 24.68, 24.64, 24.68, 24.6, 24.64, 24.64, 24.6, 24.6, 24.4, 24.28, 24.36, 24.4, 24.48, 24.44, 24.48, 24.52, 24.36, 24.36, 24.44, 24.32, 24.24, 24.44, 24.44, 24.68, 24.56, 24.4, 24.52, 24.4, 24.24, 24.24, 24.4, 24.44, 24.48, 24.6, 24.76, 24.76, 24.64, 24.72, 24.44, 24.48, 24.6, 24.56, 24.56, 24.64, 24.72, 24.56, 24.4, 24.24, 24.2, 24.16, 24.08, 24.04, 24.12, 24.32, 24.32, 24.36, 24.36, 24.6, 24.56, 24.52, 24.76, 24.6, 24.68, 24.56, 24.76, 24.8, 24.76, 24.76, 24.88, 24.64, 24.76, 24.52, 24.44, 24.2, 24.68, 24.48, 24.84, 25.0, 25.12, 25.12, 24.96, 24.84, 24.52, 24.28, 24.2, 24.48, 24.48, 24.52, 24.52, 24.56, 24.6, 24.6, 24.76, 24.76, 25.0, 24.88, 24.92, 24.88, 24.64, 24.72, 24.64, 24.68, 24.68, 24.56, 24.56, 24.36, 24.28, 24.32]
430.982666015625
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 424.4943735599518, 'TIME_S_1KI': 424.4943735599518, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 10473.538437499996, 'W': 24.301530579701517}
[20.56, 20.4, 20.48, 20.48, 20.56, 20.8, 20.96, 20.68, 20.68, 20.52, 20.2, 20.36, 20.48, 20.56, 20.28, 20.2, 20.2, 20.24, 20.52, 20.4]
368.72
18.436
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 424.4943735599518, 'TIME_S_1KI': 424.4943735599518, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 10473.538437499996, 'W': 24.301530579701517, 'J_1KI': 10473.538437499996, 'W_1KI': 24.301530579701517, 'W_D': 5.865530579701517, 'J_D': 2527.942006835934, 'W_D_1KI': 5.865530579701517, 'J_D_1KI': 5.865530579701517}
[20.44, 20.28, 20.36, 20.28, 20.32, 20.16, 20.32, 20.32, 20.52, 20.56]
[20.6, 20.56, 20.76, 24.64, 26.0, 26.92, 30.76, 29.04, 28.68, 29.52, 30.24, 30.24, 27.76, 27.48, 26.4, 24.64, 24.48, 24.32, 24.28, 24.28, 24.28, 24.08, 24.28, 24.28, 24.36, 24.2, 24.24, 24.6, 24.64, 24.52, 24.72, 24.48, 24.64, 24.4, 24.52, 24.52, 24.36, 24.4, 24.4, 24.4, 24.48, 24.68, 24.76, 24.56, 24.36, 24.16, 24.24, 24.4, 24.4, 24.76, 24.88, 24.96, 24.92]
53.668771743774414
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 42.006776571273804, 'TIME_S_1KI': 420.06776571273804, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1323.3857918548583, 'W': 24.65839535462019}
[20.44, 20.28, 20.36, 20.28, 20.32, 20.16, 20.32, 20.32, 20.52, 20.56, 20.12, 19.92, 20.0, 20.28, 20.44, 20.72, 20.96, 21.04, 21.04, 21.04]
368.03999999999996
18.401999999999997
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 42.006776571273804, 'TIME_S_1KI': 420.06776571273804, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1323.3857918548583, 'W': 24.65839535462019, 'J_1KI': 13233.857918548583, 'W_1KI': 246.58395354620188, 'W_D': 6.256395354620192, 'J_D': 335.77305422592167, 'W_D_1KI': 62.56395354620192, 'J_D_1KI': 625.6395354620192}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.3.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 637.8268127441406, "TIME_S_1KI": 637.8268127441406, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 15996.775521488189, "W": 24.370595996658984, "J_1KI": 15996.775521488189, "W_1KI": 24.370595996658984, "W_D": 5.917595996658985, "J_D": 3884.2896906743035, "W_D_1KI": 5.917595996658985, "J_D_1KI": 5.917595996658985}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 67.20304369926453, "TIME_S_1KI": 672.0304369926453, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1846.4787199783327, "W": 24.29422238106021, "J_1KI": 18464.787199783328, "W_1KI": 242.9422238106021, "W_D": 5.818222381060206, "J_D": 442.2131174325942, "W_D_1KI": 58.18222381060206, "J_D_1KI": 581.8222381060207}

44
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_0.3.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 0.3 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 637.8268127441406}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 0.3 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 67.20304369926453}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2961, 5942, ..., 29993896,
29996981, 30000000]),
col_indices=tensor([ 2, 5, 14, ..., 9994, 9995, 9996]),
values=tensor([0.0155, 0.6045, 0.5805, ..., 0.2136, 0.4050, 0.0107]),
tensor(crow_indices=tensor([ 0, 2926, 5920, ..., 29993999,
29997022, 30000000]),
col_indices=tensor([ 1, 4, 6, ..., 9978, 9982, 9992]),
values=tensor([0.3929, 0.6592, 0.7367, ..., 0.3321, 0.3012, 0.1502]),
size=(10000, 10000), nnz=30000000, layout=torch.sparse_csr)
tensor([0.2173, 0.9101, 0.0107, ..., 0.2401, 0.6161, 0.6478])
tensor([0.6782, 0.5388, 0.0901, ..., 0.7339, 0.4235, 0.1483])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,16 +16,16 @@ Rows: 10000
Size: 100000000
NNZ: 30000000
Density: 0.3
Time: 637.8268127441406 seconds
Time: 67.20304369926453 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2961, 5942, ..., 29993896,
29996981, 30000000]),
col_indices=tensor([ 2, 5, 14, ..., 9994, 9995, 9996]),
values=tensor([0.0155, 0.6045, 0.5805, ..., 0.2136, 0.4050, 0.0107]),
tensor(crow_indices=tensor([ 0, 2926, 5920, ..., 29993999,
29997022, 30000000]),
col_indices=tensor([ 1, 4, 6, ..., 9978, 9982, 9992]),
values=tensor([0.3929, 0.6592, 0.7367, ..., 0.3321, 0.3012, 0.1502]),
size=(10000, 10000), nnz=30000000, layout=torch.sparse_csr)
tensor([0.2173, 0.9101, 0.0107, ..., 0.2401, 0.6161, 0.6478])
tensor([0.6782, 0.5388, 0.0901, ..., 0.7339, 0.4235, 0.1483])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -33,13 +33,13 @@ Rows: 10000
Size: 100000000
NNZ: 30000000
Density: 0.3
Time: 637.8268127441406 seconds
Time: 67.20304369926453 seconds
[20.6, 20.64, 20.76, 20.6, 20.44, 20.44, 20.48, 20.64, 20.64, 20.88]
[20.96, 21.0, 21.0, 25.76, 26.8, 28.52, 29.6, 31.28, 31.2, 30.36, 31.16, 30.84, 30.84, 29.04, 28.4, 28.24, 27.2, 26.12, 24.68, 24.6, 24.44, 24.4, 24.4, 24.64, 24.72, 24.72, 24.96, 24.92, 24.64, 24.76, 24.76, 25.12, 25.12, 25.2, 24.92, 24.68, 24.48, 24.48, 24.6, 24.52, 24.44, 24.6, 24.48, 24.28, 24.24, 24.2, 24.32, 24.48, 24.6, 24.6, 24.56, 24.48, 24.4, 24.24, 24.28, 24.52, 24.4, 24.56, 24.56, 24.32, 24.48, 24.48, 24.48, 24.4, 24.64, 24.84, 24.76, 24.76, 24.8, 24.64, 24.56, 24.52, 24.28, 24.4, 24.4, 24.48, 24.72, 24.76, 24.68, 24.76, 24.76, 24.64, 24.64, 24.36, 24.2, 24.36, 24.36, 24.28, 24.48, 24.6, 24.4, 24.24, 24.4, 24.32, 24.36, 24.4, 24.52, 24.12, 24.36, 24.36, 24.6, 24.72, 24.56, 24.92, 24.48, 24.56, 24.56, 24.56, 24.16, 24.24, 24.24, 24.24, 24.36, 24.4, 24.56, 24.52, 24.72, 24.68, 24.8, 24.8, 24.76, 24.76, 24.64, 24.64, 24.64, 24.68, 24.52, 24.48, 24.56, 24.4, 24.6, 24.56, 24.6, 24.76, 24.76, 24.84, 24.84, 25.0, 24.92, 24.68, 24.72, 24.68, 24.84, 24.84, 24.72, 24.8, 24.64, 24.52, 24.52, 24.76, 24.8, 24.64, 24.6, 24.6, 24.56, 24.8, 25.0, 24.92, 24.8, 24.76, 24.52, 24.52, 24.44, 24.64, 24.76, 24.76, 24.64, 24.36, 24.36, 24.52, 24.56, 24.72, 24.96, 24.96, 24.76, 24.72, 24.84, 24.68, 24.52, 24.76, 24.68, 24.44, 24.52, 24.48, 24.52, 24.92, 24.92, 24.76, 24.52, 24.8, 24.68, 24.52, 24.72, 24.92, 24.6, 24.48, 24.44, 24.52, 24.52, 24.24, 24.4, 24.4, 24.4, 24.24, 24.28, 24.48, 24.44, 24.48, 24.8, 24.64, 24.64, 24.72, 24.8, 24.64, 24.52, 24.48, 24.28, 24.24, 24.2, 24.36, 24.4, 24.44, 24.44, 24.44, 24.72, 24.72, 24.76, 24.72, 24.8, 24.48, 24.24, 24.36, 24.36, 24.4, 24.64, 24.64, 24.76, 24.64, 24.56, 24.24, 24.32, 24.32, 24.4, 24.64, 24.76, 24.76, 24.92, 24.92, 24.96, 25.04, 24.76, 25.0, 24.8, 24.76, 24.72, 24.52, 24.52, 24.44, 24.6, 24.6, 24.6, 24.6, 24.64, 24.68, 24.68, 24.84, 25.0, 25.0, 24.88, 24.72, 24.56, 24.4, 24.4, 24.48, 24.52, 24.56, 24.44, 24.64, 24.56, 24.44, 24.36, 24.56, 24.72, 24.64, 24.92, 24.92, 24.84, 24.64, 24.68, 24.64, 24.68, 24.68, 24.8, 24.88, 24.88, 24.76, 24.96, 24.96, 24.76, 24.8, 24.64, 24.84, 24.68, 24.56, 24.72, 24.56, 24.56, 24.44, 24.28, 24.28, 24.36, 24.2, 24.04, 24.2, 24.32, 24.32, 24.52, 24.52, 24.4, 24.48, 24.72, 24.68, 24.68, 24.6, 24.68, 24.52, 24.52, 24.72, 24.64, 24.68, 24.8, 24.48, 24.6, 24.84, 24.84, 24.56, 24.56, 24.28, 24.24, 24.2, 24.32, 24.32, 24.16, 24.28, 24.44, 24.4, 24.4, 24.68, 24.84, 24.84, 24.76, 24.64, 24.6, 24.4, 24.56, 24.4, 24.4, 24.68, 24.8, 24.8, 24.72, 24.72, 24.56, 24.36, 24.48, 24.64, 24.72, 24.84, 24.88, 24.64, 24.6, 24.6, 24.56, 24.44, 24.44, 24.56, 24.6, 24.64, 24.48, 24.56, 24.28, 24.36, 24.44, 24.28, 24.28, 24.16, 24.24, 24.4, 24.36, 24.6, 24.64, 24.6, 24.6, 24.72, 24.84, 24.76, 24.76, 24.68, 24.68, 24.6, 24.52, 24.52, 24.48, 24.36, 24.32, 24.24, 24.2, 24.4, 24.4, 24.48, 24.52, 24.72, 24.68, 24.88, 24.88, 24.8, 24.92, 24.84, 24.76, 24.6, 24.44, 24.44, 24.4, 24.28, 24.24, 24.32, 24.16, 24.4, 24.52, 24.4, 24.28, 24.28, 24.2, 24.2, 24.08, 24.16, 24.28, 24.32, 24.4, 24.28, 24.24, 24.04, 24.04, 24.12, 24.16, 24.16, 24.4, 24.4, 24.48, 24.6, 24.76, 24.68, 24.6, 24.64, 24.48, 24.36, 24.36, 24.44, 24.44, 24.48, 24.72, 24.88, 24.64, 24.8, 24.64, 24.64, 24.6, 24.48, 24.48, 24.64, 24.64, 24.88, 24.64, 24.48, 24.36, 24.52, 24.48, 24.68, 24.88, 24.8, 24.92, 24.88, 24.88, 24.88, 25.04, 24.64, 24.8, 24.92, 24.8, 24.72, 24.72, 24.6, 24.4, 24.44, 24.52, 24.52, 24.36, 24.44, 24.6, 24.56, 24.48, 24.8, 24.72, 24.8, 24.72, 24.92, 24.68, 24.68, 24.52, 24.52, 24.48, 24.44, 24.36, 24.4, 24.36, 24.44, 24.4, 24.4, 24.32, 24.24, 24.24, 24.2, 24.12, 24.2, 24.36, 24.36, 24.52, 24.52, 24.4, 24.24, 24.24, 24.32, 24.32, 24.16, 24.2, 24.24, 24.28, 24.24, 24.44, 24.6, 24.44, 24.32, 24.4, 24.48, 24.48, 24.36, 24.68, 24.6, 24.52, 24.52, 24.52, 24.68, 24.48, 24.56, 24.44, 24.32, 24.28, 24.28, 24.24, 24.44, 24.4, 24.36, 24.24, 24.36, 24.72, 24.8, 24.64, 24.68, 24.44, 24.44, 24.44, 24.52, 24.68, 24.68, 24.84, 24.76, 24.96, 24.92, 24.92, 25.04, 25.48, 25.48, 25.16, 25.16, 24.76, 24.24, 24.32, 24.2, 24.32, 24.36, 24.4, 24.44, 24.56, 24.72, 24.72, 24.72, 24.64, 24.76, 24.6, 24.28, 24.12, 24.2, 24.24, 24.44, 25.04, 25.12, 25.12, 25.04, 25.08, 24.96, 24.64, 24.68, 24.88, 24.96, 25.12, 25.16, 25.16, 24.88, 25.08, 25.08, 25.12, 25.08, 24.84, 24.76, 24.56, 24.44, 24.68, 24.84, 24.64, 24.84, 24.76, 24.76, 24.56, 24.56, 24.8, 24.96, 24.96, 25.08, 24.92, 24.8, 24.76, 24.84, 24.76, 24.76, 24.8, 24.84, 24.72, 24.52, 24.6]
656.396565914154
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 637.8268127441406, 'TIME_S_1KI': 637.8268127441406, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 15996.775521488189, 'W': 24.370595996658984}
[20.6, 20.64, 20.76, 20.6, 20.44, 20.44, 20.48, 20.64, 20.64, 20.88, 20.28, 20.44, 20.44, 20.64, 20.68, 20.44, 20.36, 20.36, 20.12, 20.12]
369.06
18.453
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 637.8268127441406, 'TIME_S_1KI': 637.8268127441406, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 15996.775521488189, 'W': 24.370595996658984, 'J_1KI': 15996.775521488189, 'W_1KI': 24.370595996658984, 'W_D': 5.917595996658985, 'J_D': 3884.2896906743035, 'W_D_1KI': 5.917595996658985, 'J_D_1KI': 5.917595996658985}
[20.6, 20.8, 20.76, 20.64, 20.32, 20.48, 20.48, 20.52, 20.8, 20.6]
[20.64, 20.6, 20.6, 21.32, 22.6, 24.24, 25.68, 28.84, 30.4, 30.64, 31.4, 31.32, 28.4, 28.16, 27.72, 26.8, 26.8, 25.96, 24.6, 24.4, 24.12, 24.12, 24.12, 24.08, 24.36, 24.56, 24.8, 24.88, 24.88, 24.92, 24.92, 24.72, 24.64, 24.44, 24.28, 24.52, 24.68, 24.64, 24.68, 24.64, 24.64, 24.52, 24.76, 24.84, 24.68, 24.72, 24.68, 24.68, 24.76, 24.68, 24.52, 24.2, 24.12, 24.12, 24.24, 24.48, 24.64, 24.76, 24.76, 24.52, 24.28, 24.32, 24.12, 24.04, 24.32, 24.32, 24.6, 24.52, 24.76, 24.72, 24.36, 24.12, 24.16, 24.36, 24.48]
76.0048496723175
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 67.20304369926453, 'TIME_S_1KI': 672.0304369926453, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1846.4787199783327, 'W': 24.29422238106021}
[20.6, 20.8, 20.76, 20.64, 20.32, 20.48, 20.48, 20.52, 20.8, 20.6, 20.04, 20.16, 20.28, 20.16, 20.16, 20.64, 20.72, 20.72, 20.92, 20.68]
369.52000000000004
18.476000000000003
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 67.20304369926453, 'TIME_S_1KI': 672.0304369926453, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1846.4787199783327, 'W': 24.29422238106021, 'J_1KI': 18464.787199783328, 'W_1KI': 242.9422238106021, 'W_D': 5.818222381060206, 'J_D': 442.2131174325942, 'W_D_1KI': 58.18222381060206, 'J_D_1KI': 581.8222381060207}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 141920, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.197850227355957, "TIME_S_1KI": 0.0718563291104563, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 278.53162281036373, "W": 19.575620643059622, "J_1KI": 1.9625959893627658, "W_1KI": 0.1379341928062262, "W_D": 4.584620643059623, "J_D": 65.2322524514198, "W_D_1KI": 0.032304260449969154, "J_D_1KI": 0.00022762303022808028}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 145476, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.560847282409668, "TIME_S_1KI": 0.07259511728676668, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 314.8679765701293, "W": 22.125839021878733, "J_1KI": 2.1643980902013342, "W_1KI": 0.15209270960075016, "W_D": 3.6298390218787304, "J_D": 51.65544533538807, "W_D_1KI": 0.024951462934633413, "J_D_1KI": 0.00017151600906426773}

2538
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_1e-05.output
View File

File diff suppressed because it is too large Load Diff

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 52721, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 11.043588876724243, "TIME_S_1KI": 0.20947229522816796, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 310.41147632598876, "W": 21.8334966893401, "J_1KI": 5.8878146530981725, "W_1KI": 0.4141328254270613, "W_D": 3.2964966893401026, "J_D": 46.86699609327319, "W_D_1KI": 0.06252720337892116, "J_D_1KI": 0.0011860018470613448}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 52342, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.44128155708313, "TIME_S_1KI": 0.19948189899283808, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 334.7413140869141, "W": 23.567352524316387, "J_1KI": 6.395271752835469, "W_1KI": 0.4502570120422679, "W_D": 3.4093525243163825, "J_D": 48.42508902931206, "W_D_1KI": 0.06513607665577133, "J_D_1KI": 0.0012444323231013588}

64
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_10000_5e-05.output
View File

@ -1,13 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 10000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.24570083618164062}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 10000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.027875900268554688}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 1, ..., 5000, 5000, 5000]),
col_indices=tensor([7274, 1823, 9481, ..., 3720, 7669, 6157]),
values=tensor([0.0699, 0.4403, 0.9366, ..., 0.7220, 0.3462, 0.9666]),
tensor(crow_indices=tensor([ 0, 0, 0, ..., 5000, 5000, 5000]),
col_indices=tensor([4151, 7566, 61, ..., 1923, 6890, 8738]),
values=tensor([0.6199, 0.1524, 0.8589, ..., 0.4429, 0.5764, 0.1533]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.3652, 0.9468, 0.8818, ..., 0.3143, 0.5478, 0.8274])
tensor([0.5335, 0.6247, 0.4039, ..., 0.6064, 0.4993, 0.4017])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -15,18 +15,18 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 0.24570083618164062 seconds
Time: 0.027875900268554688 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 42734 -ss 10000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 8.510959386825562}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 37666 -ss 10000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 7.555848598480225}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 4999, 4999, 5000]),
col_indices=tensor([3889, 8009, 975, ..., 383, 3476, 3024]),
values=tensor([0.2888, 0.9236, 0.0703, ..., 0.2234, 0.4670, 0.5913]),
tensor(crow_indices=tensor([ 0, 0, 2, ..., 4999, 5000, 5000]),
col_indices=tensor([4832, 7617, 3198, ..., 2337, 8239, 2535]),
values=tensor([0.0012, 0.2497, 0.5477, ..., 0.0331, 0.3343, 0.4565]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.8206, 0.5304, 0.1258, ..., 0.8056, 0.8493, 0.1547])
tensor([0.1414, 0.6293, 0.2915, ..., 0.6179, 0.0556, 0.9688])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -34,18 +34,18 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 8.510959386825562 seconds
Time: 7.555848598480225 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 52721 -ss 10000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 11.043588876724243}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 52342 -ss 10000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.44128155708313}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 4999, 5000, 5000]),
col_indices=tensor([3785, 9007, 8216, ..., 3831, 7720, 5051]),
values=tensor([0.5982, 0.6190, 0.5749, ..., 0.7670, 0.1287, 0.1052]),
col_indices=tensor([6929, 6481, 6208, ..., 5185, 5914, 4436]),
values=tensor([0.2292, 0.3731, 0.2148, ..., 0.4978, 0.6385, 0.1071]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.6986, 0.7258, 0.0612, ..., 0.6419, 0.7078, 0.3008])
tensor([0.7171, 0.2412, 0.9457, ..., 0.4356, 0.0163, 0.8101])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -53,15 +53,15 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 11.043588876724243 seconds
Time: 10.44128155708313 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 4999, 5000, 5000]),
col_indices=tensor([3785, 9007, 8216, ..., 3831, 7720, 5051]),
values=tensor([0.5982, 0.6190, 0.5749, ..., 0.7670, 0.1287, 0.1052]),
col_indices=tensor([6929, 6481, 6208, ..., 5185, 5914, 4436]),
values=tensor([0.2292, 0.3731, 0.2148, ..., 0.4978, 0.6385, 0.1071]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.6986, 0.7258, 0.0612, ..., 0.6419, 0.7078, 0.3008])
tensor([0.7171, 0.2412, 0.9457, ..., 0.4356, 0.0163, 0.8101])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -69,13 +69,13 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 11.043588876724243 seconds
Time: 10.44128155708313 seconds
[20.64, 20.64, 20.64, 20.64, 20.48, 20.52, 20.4, 20.52, 20.6, 20.56]
[20.56, 20.6, 20.72, 22.44, 23.64, 25.6, 26.32, 25.84, 25.0, 23.2, 23.28, 23.24, 23.44, 23.44]
14.217213153839111
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 52721, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 11.043588876724243, 'TIME_S_1KI': 0.20947229522816796, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 310.41147632598876, 'W': 21.8334966893401}
[20.64, 20.64, 20.64, 20.64, 20.48, 20.52, 20.4, 20.52, 20.6, 20.56, 20.36, 20.48, 20.48, 20.44, 20.56, 20.68, 20.88, 20.96, 20.72, 20.64]
370.74
18.537
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 52721, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 11.043588876724243, 'TIME_S_1KI': 0.20947229522816796, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 310.41147632598876, 'W': 21.8334966893401, 'J_1KI': 5.8878146530981725, 'W_1KI': 0.4141328254270613, 'W_D': 3.2964966893401026, 'J_D': 46.86699609327319, 'W_D_1KI': 0.06252720337892116, 'J_D_1KI': 0.0011860018470613448}
[25.48, 25.4, 25.4, 24.8, 24.68, 24.36, 24.4, 24.16, 24.24, 23.68]
[23.36, 22.52, 25.48, 26.0, 27.84, 27.84, 27.8, 28.32, 24.88, 24.76, 23.72, 23.52, 23.72, 23.68]
14.20360279083252
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 52342, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.44128155708313, 'TIME_S_1KI': 0.19948189899283808, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 334.7413140869141, 'W': 23.567352524316387}
[25.48, 25.4, 25.4, 24.8, 24.68, 24.36, 24.4, 24.16, 24.24, 23.68, 20.0, 19.92, 19.76, 20.2, 20.12, 20.12, 20.36, 20.24, 20.24, 20.36]
403.1600000000001
20.158000000000005
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 52342, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.44128155708313, 'TIME_S_1KI': 0.19948189899283808, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 334.7413140869141, 'W': 23.567352524316387, 'J_1KI': 6.395271752835469, 'W_1KI': 0.4502570120422679, 'W_D': 3.4093525243163825, 'J_D': 48.42508902931206, 'W_D_1KI': 0.06513607665577133, 'J_D_1KI': 0.0012444323231013588}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_500000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 78.25872588157654, "TIME_S_1KI": 78.25872588157654, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 4354.838327121737, "W": 46.22946318790178, "J_1KI": 4354.838327121737, "W_1KI": 46.22946318790178, "W_D": 27.22746318790178, "J_D": 2564.8405165128734, "W_D_1KI": 27.22746318790178, "J_D_1KI": 27.22746318790178}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 137, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.396055936813354, "TIME_S_1KI": 75.88361997673981, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 753.5857475948335, "W": 33.1694657622459, "J_1KI": 5500.625894852799, "W_1KI": 242.11288877551752, "W_D": 14.2684657622459, "J_D": 324.16899673771866, "W_D_1KI": 104.14938512588249, "J_D_1KI": 760.2144899699452}

69
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_500000_0.0001.output
View File

@ -1,15 +1,15 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 500000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 78.25872588157654}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 500000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 7.629654169082642}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 62, 108, ..., 24999902,
24999957, 25000000]),
col_indices=tensor([ 8122, 13146, 17492, ..., 475877, 485043,
496973]),
values=tensor([0.3209, 0.8894, 0.3965, ..., 0.0269, 0.4047, 0.8747]),
tensor(crow_indices=tensor([ 0, 44, 92, ..., 24999905,
24999955, 25000000]),
col_indices=tensor([ 2191, 6192, 41052, ..., 471066, 488040,
493296]),
values=tensor([0.3986, 0.5227, 0.3241, ..., 0.9261, 0.7192, 0.3287]),
size=(500000, 500000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.9161, 0.4114, 0.3584, ..., 0.3416, 0.2120, 0.9282])
tensor([0.0957, 0.3468, 0.1431, ..., 0.5849, 0.2942, 0.3782])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -17,17 +17,20 @@ Rows: 500000
Size: 250000000000
NNZ: 25000000
Density: 0.0001
Time: 78.25872588157654 seconds
Time: 7.629654169082642 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 137 -ss 500000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.396055936813354}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 62, 108, ..., 24999902,
24999957, 25000000]),
col_indices=tensor([ 8122, 13146, 17492, ..., 475877, 485043,
496973]),
values=tensor([0.3209, 0.8894, 0.3965, ..., 0.0269, 0.4047, 0.8747]),
tensor(crow_indices=tensor([ 0, 53, 94, ..., 24999919,
24999951, 25000000]),
col_indices=tensor([ 2485, 12624, 22152, ..., 462150, 467889,
476331]),
values=tensor([0.9572, 0.0985, 0.5455, ..., 0.5648, 0.8530, 0.8208]),
size=(500000, 500000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.9161, 0.4114, 0.3584, ..., 0.3416, 0.2120, 0.9282])
tensor([0.3242, 0.9080, 0.9457, ..., 0.2147, 0.3332, 0.4113])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -35,13 +38,31 @@ Rows: 500000
Size: 250000000000
NNZ: 25000000
Density: 0.0001
Time: 78.25872588157654 seconds
Time: 10.396055936813354 seconds
[21.04, 21.2, 21.24, 21.2, 21.44, 21.2, 21.08, 21.04, 20.84, 20.8]
[20.96, 20.88, 21.6, 21.6, 22.56, 24.32, 26.6, 27.68, 30.32, 31.76, 31.8, 31.36, 31.8, 32.64, 38.12, 43.32, 49.08, 53.24, 53.16, 52.56, 52.56, 52.56, 52.6, 52.28, 52.68, 53.0, 53.04, 53.24, 53.08, 53.48, 53.36, 53.16, 53.16, 52.68, 52.84, 52.88, 53.12, 52.96, 52.88, 53.2, 53.16, 53.0, 53.0, 52.96, 53.12, 52.84, 52.76, 53.08, 52.96, 52.84, 52.96, 52.92, 53.12, 53.2, 53.04, 53.28, 53.16, 53.12, 52.68, 52.96, 52.88, 52.72, 53.04, 53.08, 52.76, 52.76, 52.92, 53.32, 53.28, 53.32, 53.44, 53.28, 53.32, 53.4, 53.48, 53.48, 53.4, 53.4, 53.44, 53.12, 53.32, 53.44, 53.56, 53.52, 53.4, 53.36, 53.12, 53.12, 53.12, 53.16]
94.20049524307251
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 78.25872588157654, 'TIME_S_1KI': 78.25872588157654, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 4354.838327121737, 'W': 46.22946318790178}
[21.04, 21.2, 21.24, 21.2, 21.44, 21.2, 21.08, 21.04, 20.84, 20.8, 21.64, 21.52, 21.28, 21.2, 21.12, 20.8, 20.84, 20.76, 21.0, 21.08]
380.03999999999996
19.002
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 78.25872588157654, 'TIME_S_1KI': 78.25872588157654, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 4354.838327121737, 'W': 46.22946318790178, 'J_1KI': 4354.838327121737, 'W_1KI': 46.22946318790178, 'W_D': 27.22746318790178, 'J_D': 2564.8405165128734, 'W_D_1KI': 27.22746318790178, 'J_D_1KI': 27.22746318790178}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 53, 94, ..., 24999919,
24999951, 25000000]),
col_indices=tensor([ 2485, 12624, 22152, ..., 462150, 467889,
476331]),
values=tensor([0.9572, 0.0985, 0.5455, ..., 0.5648, 0.8530, 0.8208]),
size=(500000, 500000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.3242, 0.9080, 0.9457, ..., 0.2147, 0.3332, 0.4113])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
Rows: 500000
Size: 250000000000
NNZ: 25000000
Density: 0.0001
Time: 10.396055936813354 seconds
[20.64, 20.72, 20.84, 20.96, 21.08, 21.36, 21.56, 21.56, 21.36, 21.36]
[21.36, 21.04, 20.8, 24.16, 25.16, 26.72, 28.8, 27.16, 30.52, 29.84, 29.76, 29.92, 29.92, 30.28, 32.88, 38.36, 44.52, 49.04, 53.12, 53.28, 53.2, 52.84]
22.719260931015015
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 137, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.396055936813354, 'TIME_S_1KI': 75.88361997673981, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 753.5857475948335, 'W': 33.1694657622459}
[20.64, 20.72, 20.84, 20.96, 21.08, 21.36, 21.56, 21.56, 21.36, 21.36, 20.8, 20.84, 20.88, 21.0, 21.04, 20.84, 20.96, 20.68, 20.68, 20.52]
378.02
18.901
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 137, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.396055936813354, 'TIME_S_1KI': 75.88361997673981, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 753.5857475948335, 'W': 33.1694657622459, 'J_1KI': 5500.625894852799, 'W_1KI': 242.11288877551752, 'W_D': 14.2684657622459, 'J_D': 324.16899673771866, 'W_D_1KI': 104.14938512588249, 'J_D_1KI': 760.2144899699452}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_500000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1484, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.88543152809143, "TIME_S_1KI": 7.335196447500964, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 480.2575412559509, "W": 32.95171788766838, "J_1KI": 323.6236800916111, "W_1KI": 22.204661649372223, "W_D": 16.90171788766838, "J_D": 246.33548707246783, "W_D_1KI": 11.389297767970607, "J_D_1KI": 7.67472895415809}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1548, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 11.061279058456421, "TIME_S_1KI": 7.145529107529987, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 576.3272948646545, "W": 36.84956710355697, "J_1KI": 372.3044540469344, "W_1KI": 23.80462991185851, "W_D": 18.03556710355697, "J_D": 282.0763014917373, "W_D_1KI": 11.65088314183267, "J_D_1KI": 7.526410298341518}

66
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_500000_1e-05.output
View File

@ -1,15 +1,15 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 500000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 7.073613166809082}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 500000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.6782102584838867}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 13, ..., 2499990,
2499995, 2500000]),
col_indices=tensor([ 8141, 69274, 149925, ..., 390687, 407872,
439375]),
values=tensor([0.4271, 0.3560, 0.2859, ..., 0.3294, 0.0849, 0.5690]),
tensor(crow_indices=tensor([ 0, 8, 13, ..., 2499990,
2499997, 2500000]),
col_indices=tensor([ 40175, 122073, 147940, ..., 245767, 297950,
495791]),
values=tensor([0.1248, 0.8645, 0.7112, ..., 0.2227, 0.8085, 0.2637]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.1896, 0.3447, 0.8973, ..., 0.8957, 0.5716, 0.6993])
tensor([0.3055, 0.1588, 0.3916, ..., 0.3608, 0.8122, 0.4114])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -17,20 +17,20 @@ Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 7.073613166809082 seconds
Time: 0.6782102584838867 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1484 -ss 500000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.88543152809143}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1548 -ss 500000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 11.061279058456421}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 10, 15, ..., 2499994,
2500000, 2500000]),
col_indices=tensor([ 19808, 30523, 42041, ..., 253465, 473291,
475423]),
values=tensor([0.2655, 0.1335, 0.5252, ..., 0.0072, 0.8874, 0.1974]),
tensor(crow_indices=tensor([ 0, 0, 4, ..., 2499989,
2499994, 2500000]),
col_indices=tensor([ 33871, 87157, 252512, ..., 380315, 410804,
497208]),
values=tensor([0.0607, 0.8545, 0.0688, ..., 0.9965, 0.6178, 0.6113]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.7673, 0.2797, 0.0430, ..., 0.8352, 0.7956, 0.1250])
tensor([0.3094, 0.9384, 0.5289, ..., 0.5205, 0.7717, 0.7334])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -38,17 +38,17 @@ Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 10.88543152809143 seconds
Time: 11.061279058456421 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 10, 15, ..., 2499994,
2500000, 2500000]),
col_indices=tensor([ 19808, 30523, 42041, ..., 253465, 473291,
475423]),
values=tensor([0.2655, 0.1335, 0.5252, ..., 0.0072, 0.8874, 0.1974]),
tensor(crow_indices=tensor([ 0, 0, 4, ..., 2499989,
2499994, 2500000]),
col_indices=tensor([ 33871, 87157, 252512, ..., 380315, 410804,
497208]),
values=tensor([0.0607, 0.8545, 0.0688, ..., 0.9965, 0.6178, 0.6113]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.7673, 0.2797, 0.0430, ..., 0.8352, 0.7956, 0.1250])
tensor([0.3094, 0.9384, 0.5289, ..., 0.5205, 0.7717, 0.7334])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -56,13 +56,13 @@ Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 10.88543152809143 seconds
Time: 11.061279058456421 seconds
[20.16, 19.76, 19.4, 19.48, 19.12, 18.68, 18.6, 18.12, 18.12, 17.68]
[17.44, 17.16, 17.36, 21.36, 23.28, 27.36, 35.28, 38.6, 44.16, 47.92, 48.84, 48.56, 48.96, 49.04]
14.574582815170288
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1484, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.88543152809143, 'TIME_S_1KI': 7.335196447500964, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 480.2575412559509, 'W': 32.95171788766838}
[20.16, 19.76, 19.4, 19.48, 19.12, 18.68, 18.6, 18.12, 18.12, 17.68, 16.68, 16.56, 16.56, 16.56, 16.72, 16.64, 16.76, 17.08, 17.04, 17.08]
321.0
16.05
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1484, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.88543152809143, 'TIME_S_1KI': 7.335196447500964, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 480.2575412559509, 'W': 32.95171788766838, 'J_1KI': 323.6236800916111, 'W_1KI': 22.204661649372223, 'W_D': 16.90171788766838, 'J_D': 246.33548707246783, 'W_D_1KI': 11.389297767970607, 'J_D_1KI': 7.67472895415809}
[20.72, 20.76, 21.12, 20.92, 20.92, 20.92, 20.88, 20.44, 20.48, 20.48]
[20.56, 20.56, 21.24, 22.4, 24.48, 28.6, 36.72, 41.88, 48.52, 48.52, 52.36, 52.92, 53.48, 53.4, 53.48]
15.640001773834229
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1548, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 11.061279058456421, 'TIME_S_1KI': 7.145529107529987, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 576.3272948646545, 'W': 36.84956710355697}
[20.72, 20.76, 21.12, 20.92, 20.92, 20.92, 20.88, 20.44, 20.48, 20.48, 20.6, 20.68, 20.88, 21.24, 21.48, 21.32, 21.48, 20.96, 20.64, 20.52]
376.28
18.814
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1548, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 11.061279058456421, 'TIME_S_1KI': 7.145529107529987, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 576.3272948646545, 'W': 36.84956710355697, 'J_1KI': 372.3044540469344, 'W_1KI': 23.80462991185851, 'W_D': 18.03556710355697, 'J_D': 282.0763014917373, 'W_D_1KI': 11.65088314183267, 'J_D_1KI': 7.526410298341518}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_500000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 37.50764799118042, "TIME_S_1KI": 37.50764799118042, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2045.1465811538696, "W": 43.475713756728496, "J_1KI": 2045.1465811538699, "W_1KI": 43.475713756728496, "W_D": 24.798713756728496, "J_D": 1166.5594483480452, "W_D_1KI": 24.798713756728496, "J_D_1KI": 24.798713756728496}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 285, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 11.060697317123413, "TIME_S_1KI": 38.809464270608466, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 641.8192787170411, "W": 34.323302144261945, "J_1KI": 2251.9974691826005, "W_1KI": 120.43263910267349, "W_D": 15.818302144261942, "J_D": 295.7900504469872, "W_D_1KI": 55.50281454126997, "J_D_1KI": 194.74671768866656}

67
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_500000_5e-05.output
View File

@ -1,15 +1,15 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 500000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 37.50764799118042}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 500000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 3.6834404468536377}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 37, 66, ..., 12499959,
tensor(crow_indices=tensor([ 0, 36, 64, ..., 12499962,
12499977, 12500000]),
col_indices=tensor([ 2833, 12397, 17330, ..., 457394, 482426,
493028]),
values=tensor([0.7637, 0.4211, 0.1191, ..., 0.6468, 0.8330, 0.4871]),
col_indices=tensor([ 6540, 37225, 45963, ..., 476281, 491551,
491729]),
values=tensor([0.4995, 0.3434, 0.2289, ..., 0.9980, 0.3953, 0.2839]),
size=(500000, 500000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.3810, 0.7581, 0.5448, ..., 0.8790, 0.4682, 0.1184])
tensor([0.6494, 0.0196, 0.1697, ..., 0.1655, 0.3294, 0.8926])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -17,17 +17,20 @@ Rows: 500000
Size: 250000000000
NNZ: 12500000
Density: 5e-05
Time: 37.50764799118042 seconds
Time: 3.6834404468536377 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 285 -ss 500000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 11.060697317123413}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 37, 66, ..., 12499959,
12499977, 12500000]),
col_indices=tensor([ 2833, 12397, 17330, ..., 457394, 482426,
493028]),
values=tensor([0.7637, 0.4211, 0.1191, ..., 0.6468, 0.8330, 0.4871]),
tensor(crow_indices=tensor([ 0, 30, 51, ..., 12499947,
12499979, 12500000]),
col_indices=tensor([ 32854, 40713, 51141, ..., 464012, 471829,
496055]),
values=tensor([0.7704, 0.8573, 0.2864, ..., 0.9432, 0.9508, 0.7094]),
size=(500000, 500000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.3810, 0.7581, 0.5448, ..., 0.8790, 0.4682, 0.1184])
tensor([0.9627, 0.2273, 0.6691, ..., 0.1238, 0.9472, 0.0057])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -35,13 +38,31 @@ Rows: 500000
Size: 250000000000
NNZ: 12500000
Density: 5e-05
Time: 37.50764799118042 seconds
Time: 11.060697317123413 seconds
[21.08, 20.96, 20.88, 20.72, 20.6, 20.6, 20.64, 20.2, 20.28, 20.6]
[20.72, 20.68, 21.2, 23.0, 24.88, 26.4, 29.84, 29.72, 30.96, 36.72, 36.72, 41.28, 45.64, 50.76, 52.76, 52.48, 53.28, 52.92, 52.8, 52.76, 52.48, 52.52, 52.68, 52.72, 53.4, 53.44, 53.4, 53.36, 53.12, 53.4, 53.2, 53.0, 53.0, 53.24, 53.08, 52.92, 53.0, 52.72, 52.8, 52.64, 52.8, 52.96, 52.8, 52.8, 52.68]
47.04112720489502
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 37.50764799118042, 'TIME_S_1KI': 37.50764799118042, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2045.1465811538696, 'W': 43.475713756728496}
[21.08, 20.96, 20.88, 20.72, 20.6, 20.6, 20.64, 20.2, 20.28, 20.6, 20.92, 21.12, 21.0, 20.84, 20.92, 20.72, 20.72, 20.56, 21.0, 20.96]
373.54
18.677
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 37.50764799118042, 'TIME_S_1KI': 37.50764799118042, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2045.1465811538696, 'W': 43.475713756728496, 'J_1KI': 2045.1465811538699, 'W_1KI': 43.475713756728496, 'W_D': 24.798713756728496, 'J_D': 1166.5594483480452, 'W_D_1KI': 24.798713756728496, 'J_D_1KI': 24.798713756728496}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 30, 51, ..., 12499947,
12499979, 12500000]),
col_indices=tensor([ 32854, 40713, 51141, ..., 464012, 471829,
496055]),
values=tensor([0.7704, 0.8573, 0.2864, ..., 0.9432, 0.9508, 0.7094]),
size=(500000, 500000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.9627, 0.2273, 0.6691, ..., 0.1238, 0.9472, 0.0057])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
Rows: 500000
Size: 250000000000
NNZ: 12500000
Density: 5e-05
Time: 11.060697317123413 seconds
[20.52, 20.84, 20.92, 20.88, 20.88, 20.72, 20.52, 20.32, 20.36, 20.2]
[20.44, 20.6, 21.48, 21.48, 23.84, 25.24, 27.32, 30.28, 29.76, 32.76, 37.84, 41.44, 46.92, 51.96, 52.12, 52.64, 52.6, 52.8]
18.699228763580322
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 285, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 11.060697317123413, 'TIME_S_1KI': 38.809464270608466, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 641.8192787170411, 'W': 34.323302144261945}
[20.52, 20.84, 20.92, 20.88, 20.88, 20.72, 20.52, 20.32, 20.36, 20.2, 20.52, 20.44, 20.6, 20.44, 20.44, 20.48, 20.48, 20.36, 20.44, 20.72]
370.1
18.505000000000003
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 285, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 11.060697317123413, 'TIME_S_1KI': 38.809464270608466, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 641.8192787170411, 'W': 34.323302144261945, 'J_1KI': 2251.9974691826005, 'W_1KI': 120.43263910267349, 'W_D': 15.818302144261942, 'J_D': 295.7900504469872, 'W_D_1KI': 55.50281454126997, 'J_D_1KI': 194.74671768866656}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 3392, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.727018594741821, "TIME_S_1KI": 3.1624465196762443, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 423.8805270671845, "W": 29.00744018011372, "J_1KI": 124.96477802688223, "W_1KI": 8.55172175121277, "W_D": 13.914440180113719, "J_D": 203.32922177100187, "W_D_1KI": 4.102134487061827, "J_D_1KI": 1.2093556860441705}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 3424, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.731184720993042, "TIME_S_1KI": 3.1341076872059115, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 463.79336466789243, "W": 31.765422510706816, "J_1KI": 135.4536695875854, "W_1KI": 9.27728461177185, "W_D": 13.349422510706813, "J_D": 194.90921553230277, "W_D_1KI": 3.8987799388746534, "J_D_1KI": 1.1386623653255412}

74
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_0.0001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 50000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 3.0953831672668457}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 50000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.4611988067626953}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 13, ..., 249985, 249991,
tensor(crow_indices=tensor([ 0, 7, 9, ..., 249990, 249996,
250000]),
col_indices=tensor([ 782, 10679, 21591, ..., 21721, 25862, 26402]),
values=tensor([0.1080, 0.2599, 0.9753, ..., 0.8598, 0.0309, 0.7621]),
col_indices=tensor([ 1266, 4071, 18947, ..., 33754, 36171, 46993]),
values=tensor([0.2894, 0.3028, 0.5808, ..., 0.9499, 0.5530, 0.4490]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.0624, 0.3415, 0.4601, ..., 0.0482, 0.7737, 0.1465])
tensor([0.9097, 0.0887, 0.0049, ..., 0.6179, 0.8641, 0.1772])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -16,19 +16,19 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 3.0953831672668457 seconds
Time: 0.4611988067626953 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 3392 -ss 50000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.727018594741821}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 2276 -ss 50000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 6.978086233139038}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 3, 6, ..., 249992, 249997,
tensor(crow_indices=tensor([ 0, 5, 9, ..., 249990, 249995,
250000]),
col_indices=tensor([29888, 37512, 45145, ..., 10362, 27481, 28096]),
values=tensor([0.5987, 0.4413, 0.1210, ..., 0.9023, 0.1888, 0.1246]),
col_indices=tensor([ 2233, 6887, 19755, ..., 38632, 41476, 48223]),
values=tensor([0.3109, 0.9167, 0.4160, ..., 0.6671, 0.8506, 0.5777]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.0260, 0.0462, 0.3716, ..., 0.4992, 0.3586, 0.2225])
tensor([0.7812, 0.4224, 0.5960, ..., 0.2514, 0.6292, 0.3012])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -36,16 +36,19 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 10.727018594741821 seconds
Time: 6.978086233139038 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 3424 -ss 50000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.731184720993042}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 3, 6, ..., 249992, 249997,
tensor(crow_indices=tensor([ 0, 6, 7, ..., 249990, 249996,
250000]),
col_indices=tensor([29888, 37512, 45145, ..., 10362, 27481, 28096]),
values=tensor([0.5987, 0.4413, 0.1210, ..., 0.9023, 0.1888, 0.1246]),
col_indices=tensor([12104, 14436, 24112, ..., 12878, 32819, 38734]),
values=tensor([0.5759, 0.9600, 0.3696, ..., 0.0040, 0.7766, 0.9665]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.0260, 0.0462, 0.3716, ..., 0.4992, 0.3586, 0.2225])
tensor([0.4011, 0.3434, 0.3941, ..., 0.7256, 0.6030, 0.5117])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -53,13 +56,30 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 10.727018594741821 seconds
Time: 10.731184720993042 seconds
[16.52, 16.28, 16.24, 16.16, 16.44, 16.44, 16.6, 16.52, 16.56, 16.68]
[16.44, 16.64, 17.44, 19.52, 22.44, 27.08, 32.32, 35.8, 38.92, 39.84, 40.12, 40.12, 40.24, 40.6]
14.612820863723755
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3392, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.727018594741821, 'TIME_S_1KI': 3.1624465196762443, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 423.8805270671845, 'W': 29.00744018011372}
[16.52, 16.28, 16.24, 16.16, 16.44, 16.44, 16.6, 16.52, 16.56, 16.68, 16.84, 16.68, 16.84, 17.24, 17.32, 17.48, 17.4, 17.24, 16.96, 16.88]
301.86
15.093
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3392, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.727018594741821, 'TIME_S_1KI': 3.1624465196762443, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 423.8805270671845, 'W': 29.00744018011372, 'J_1KI': 124.96477802688223, 'W_1KI': 8.55172175121277, 'W_D': 13.914440180113719, 'J_D': 203.32922177100187, 'W_D_1KI': 4.102134487061827, 'J_D_1KI': 1.2093556860441705}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 7, ..., 249990, 249996,
250000]),
col_indices=tensor([12104, 14436, 24112, ..., 12878, 32819, 38734]),
values=tensor([0.5759, 0.9600, 0.3696, ..., 0.0040, 0.7766, 0.9665]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.4011, 0.3434, 0.3941, ..., 0.7256, 0.6030, 0.5117])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 10.731184720993042 seconds
[20.72, 20.68, 20.48, 20.6, 20.64, 20.32, 20.32, 20.36, 20.32, 20.4]
[20.52, 20.76, 22.24, 22.24, 24.16, 27.64, 32.84, 38.08, 39.84, 44.4, 44.52, 43.96, 44.12, 44.28]
14.60057282447815
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3424, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.731184720993042, 'TIME_S_1KI': 3.1341076872059115, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 463.79336466789243, 'W': 31.765422510706816}
[20.72, 20.68, 20.48, 20.6, 20.64, 20.32, 20.32, 20.36, 20.32, 20.4, 20.68, 20.44, 20.48, 20.48, 20.2, 20.32, 20.44, 20.44, 20.6, 20.6]
368.32000000000005
18.416000000000004
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 3424, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.731184720993042, 'TIME_S_1KI': 3.1341076872059115, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 463.79336466789243, 'W': 31.765422510706816, 'J_1KI': 135.4536695875854, 'W_1KI': 9.27728461177185, 'W_D': 13.349422510706813, 'J_D': 194.90921553230277, 'W_D_1KI': 3.8987799388746534, 'J_D_1KI': 1.1386623653255412}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_0.001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 29.441463470458984, "TIME_S_1KI": 29.441463470458984, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1067.5749865341186, "W": 32.87033016720936, "J_1KI": 1067.5749865341186, "W_1KI": 32.87033016720936, "W_D": 17.51733016720936, "J_D": 568.9344591989515, "W_D_1KI": 17.51733016720936, "J_D_1KI": 17.51733016720936}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 385, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 11.064192295074463, "TIME_S_1KI": 28.738161805388216, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 468.0283072185517, "W": 32.037671646494225, "J_1KI": 1215.657940827407, "W_1KI": 83.21473154933565, "W_D": 13.341671646494227, "J_D": 194.90430094528205, "W_D_1KI": 34.65369258829669, "J_D_1KI": 90.00959113843297}

84
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_0.001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 50000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 29.441463470458984}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 50000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 2.9453940391540527}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 37, 79, ..., 2499907,
2499951, 2500000]),
col_indices=tensor([ 2466, 3763, 4276, ..., 47502, 48879, 49149]),
values=tensor([0.0148, 0.9908, 0.2997, ..., 0.9281, 0.7443, 0.4383]),
tensor(crow_indices=tensor([ 0, 37, 86, ..., 2499902,
2499952, 2500000]),
col_indices=tensor([ 541, 1139, 1813, ..., 42919, 43072, 44933]),
values=tensor([0.0452, 0.1724, 0.8861, ..., 0.4157, 0.9772, 0.2120]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.6700, 0.5614, 0.5608, ..., 0.8928, 0.8615, 0.5607])
tensor([0.1045, 0.1557, 0.1178, ..., 0.5894, 0.9079, 0.5773])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -16,16 +16,19 @@ Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 29.441463470458984 seconds
Time: 2.9453940391540527 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 356 -ss 50000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 9.708060026168823}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 37, 79, ..., 2499907,
2499951, 2500000]),
col_indices=tensor([ 2466, 3763, 4276, ..., 47502, 48879, 49149]),
values=tensor([0.0148, 0.9908, 0.2997, ..., 0.9281, 0.7443, 0.4383]),
tensor(crow_indices=tensor([ 0, 46, 99, ..., 2499902,
2499948, 2500000]),
col_indices=tensor([ 4031, 7226, 7309, ..., 44877, 48582, 49711]),
values=tensor([0.9329, 0.4420, 0.5313, ..., 0.9423, 0.2849, 0.2389]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.6700, 0.5614, 0.5608, ..., 0.8928, 0.8615, 0.5607])
tensor([0.0088, 0.8123, 0.3302, ..., 0.9483, 0.6171, 0.9552])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -33,13 +36,50 @@ Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 29.441463470458984 seconds
Time: 9.708060026168823 seconds
[16.92, 17.36, 17.08, 16.84, 17.2, 17.28, 17.4, 17.44, 17.36, 17.36]
[17.08, 17.24, 17.08, 21.52, 22.36, 25.24, 29.48, 32.12, 34.64, 38.4, 38.76, 38.64, 38.8, 39.0, 39.0, 39.44, 39.4, 39.28, 39.32, 39.32, 39.24, 39.12, 39.0, 39.08, 39.32, 39.36, 39.28, 39.28, 39.16, 38.92, 39.08]
32.47837734222412
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 29.441463470458984, 'TIME_S_1KI': 29.441463470458984, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1067.5749865341186, 'W': 32.87033016720936}
[16.92, 17.36, 17.08, 16.84, 17.2, 17.28, 17.4, 17.44, 17.36, 17.36, 16.68, 16.88, 16.76, 16.92, 16.84, 17.04, 17.0, 17.0, 16.68, 17.0]
307.06000000000006
15.353000000000003
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 29.441463470458984, 'TIME_S_1KI': 29.441463470458984, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1067.5749865341186, 'W': 32.87033016720936, 'J_1KI': 1067.5749865341186, 'W_1KI': 32.87033016720936, 'W_D': 17.51733016720936, 'J_D': 568.9344591989515, 'W_D_1KI': 17.51733016720936, 'J_D_1KI': 17.51733016720936}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 385 -ss 50000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 11.064192295074463}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 67, 125, ..., 2499902,
2499956, 2500000]),
col_indices=tensor([ 1129, 2884, 2891, ..., 49010, 49022, 49816]),
values=tensor([0.8127, 0.7656, 0.2912, ..., 0.8978, 0.1718, 0.1428]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.7569, 0.5985, 0.1427, ..., 0.6714, 0.1732, 0.3064])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 11.064192295074463 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 67, 125, ..., 2499902,
2499956, 2500000]),
col_indices=tensor([ 1129, 2884, 2891, ..., 49010, 49022, 49816]),
values=tensor([0.8127, 0.7656, 0.2912, ..., 0.8978, 0.1718, 0.1428]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.7569, 0.5985, 0.1427, ..., 0.6714, 0.1732, 0.3064])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 11.064192295074463 seconds
[20.72, 20.52, 20.56, 20.48, 20.48, 20.52, 20.56, 20.6, 20.36, 20.6]
[20.48, 20.52, 21.48, 23.4, 25.44, 29.64, 35.04, 38.48, 42.0, 43.12, 43.12, 43.84, 43.68, 43.32]
14.608686685562134
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 385, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 11.064192295074463, 'TIME_S_1KI': 28.738161805388216, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 468.0283072185517, 'W': 32.037671646494225}
[20.72, 20.52, 20.56, 20.48, 20.48, 20.52, 20.56, 20.6, 20.36, 20.6, 20.84, 20.8, 21.0, 21.24, 21.16, 21.08, 21.28, 20.92, 20.84, 20.88]
373.91999999999996
18.695999999999998
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 385, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 11.064192295074463, 'TIME_S_1KI': 28.738161805388216, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 468.0283072185517, 'W': 32.037671646494225, 'J_1KI': 1215.657940827407, 'W_1KI': 83.21473154933565, 'W_D': 13.341671646494227, 'J_D': 194.90430094528205, 'W_D_1KI': 34.65369258829669, 'J_D_1KI': 90.00959113843297}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_0.01.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.01, "TIME_S": 324.79648518562317, "TIME_S_1KI": 324.79648518562317, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 11877.425473241823, "W": 35.646067141867775, "J_1KI": 11877.425473241823, "W_1KI": 35.646067141867775, "W_D": 16.970067141867776, "J_D": 5654.50059192373, "W_D_1KI": 16.970067141867776, "J_D_1KI": 16.970067141867776}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.01, "TIME_S": 32.80737018585205, "TIME_S_1KI": 328.0737018585205, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1398.267660713196, "W": 32.405416577679354, "J_1KI": 13982.676607131958, "W_1KI": 324.0541657767935, "W_D": 13.697416577679352, "J_D": 591.0325080986024, "W_D_1KI": 136.97416577679354, "J_D_1KI": 1369.7416577679353}

44
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_0.01.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 50000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.01, "TIME_S": 324.79648518562317}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 50000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.01, "TIME_S": 32.80737018585205}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 495, 976, ..., 24999061,
24999546, 25000000]),
col_indices=tensor([ 151, 490, 496, ..., 49361, 49363, 49505]),
values=tensor([0.8177, 0.6830, 0.3837, ..., 0.7178, 0.0658, 0.9045]),
tensor(crow_indices=tensor([ 0, 489, 963, ..., 24999055,
24999529, 25000000]),
col_indices=tensor([ 18, 157, 241, ..., 49747, 49771, 49960]),
values=tensor([0.7706, 0.7949, 0.9210, ..., 0.0962, 0.6322, 0.0053]),
size=(50000, 50000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.7632, 0.2578, 0.8207, ..., 0.6267, 0.5426, 0.4264])
tensor([0.2729, 0.2896, 0.6966, ..., 0.6831, 0.4086, 0.6520])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -16,16 +16,16 @@ Rows: 50000
Size: 2500000000
NNZ: 25000000
Density: 0.01
Time: 324.79648518562317 seconds
Time: 32.80737018585205 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 495, 976, ..., 24999061,
24999546, 25000000]),
col_indices=tensor([ 151, 490, 496, ..., 49361, 49363, 49505]),
values=tensor([0.8177, 0.6830, 0.3837, ..., 0.7178, 0.0658, 0.9045]),
tensor(crow_indices=tensor([ 0, 489, 963, ..., 24999055,
24999529, 25000000]),
col_indices=tensor([ 18, 157, 241, ..., 49747, 49771, 49960]),
values=tensor([0.7706, 0.7949, 0.9210, ..., 0.0962, 0.6322, 0.0053]),
size=(50000, 50000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.7632, 0.2578, 0.8207, ..., 0.6267, 0.5426, 0.4264])
tensor([0.2729, 0.2896, 0.6966, ..., 0.6831, 0.4086, 0.6520])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -33,13 +33,13 @@ Rows: 50000
Size: 2500000000
NNZ: 25000000
Density: 0.01
Time: 324.79648518562317 seconds
Time: 32.80737018585205 seconds
[20.76, 20.6, 20.88, 20.96, 20.84, 20.88, 20.88, 20.68, 20.8, 20.88]
[21.16, 21.04, 21.12, 26.2, 28.16, 29.12, 31.64, 29.84, 30.96, 30.96, 31.32, 31.32, 31.52, 31.84, 31.92, 33.6, 35.32, 36.84, 37.0, 36.88, 36.84, 36.12, 36.76, 36.76, 36.4, 37.24, 37.32, 37.12, 37.84, 37.0, 37.76, 37.32, 37.0, 35.96, 36.28, 35.4, 35.8, 36.36, 37.12, 37.12, 38.8, 38.96, 38.72, 38.44, 37.4, 38.36, 38.0, 38.04, 37.84, 37.56, 36.92, 36.92, 36.84, 36.48, 36.64, 36.64, 36.52, 35.64, 36.12, 36.84, 37.56, 37.52, 38.6, 37.96, 37.32, 37.44, 36.76, 37.36, 37.44, 37.76, 37.8, 37.8, 38.68, 38.12, 37.96, 37.72, 37.88, 37.92, 37.84, 36.92, 36.6, 36.4, 36.48, 36.4, 37.2, 37.04, 37.16, 36.8, 36.8, 36.72, 37.72, 38.0, 37.96, 37.76, 37.08, 37.48, 36.64, 36.8, 36.88, 37.2, 37.24, 37.44, 37.48, 38.04, 38.04, 38.16, 38.68, 37.96, 38.24, 37.16, 36.68, 36.84, 36.88, 37.16, 37.76, 37.92, 37.76, 37.56, 36.52, 36.0, 37.0, 36.44, 36.36, 36.68, 37.08, 37.4, 37.24, 37.32, 36.6, 36.2, 37.16, 37.32, 37.6, 37.6, 37.6, 37.56, 37.56, 37.24, 36.48, 36.28, 36.48, 36.64, 37.68, 38.24, 37.72, 37.64, 38.24, 37.6, 37.0, 37.0, 36.88, 36.88, 37.28, 38.48, 39.08, 38.28, 38.04, 37.48, 36.64, 36.72, 36.84, 36.84, 37.2, 37.36, 37.76, 37.96, 38.24, 37.88, 37.88, 37.12, 37.6, 36.76, 37.52, 37.68, 36.76, 37.72, 37.48, 38.04, 37.88, 37.6, 37.56, 36.96, 37.0, 37.0, 37.92, 37.08, 37.44, 36.8, 36.84, 36.08, 36.52, 36.48, 36.56, 36.84, 37.2, 37.36, 36.52, 37.0, 36.12, 36.12, 37.0, 36.68, 36.88, 37.56, 37.72, 38.36, 38.2, 38.48, 38.72, 38.36, 38.28, 37.96, 37.76, 36.36, 37.0, 36.48, 36.52, 36.52, 37.16, 36.6, 36.52, 36.6, 37.52, 37.12, 37.8, 37.88, 37.04, 36.64, 36.44, 36.04, 36.32, 37.68, 37.88, 37.88, 38.04, 38.04, 37.68, 37.92, 37.96, 36.92, 37.64, 36.4, 36.32, 36.4, 36.32, 36.2, 37.04, 37.16, 37.68, 37.64, 37.64, 38.12, 38.04, 37.64, 37.24, 36.56, 36.48, 37.28, 36.6, 36.44, 37.08, 37.08, 36.56, 37.48, 38.08, 37.2, 37.2, 36.96, 36.72, 36.64, 36.24, 37.32, 37.96, 38.2, 38.28, 38.36, 38.36, 37.88, 38.36, 37.64, 36.88, 36.88, 37.2, 36.4, 36.52, 37.2, 37.52, 37.44, 36.8, 37.48, 36.92, 37.32, 38.0, 37.76, 36.72, 37.84, 37.64, 37.64, 38.32, 38.32, 37.88, 38.16, 38.24, 37.64, 37.76, 37.12, 37.04, 36.24, 36.68, 36.4, 36.6, 36.6, 36.64, 37.16, 38.08, 37.28, 37.28, 37.24, 36.52, 36.4, 36.88, 36.2, 36.92]
333.20437359809875
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 324.79648518562317, 'TIME_S_1KI': 324.79648518562317, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 11877.425473241823, 'W': 35.646067141867775}
[20.76, 20.6, 20.88, 20.96, 20.84, 20.88, 20.88, 20.68, 20.8, 20.88, 20.44, 20.44, 20.64, 20.68, 20.68, 20.68, 20.68, 20.84, 20.88, 20.88]
373.52
18.676
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 324.79648518562317, 'TIME_S_1KI': 324.79648518562317, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 11877.425473241823, 'W': 35.646067141867775, 'J_1KI': 11877.425473241823, 'W_1KI': 35.646067141867775, 'W_D': 16.970067141867776, 'J_D': 5654.50059192373, 'W_D_1KI': 16.970067141867776, 'J_D_1KI': 16.970067141867776}
[20.96, 20.68, 20.68, 20.68, 20.72, 20.64, 20.64, 20.76, 20.84, 20.92]
[20.92, 20.84, 21.32, 22.6, 25.12, 26.44, 26.44, 28.96, 28.92, 32.68, 31.76, 31.16, 31.4, 31.52, 32.12, 34.6, 36.28, 37.8, 37.28, 37.56, 37.96, 37.96, 38.28, 38.36, 37.96, 37.32, 36.48, 36.68, 36.84, 37.44, 37.44, 37.76, 38.0, 39.0, 37.68, 36.84, 37.4, 37.4, 36.8, 37.32, 37.6, 37.72]
43.14919567108154
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 32.80737018585205, 'TIME_S_1KI': 328.0737018585205, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1398.267660713196, 'W': 32.405416577679354}
[20.96, 20.68, 20.68, 20.68, 20.72, 20.64, 20.64, 20.76, 20.84, 20.92, 21.04, 20.84, 20.72, 20.68, 20.72, 20.6, 20.92, 21.2, 20.96, 20.84]
374.16
18.708000000000002
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 32.80737018585205, 'TIME_S_1KI': 328.0737018585205, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1398.267660713196, 'W': 32.405416577679354, 'J_1KI': 13982.676607131958, 'W_1KI': 324.0541657767935, 'W_D': 13.697416577679352, 'J_D': 591.0325080986024, 'W_D_1KI': 136.97416577679354, 'J_D_1KI': 1369.7416577679353}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 20098, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.650188207626343, "TIME_S_1KI": 0.5299128374776765, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 402.8303679275513, "W": 27.52967890413959, "J_1KI": 20.04330619601708, "W_1KI": 1.3697720621026763, "W_D": 12.350678904139592, "J_D": 180.72235947370535, "W_D_1KI": 0.6145227835674988, "J_D_1KI": 0.030576315233729664}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 19951, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.462696075439453, "TIME_S_1KI": 0.5244196318700542, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 488.71538955688476, "W": 33.3802333465878, "J_1KI": 24.49578414900931, "W_1KI": 1.6731107887618566, "W_D": 15.011233346587801, "J_D": 219.77739569807053, "W_D_1KI": 0.752405059725718, "J_D_1KI": 0.03771264897627778}

68
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_1e-05.output
View File

@ -1,13 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 50000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.648245096206665}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 50000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.05949115753173828}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 24997, 24999, 25000]),
col_indices=tensor([ 889, 16856, 49649, ..., 20622, 24354, 47394]),
values=tensor([0.8512, 0.0995, 0.9072, ..., 0.9114, 0.3857, 0.4483]),
tensor(crow_indices=tensor([ 0, 0, 0, ..., 24998, 24999, 25000]),
col_indices=tensor([35821, 49411, 3789, ..., 32092, 27347, 39445]),
values=tensor([0.1439, 0.1701, 0.0383, ..., 0.6521, 0.3755, 0.5678]),
size=(50000, 50000), nnz=25000, layout=torch.sparse_csr)
tensor([0.8531, 0.5584, 0.8209, ..., 0.8853, 0.7506, 0.6837])
tensor([0.8709, 0.6173, 0.3475, ..., 0.7020, 0.1451, 0.7453])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -15,18 +15,18 @@ Rows: 50000
Size: 2500000000
NNZ: 25000
Density: 1e-05
Time: 0.648245096206665 seconds
Time: 0.05949115753173828 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 16197 -ss 50000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 8.461615800857544}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 17649 -ss 50000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 9.288093090057373}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 25000, 25000, 25000]),
col_indices=tensor([37259, 33129, 13575, ..., 31298, 24333, 9136]),
values=tensor([0.0302, 0.8728, 0.1875, ..., 0.5590, 0.6136, 0.6206]),
tensor(crow_indices=tensor([ 0, 0, 2, ..., 24999, 25000, 25000]),
col_indices=tensor([10903, 22613, 1325, ..., 4616, 25772, 38217]),
values=tensor([0.1548, 0.5404, 0.0562, ..., 0.6796, 0.5534, 0.6437]),
size=(50000, 50000), nnz=25000, layout=torch.sparse_csr)
tensor([0.6191, 0.3887, 0.4199, ..., 0.2754, 0.8424, 0.8817])
tensor([0.8066, 0.3465, 0.3699, ..., 0.5654, 0.2544, 0.1290])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -34,18 +34,18 @@ Rows: 50000
Size: 2500000000
NNZ: 25000
Density: 1e-05
Time: 8.461615800857544 seconds
Time: 9.288093090057373 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 20098 -ss 50000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.650188207626343}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 19951 -ss 50000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.462696075439453}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 2, ..., 24999, 25000, 25000]),
col_indices=tensor([ 35, 8013, 35741, ..., 26171, 43365, 6398]),
values=tensor([0.7135, 0.5997, 0.1893, ..., 0.8752, 0.2236, 0.9882]),
tensor(crow_indices=tensor([ 0, 1, 2, ..., 24997, 24998, 25000]),
col_indices=tensor([36526, 27522, 9271, ..., 28337, 20494, 41611]),
values=tensor([0.2838, 0.5711, 0.3512, ..., 0.1758, 0.7475, 0.3339]),
size=(50000, 50000), nnz=25000, layout=torch.sparse_csr)
tensor([0.7523, 0.4685, 0.7648, ..., 0.0829, 0.9708, 0.7467])
tensor([0.9803, 0.0496, 0.4924, ..., 0.5397, 0.0486, 0.3592])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -53,15 +53,15 @@ Rows: 50000
Size: 2500000000
NNZ: 25000
Density: 1e-05
Time: 10.650188207626343 seconds
Time: 10.462696075439453 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 2, ..., 24999, 25000, 25000]),
col_indices=tensor([ 35, 8013, 35741, ..., 26171, 43365, 6398]),
values=tensor([0.7135, 0.5997, 0.1893, ..., 0.8752, 0.2236, 0.9882]),
tensor(crow_indices=tensor([ 0, 1, 2, ..., 24997, 24998, 25000]),
col_indices=tensor([36526, 27522, 9271, ..., 28337, 20494, 41611]),
values=tensor([0.2838, 0.5711, 0.3512, ..., 0.1758, 0.7475, 0.3339]),
size=(50000, 50000), nnz=25000, layout=torch.sparse_csr)
tensor([0.7523, 0.4685, 0.7648, ..., 0.0829, 0.9708, 0.7467])
tensor([0.9803, 0.0496, 0.4924, ..., 0.5397, 0.0486, 0.3592])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -69,13 +69,13 @@ Rows: 50000
Size: 2500000000
NNZ: 25000
Density: 1e-05
Time: 10.650188207626343 seconds
Time: 10.462696075439453 seconds
[16.64, 16.56, 16.56, 16.96, 17.08, 17.16, 16.88, 16.96, 16.48, 16.36]
[16.44, 16.44, 16.68, 18.12, 19.04, 22.72, 28.64, 33.16, 36.92, 39.76, 39.32, 39.28, 39.68, 39.6]
14.632585048675537
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 20098, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.650188207626343, 'TIME_S_1KI': 0.5299128374776765, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 402.8303679275513, 'W': 27.52967890413959}
[16.64, 16.56, 16.56, 16.96, 17.08, 17.16, 16.88, 16.96, 16.48, 16.36, 17.12, 16.92, 16.68, 16.88, 16.76, 16.8, 16.8, 17.12, 17.32, 17.2]
303.58
15.178999999999998
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 20098, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.650188207626343, 'TIME_S_1KI': 0.5299128374776765, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 402.8303679275513, 'W': 27.52967890413959, 'J_1KI': 20.04330619601708, 'W_1KI': 1.3697720621026763, 'W_D': 12.350678904139592, 'J_D': 180.72235947370535, 'W_D_1KI': 0.6145227835674988, 'J_D_1KI': 0.030576315233729664}
[20.44, 20.56, 20.6, 20.64, 20.64, 20.44, 20.44, 20.44, 20.4, 20.56]
[20.44, 20.64, 23.28, 24.2, 26.6, 32.52, 37.24, 40.44, 44.28, 44.96, 44.96, 44.92, 44.44, 43.84]
14.640861988067627
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 19951, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.462696075439453, 'TIME_S_1KI': 0.5244196318700542, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 488.71538955688476, 'W': 33.3802333465878}
[20.44, 20.56, 20.6, 20.64, 20.64, 20.44, 20.44, 20.44, 20.4, 20.56, 20.32, 20.32, 20.12, 20.24, 20.32, 20.56, 20.52, 20.28, 20.2, 20.0]
367.38
18.369
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 19951, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.462696075439453, 'TIME_S_1KI': 0.5244196318700542, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 488.71538955688476, 'W': 33.3802333465878, 'J_1KI': 24.49578414900931, 'W_1KI': 1.6731107887618566, 'W_D': 15.011233346587801, 'J_D': 219.77739569807053, 'W_D_1KI': 0.752405059725718, 'J_D_1KI': 0.03771264897627778}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 6265, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 125000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.399809837341309, "TIME_S_1KI": 1.6599856085141753, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 409.2376235961914, "W": 30.158092261325162, "J_1KI": 65.32124877832264, "W_1KI": 4.813741781536339, "W_D": 11.570092261325161, "J_D": 157.0032023506164, "W_D_1KI": 1.846782483850784, "J_D_1KI": 0.2947777308620565}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 6322, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 125000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.657772779464722, "TIME_S_1KI": 1.68582296416715, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 425.8721401214599, "W": 31.344652680689325, "J_1KI": 67.3635147297469, "W_1KI": 4.958027946961298, "W_D": 12.573652680689325, "J_D": 170.83514789009087, "W_D_1KI": 1.9888726163697126, "J_D_1KI": 0.31459547870447846}

74
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_50000_5e-05.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 50000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 125000, "MATRIX_DENSITY": 5e-05, "TIME_S": 1.6757559776306152}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 50000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 125000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.3423471450805664}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 3, 4, ..., 124996, 124997,
tensor(crow_indices=tensor([ 0, 4, 7, ..., 124998, 124999,
125000]),
col_indices=tensor([11324, 36531, 41582, ..., 26561, 37075, 42675]),
values=tensor([0.0907, 0.5500, 0.9495, ..., 0.7742, 0.3202, 0.5187]),
col_indices=tensor([ 303, 26221, 28347, ..., 8622, 14261, 4291]),
values=tensor([0.9240, 0.5223, 0.0365, ..., 0.6044, 0.0072, 0.5479]),
size=(50000, 50000), nnz=125000, layout=torch.sparse_csr)
tensor([0.4295, 0.8994, 0.1269, ..., 0.0289, 0.7051, 0.4729])
tensor([0.1523, 0.9417, 0.1754, ..., 0.6908, 0.2427, 0.5501])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -16,19 +16,19 @@ Rows: 50000
Size: 2500000000
NNZ: 125000
Density: 5e-05
Time: 1.6757559776306152 seconds
Time: 0.3423471450805664 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 6265 -ss 50000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 125000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.399809837341309}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 3067 -ss 50000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 125000, "MATRIX_DENSITY": 5e-05, "TIME_S": 5.0935986042022705}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 3, 6, ..., 124998, 125000,
tensor(crow_indices=tensor([ 0, 1, 3, ..., 124997, 125000,
125000]),
col_indices=tensor([ 5059, 34750, 35724, ..., 34703, 6591, 31118]),
values=tensor([0.4217, 0.1867, 0.1593, ..., 0.5339, 0.2274, 0.5888]),
col_indices=tensor([ 1194, 5034, 6320, ..., 11179, 21504, 33093]),
values=tensor([0.7209, 0.3055, 0.4482, ..., 0.3076, 0.8643, 0.0918]),
size=(50000, 50000), nnz=125000, layout=torch.sparse_csr)
tensor([0.7798, 0.1756, 0.8709, ..., 0.5047, 0.8577, 0.3016])
tensor([0.9680, 0.6265, 0.9723, ..., 0.1304, 0.1284, 0.7215])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -36,16 +36,19 @@ Rows: 50000
Size: 2500000000
NNZ: 125000
Density: 5e-05
Time: 10.399809837341309 seconds
Time: 5.0935986042022705 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 6322 -ss 50000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 125000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.657772779464722}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 3, 6, ..., 124998, 125000,
tensor(crow_indices=tensor([ 0, 1, 2, ..., 124992, 124996,
125000]),
col_indices=tensor([ 5059, 34750, 35724, ..., 34703, 6591, 31118]),
values=tensor([0.4217, 0.1867, 0.1593, ..., 0.5339, 0.2274, 0.5888]),
col_indices=tensor([41720, 5446, 23409, ..., 23991, 37197, 42632]),
values=tensor([0.7857, 0.2010, 0.0929, ..., 0.8446, 0.3352, 0.3559]),
size=(50000, 50000), nnz=125000, layout=torch.sparse_csr)
tensor([0.7798, 0.1756, 0.8709, ..., 0.5047, 0.8577, 0.3016])
tensor([0.5851, 0.1828, 0.1733, ..., 0.7326, 0.4663, 0.8685])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -53,13 +56,30 @@ Rows: 50000
Size: 2500000000
NNZ: 125000
Density: 5e-05
Time: 10.399809837341309 seconds
Time: 10.657772779464722 seconds
[20.92, 20.84, 20.96, 20.84, 20.68, 20.64, 20.84, 20.76, 20.84, 21.08]
[21.08, 21.16, 21.84, 21.84, 23.04, 26.04, 31.36, 35.8, 40.08, 43.12, 43.68, 43.84, 43.84]
13.569745063781738
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 6265, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 125000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.399809837341309, 'TIME_S_1KI': 1.6599856085141753, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 409.2376235961914, 'W': 30.158092261325162}
[20.92, 20.84, 20.96, 20.84, 20.68, 20.64, 20.84, 20.76, 20.84, 21.08, 20.04, 19.96, 20.16, 20.28, 20.56, 20.88, 20.76, 20.68, 20.68, 20.76]
371.76000000000005
18.588
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 6265, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 125000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.399809837341309, 'TIME_S_1KI': 1.6599856085141753, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 409.2376235961914, 'W': 30.158092261325162, 'J_1KI': 65.32124877832264, 'W_1KI': 4.813741781536339, 'W_D': 11.570092261325161, 'J_D': 157.0032023506164, 'W_D_1KI': 1.846782483850784, 'J_D_1KI': 0.2947777308620565}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 2, ..., 124992, 124996,
125000]),
col_indices=tensor([41720, 5446, 23409, ..., 23991, 37197, 42632]),
values=tensor([0.7857, 0.2010, 0.0929, ..., 0.8446, 0.3352, 0.3559]),
size=(50000, 50000), nnz=125000, layout=torch.sparse_csr)
tensor([0.5851, 0.1828, 0.1733, ..., 0.7326, 0.4663, 0.8685])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
Rows: 50000
Size: 2500000000
NNZ: 125000
Density: 5e-05
Time: 10.657772779464722 seconds
[20.44, 20.32, 20.44, 20.56, 20.76, 20.84, 20.88, 21.04, 21.08, 21.08]
[21.08, 20.88, 20.92, 24.6, 25.64, 31.32, 35.8, 37.92, 41.28, 43.2, 43.12, 43.4, 43.48]
13.586755752563477
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 6322, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 125000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.657772779464722, 'TIME_S_1KI': 1.68582296416715, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 425.8721401214599, 'W': 31.344652680689325}
[20.44, 20.32, 20.44, 20.56, 20.76, 20.84, 20.88, 21.04, 21.08, 21.08, 20.84, 20.88, 20.84, 20.72, 20.96, 20.96, 21.08, 21.16, 21.2, 21.04]
375.42
18.771
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 6322, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 125000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.657772779464722, 'TIME_S_1KI': 1.68582296416715, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 425.8721401214599, 'W': 31.344652680689325, 'J_1KI': 67.3635147297469, 'W_1KI': 4.958027946961298, 'W_D': 12.573652680689325, 'J_D': 170.83514789009087, 'W_D_1KI': 1.9888726163697126, 'J_D_1KI': 0.31459547870447846}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 96690, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.663233041763306, "TIME_S_1KI": 0.11028268736956569, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 262.53495136260995, "W": 18.48739374467239, "J_1KI": 2.7152234084456506, "W_1KI": 0.19120274841940624, "W_D": 3.6973937446723912, "J_D": 52.505783147812, "W_D_1KI": 0.038239670541652615, "J_D_1KI": 0.0003954873362462779}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 98325, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.519834756851196, "TIME_S_1KI": 0.10699043739487614, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 316.21768825531, "W": 22.200046075478067, "J_1KI": 3.216045647142741, "W_1KI": 0.22578231452304162, "W_D": 3.72304607547807, "J_D": 53.03110719919203, "W_D_1KI": 0.03786469438574187, "J_D_1KI": 0.0003850973240350051}

93
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.0001.output
View File

@ -1,32 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.11520600318908691}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 2498, 2498, 2500]),
col_indices=tensor([4712, 1560, 1507, ..., 2651, 244, 3781]),
values=tensor([0.1646, 0.3564, 0.3355, ..., 0.5785, 0.6935, 0.4198]),
size=(5000, 5000), nnz=2500, layout=torch.sparse_csr)
tensor([0.6842, 0.2217, 0.0992, ..., 0.1824, 0.3701, 0.4149])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 2500
Density: 0.0001
Time: 0.11520600318908691 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 91141 -ss 5000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 9.897401094436646}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.018892765045166016}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 2499, 2500, 2500]),
col_indices=tensor([1451, 2006, 3586, ..., 3975, 4446, 2086]),
values=tensor([0.6609, 0.8356, 0.1353, ..., 0.7408, 0.3224, 0.8471]),
col_indices=tensor([3456, 1605, 749, ..., 2516, 837, 4620]),
values=tensor([0.8429, 0.4221, 0.2092, ..., 0.3256, 0.3578, 0.9398]),
size=(5000, 5000), nnz=2500, layout=torch.sparse_csr)
tensor([0.2892, 0.1223, 0.3419, ..., 0.7884, 0.7802, 0.0113])
tensor([0.1595, 0.2560, 0.8545, ..., 0.4673, 0.4412, 0.6412])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -34,18 +15,19 @@ Rows: 5000
Size: 25000000
NNZ: 2500
Density: 0.0001
Time: 9.897401094436646 seconds
Time: 0.018892765045166016 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 96690 -ss 5000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.663233041763306}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 55576 -ss 5000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 5.934850692749023}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 2, ..., 2499, 2499, 2500]),
col_indices=tensor([2205, 2444, 2425, ..., 3761, 2544, 2990]),
values=tensor([0.2656, 0.9114, 0.3983, ..., 0.8675, 0.7517, 0.7885]),
size=(5000, 5000), nnz=2500, layout=torch.sparse_csr)
tensor([0.1659, 0.6941, 0.7553, ..., 0.7483, 0.8019, 0.7277])
tensor(crow_indices=tensor([ 0, 2, 2, ..., 2499, 2500, 2500]),
col_indices=tensor([2304, 3497, 2599, ..., 3517, 2336, 3180]),
values=tensor([7.9793e-01, 1.3489e-04, 7.1193e-01, ...,
7.4115e-01, 8.0632e-01, 9.8789e-03]), size=(5000, 5000),
nnz=2500, layout=torch.sparse_csr)
tensor([0.4232, 0.5545, 0.0889, ..., 0.2237, 0.6245, 0.5041])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -53,15 +35,18 @@ Rows: 5000
Size: 25000000
NNZ: 2500
Density: 0.0001
Time: 10.663233041763306 seconds
Time: 5.934850692749023 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 98325 -ss 5000 -sd 0.0001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.519834756851196}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 2, ..., 2499, 2499, 2500]),
col_indices=tensor([2205, 2444, 2425, ..., 3761, 2544, 2990]),
values=tensor([0.2656, 0.9114, 0.3983, ..., 0.8675, 0.7517, 0.7885]),
tensor(crow_indices=tensor([ 0, 1, 1, ..., 2500, 2500, 2500]),
col_indices=tensor([ 417, 1523, 4116, ..., 1599, 2107, 3220]),
values=tensor([0.7284, 0.4903, 0.1270, ..., 0.3684, 0.2323, 0.2388]),
size=(5000, 5000), nnz=2500, layout=torch.sparse_csr)
tensor([0.1659, 0.6941, 0.7553, ..., 0.7483, 0.8019, 0.7277])
tensor([0.8570, 0.2399, 0.2271, ..., 0.1785, 0.2270, 0.3588])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -69,13 +54,29 @@ Rows: 5000
Size: 25000000
NNZ: 2500
Density: 0.0001
Time: 10.663233041763306 seconds
Time: 10.519834756851196 seconds
[16.32, 16.04, 16.32, 16.56, 16.56, 16.64, 16.96, 16.96, 16.76, 16.76]
[16.76, 16.88, 17.08, 18.96, 20.48, 21.96, 22.64, 22.36, 21.0, 19.8, 19.8, 19.6, 19.72, 19.8]
14.20075511932373
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 96690, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.663233041763306, 'TIME_S_1KI': 0.11028268736956569, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 262.53495136260995, 'W': 18.48739374467239}
[16.32, 16.04, 16.32, 16.56, 16.56, 16.64, 16.96, 16.96, 16.76, 16.76, 16.48, 16.4, 16.2, 16.28, 16.28, 16.12, 16.12, 16.28, 16.36, 16.36]
295.79999999999995
14.789999999999997
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 96690, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.663233041763306, 'TIME_S_1KI': 0.11028268736956569, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 262.53495136260995, 'W': 18.48739374467239, 'J_1KI': 2.7152234084456506, 'W_1KI': 0.19120274841940624, 'W_D': 3.6973937446723912, 'J_D': 52.505783147812, 'W_D_1KI': 0.038239670541652615, 'J_D_1KI': 0.0003954873362462779}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 2500, 2500, 2500]),
col_indices=tensor([ 417, 1523, 4116, ..., 1599, 2107, 3220]),
values=tensor([0.7284, 0.4903, 0.1270, ..., 0.3684, 0.2323, 0.2388]),
size=(5000, 5000), nnz=2500, layout=torch.sparse_csr)
tensor([0.8570, 0.2399, 0.2271, ..., 0.1785, 0.2270, 0.3588])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 2500
Density: 0.0001
Time: 10.519834756851196 seconds
[20.24, 20.24, 20.16, 20.08, 20.32, 20.6, 20.6, 20.56, 20.52, 20.6]
[20.68, 20.8, 21.12, 22.88, 24.6, 25.72, 26.32, 26.12, 25.12, 25.12, 23.36, 23.52, 23.52, 23.76]
14.24401044845581
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 98325, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.519834756851196, 'TIME_S_1KI': 0.10699043739487614, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 316.21768825531, 'W': 22.200046075478067}
[20.24, 20.24, 20.16, 20.08, 20.32, 20.6, 20.6, 20.56, 20.52, 20.6, 20.68, 20.68, 20.72, 20.52, 20.44, 20.56, 20.8, 20.84, 20.76, 20.76]
369.53999999999996
18.476999999999997
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 98325, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.519834756851196, 'TIME_S_1KI': 0.10699043739487614, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 316.21768825531, 'W': 22.200046075478067, 'J_1KI': 3.216045647142741, 'W_1KI': 0.22578231452304162, 'W_D': 3.72304607547807, 'J_D': 53.03110719919203, 'W_D_1KI': 0.03786469438574187, 'J_D_1KI': 0.0003850973240350051}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.001.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 17852, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.54783010482788, "TIME_S_1KI": 0.5908486502816425, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 261.2027643966674, "W": 18.408959018952206, "J_1KI": 14.6315686979984, "W_1KI": 1.0311986902841253, "W_D": 3.4579590189522076, "J_D": 49.0646132674217, "W_D_1KI": 0.19370149109075777, "J_D_1KI": 0.010850408418707023}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 17780, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.605318784713745, "TIME_S_1KI": 0.5964746223123591, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 313.57909806251524, "W": 22.019197605482734, "J_1KI": 17.636619688555413, "W_1KI": 1.2384250621756319, "W_D": 3.4661976054827335, "J_D": 49.36270332407948, "W_D_1KI": 0.19494924665257218, "J_D_1KI": 0.010964524558637355}

68
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.001.output
View File

@ -1,13 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.6220724582672119}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.06766819953918457}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 4, ..., 24988, 24993, 25000]),
col_indices=tensor([2208, 3192, 3630, ..., 2657, 2751, 4682]),
values=tensor([0.3516, 0.9043, 0.4344, ..., 0.9354, 0.2858, 0.8708]),
tensor(crow_indices=tensor([ 0, 7, 15, ..., 24992, 24996, 25000]),
col_indices=tensor([ 734, 800, 1880, ..., 3125, 3280, 3794]),
values=tensor([0.0540, 0.4911, 0.3592, ..., 0.2590, 0.5736, 0.3057]),
size=(5000, 5000), nnz=25000, layout=torch.sparse_csr)
tensor([0.1847, 0.5253, 0.6086, ..., 0.9552, 0.0514, 0.1920])
tensor([0.9823, 0.9343, 0.9377, ..., 0.0786, 0.0908, 0.1511])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -15,18 +15,18 @@ Rows: 5000
Size: 25000000
NNZ: 25000
Density: 0.001
Time: 0.6220724582672119 seconds
Time: 0.06766819953918457 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 16879 -ss 5000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 9.927400827407837}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 15516 -ss 5000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 9.162637948989868}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 7, 17, ..., 24988, 24992, 25000]),
col_indices=tensor([1765, 1880, 2380, ..., 3402, 4335, 4928]),
values=tensor([0.8113, 0.6065, 0.0419, ..., 0.8515, 0.2786, 0.9879]),
tensor(crow_indices=tensor([ 0, 8, 11, ..., 24988, 24995, 25000]),
col_indices=tensor([ 62, 227, 575, ..., 2337, 2631, 3700]),
values=tensor([0.5265, 0.4146, 0.5026, ..., 0.0706, 0.1241, 0.5991]),
size=(5000, 5000), nnz=25000, layout=torch.sparse_csr)
tensor([0.6729, 0.2847, 0.7618, ..., 0.5837, 0.8359, 0.7138])
tensor([0.6610, 0.4053, 0.0257, ..., 0.7779, 0.2973, 0.6422])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -34,18 +34,18 @@ Rows: 5000
Size: 25000000
NNZ: 25000
Density: 0.001
Time: 9.927400827407837 seconds
Time: 9.162637948989868 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 17852 -ss 5000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.54783010482788}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 17780 -ss 5000 -sd 0.001 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 25000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.605318784713745}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 5, 10, ..., 24988, 24991, 25000]),
col_indices=tensor([ 732, 2237, 2424, ..., 1459, 1662, 4133]),
values=tensor([0.5131, 0.9715, 0.7721, ..., 0.9714, 0.5730, 0.7149]),
tensor(crow_indices=tensor([ 0, 6, 8, ..., 24994, 24997, 25000]),
col_indices=tensor([ 423, 1662, 2124, ..., 288, 1379, 2658]),
values=tensor([0.1096, 0.1453, 0.3978, ..., 0.4089, 0.5724, 0.6122]),
size=(5000, 5000), nnz=25000, layout=torch.sparse_csr)
tensor([0.7581, 0.5458, 0.9932, ..., 0.3205, 0.5744, 0.9847])
tensor([0.2174, 0.6127, 0.5782, ..., 0.6057, 0.7055, 0.7233])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -53,15 +53,15 @@ Rows: 5000
Size: 25000000
NNZ: 25000
Density: 0.001
Time: 10.54783010482788 seconds
Time: 10.605318784713745 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 5, 10, ..., 24988, 24991, 25000]),
col_indices=tensor([ 732, 2237, 2424, ..., 1459, 1662, 4133]),
values=tensor([0.5131, 0.9715, 0.7721, ..., 0.9714, 0.5730, 0.7149]),
tensor(crow_indices=tensor([ 0, 6, 8, ..., 24994, 24997, 25000]),
col_indices=tensor([ 423, 1662, 2124, ..., 288, 1379, 2658]),
values=tensor([0.1096, 0.1453, 0.3978, ..., 0.4089, 0.5724, 0.6122]),
size=(5000, 5000), nnz=25000, layout=torch.sparse_csr)
tensor([0.7581, 0.5458, 0.9932, ..., 0.3205, 0.5744, 0.9847])
tensor([0.2174, 0.6127, 0.5782, ..., 0.6057, 0.7055, 0.7233])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -69,13 +69,13 @@ Rows: 5000
Size: 25000000
NNZ: 25000
Density: 0.001
Time: 10.54783010482788 seconds
Time: 10.605318784713745 seconds
[16.36, 16.36, 16.32, 16.6, 16.8, 17.04, 17.2, 17.08, 16.8, 16.52]
[16.48, 16.4, 17.36, 19.52, 19.52, 21.32, 21.84, 22.56, 21.0, 20.28, 19.68, 19.84, 19.92, 19.92]
14.188893795013428
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 17852, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.54783010482788, 'TIME_S_1KI': 0.5908486502816425, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 261.2027643966674, 'W': 18.408959018952206}
[16.36, 16.36, 16.32, 16.6, 16.8, 17.04, 17.2, 17.08, 16.8, 16.52, 16.4, 16.36, 16.32, 16.32, 16.76, 16.96, 16.72, 16.44, 16.2, 16.2]
299.02
14.950999999999999
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 17852, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.54783010482788, 'TIME_S_1KI': 0.5908486502816425, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 261.2027643966674, 'W': 18.408959018952206, 'J_1KI': 14.6315686979984, 'W_1KI': 1.0311986902841253, 'W_D': 3.4579590189522076, 'J_D': 49.0646132674217, 'W_D_1KI': 0.19370149109075777, 'J_D_1KI': 0.010850408418707023}
[20.64, 20.56, 20.28, 20.4, 20.28, 20.28, 20.36, 20.44, 20.4, 20.4]
[20.64, 20.56, 20.76, 22.16, 23.52, 25.36, 26.08, 26.2, 25.48, 23.84, 23.92, 23.84, 23.84, 23.88]
14.24116826057434
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 17780, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.605318784713745, 'TIME_S_1KI': 0.5964746223123591, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 313.57909806251524, 'W': 22.019197605482734}
[20.64, 20.56, 20.28, 20.4, 20.28, 20.28, 20.36, 20.44, 20.4, 20.4, 20.84, 20.68, 20.92, 20.96, 20.88, 20.88, 21.0, 20.8, 20.68, 20.64]
371.06
18.553
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 17780, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 25000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.605318784713745, 'TIME_S_1KI': 0.5964746223123591, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 313.57909806251524, 'W': 22.019197605482734, 'J_1KI': 17.636619688555413, 'W_1KI': 1.2384250621756319, 'W_D': 3.4661976054827335, 'J_D': 49.36270332407948, 'W_D_1KI': 0.19494924665257218, 'J_D_1KI': 0.010964524558637355}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.01.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1933, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.477078676223755, "TIME_S_1KI": 5.420113127896407, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 268.2967868804932, "W": 18.818848497168876, "J_1KI": 138.79813082281075, "W_1KI": 9.7355656995183, "W_D": 3.947848497168877, "J_D": 56.283734206199696, "W_D_1KI": 2.0423427300408057, "J_D_1KI": 1.0565663373206444}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1921, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.388477563858032, "TIME_S_1KI": 5.407848809920892, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 312.91259566307076, "W": 21.966357313024826, "J_1KI": 162.89047145396708, "W_1KI": 11.434855446655297, "W_D": 3.429357313024827, "J_D": 48.85148151707659, "W_D_1KI": 1.785193812089967, "J_D_1KI": 0.9293044310723411}

74
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.01.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.01, "TIME_S": 5.431562900543213}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.01, "TIME_S": 0.593717098236084}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 36, 79, ..., 249900, 249941,
tensor(crow_indices=tensor([ 0, 39, 88, ..., 249897, 249951,
250000]),
col_indices=tensor([ 80, 388, 404, ..., 4737, 4807, 4857]),
values=tensor([0.4885, 0.5213, 0.1721, ..., 0.5810, 0.1625, 0.7107]),
col_indices=tensor([ 1, 41, 120, ..., 4868, 4902, 4963]),
values=tensor([0.6487, 0.6379, 0.3189, ..., 0.3941, 0.1960, 0.9453]),
size=(5000, 5000), nnz=250000, layout=torch.sparse_csr)
tensor([0.1545, 0.4718, 0.9539, ..., 0.2261, 0.6017, 0.7355])
tensor([0.9493, 0.7713, 0.4212, ..., 0.5345, 0.1694, 0.1229])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -16,19 +16,19 @@ Rows: 5000
Size: 25000000
NNZ: 250000
Density: 0.01
Time: 5.431562900543213 seconds
Time: 0.593717098236084 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1933 -ss 5000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.477078676223755}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1768 -ss 5000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.01, "TIME_S": 9.65909719467163}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 40, 89, ..., 249917, 249964,
tensor(crow_indices=tensor([ 0, 53, 105, ..., 249907, 249948,
250000]),
col_indices=tensor([ 165, 177, 195, ..., 4656, 4719, 4927]),
values=tensor([0.2100, 0.9405, 0.2582, ..., 0.7931, 0.5258, 0.8197]),
col_indices=tensor([ 103, 261, 471, ..., 4857, 4933, 4959]),
values=tensor([0.8889, 0.3073, 0.1638, ..., 0.6109, 0.3049, 0.0052]),
size=(5000, 5000), nnz=250000, layout=torch.sparse_csr)
tensor([0.9173, 0.2185, 0.4076, ..., 0.3362, 0.1795, 0.2923])
tensor([0.5335, 0.0728, 0.9615, ..., 0.8926, 0.1348, 0.8188])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -36,16 +36,19 @@ Rows: 5000
Size: 25000000
NNZ: 250000
Density: 0.01
Time: 10.477078676223755 seconds
Time: 9.65909719467163 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1921 -ss 5000 -sd 0.01 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.388477563858032}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 40, 89, ..., 249917, 249964,
tensor(crow_indices=tensor([ 0, 65, 98, ..., 249897, 249948,
250000]),
col_indices=tensor([ 165, 177, 195, ..., 4656, 4719, 4927]),
values=tensor([0.2100, 0.9405, 0.2582, ..., 0.7931, 0.5258, 0.8197]),
col_indices=tensor([ 141, 179, 219, ..., 4719, 4923, 4985]),
values=tensor([0.6589, 0.9882, 0.9555, ..., 0.3007, 0.0365, 0.3378]),
size=(5000, 5000), nnz=250000, layout=torch.sparse_csr)
tensor([0.9173, 0.2185, 0.4076, ..., 0.3362, 0.1795, 0.2923])
tensor([0.9859, 0.3282, 0.7924, ..., 0.6550, 0.5905, 0.4141])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -53,13 +56,30 @@ Rows: 5000
Size: 25000000
NNZ: 250000
Density: 0.01
Time: 10.477078676223755 seconds
Time: 10.388477563858032 seconds
[17.04, 17.04, 16.96, 17.04, 16.64, 16.4, 16.44, 16.32, 16.24, 16.36]
[16.44, 16.52, 17.76, 19.52, 21.52, 21.52, 22.36, 22.76, 21.68, 21.36, 19.84, 20.04, 20.0, 20.08]
14.25681209564209
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1933, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.477078676223755, 'TIME_S_1KI': 5.420113127896407, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 268.2967868804932, 'W': 18.818848497168876}
[17.04, 17.04, 16.96, 17.04, 16.64, 16.4, 16.44, 16.32, 16.24, 16.36, 16.4, 16.24, 16.32, 16.36, 16.36, 16.48, 16.68, 16.56, 16.4, 16.08]
297.41999999999996
14.870999999999999
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1933, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.477078676223755, 'TIME_S_1KI': 5.420113127896407, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 268.2967868804932, 'W': 18.818848497168876, 'J_1KI': 138.79813082281075, 'W_1KI': 9.7355656995183, 'W_D': 3.947848497168877, 'J_D': 56.283734206199696, 'W_D_1KI': 2.0423427300408057, 'J_D_1KI': 1.0565663373206444}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 65, 98, ..., 249897, 249948,
250000]),
col_indices=tensor([ 141, 179, 219, ..., 4719, 4923, 4985]),
values=tensor([0.6589, 0.9882, 0.9555, ..., 0.3007, 0.0365, 0.3378]),
size=(5000, 5000), nnz=250000, layout=torch.sparse_csr)
tensor([0.9859, 0.3282, 0.7924, ..., 0.6550, 0.5905, 0.4141])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 250000
Density: 0.01
Time: 10.388477563858032 seconds
[21.12, 20.84, 20.6, 20.44, 20.44, 20.44, 20.52, 20.52, 20.6, 20.96]
[20.8, 20.8, 20.76, 21.68, 22.28, 24.84, 25.84, 26.2, 25.88, 24.96, 23.84, 23.92, 23.72, 23.88]
14.245083570480347
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1921, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.388477563858032, 'TIME_S_1KI': 5.407848809920892, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 312.91259566307076, 'W': 21.966357313024826}
[21.12, 20.84, 20.6, 20.44, 20.44, 20.44, 20.52, 20.52, 20.6, 20.96, 20.44, 20.44, 20.44, 20.48, 20.56, 20.68, 20.68, 20.56, 20.84, 20.8]
370.74
18.537
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1921, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.388477563858032, 'TIME_S_1KI': 5.407848809920892, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 312.91259566307076, 'W': 21.966357313024826, 'J_1KI': 162.89047145396708, 'W_1KI': 11.434855446655297, 'W_D': 3.429357313024827, 'J_D': 48.85148151707659, 'W_D_1KI': 1.785193812089967, 'J_D_1KI': 0.9293044310723411}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250000, "MATRIX_DENSITY": 0.05, "TIME_S": 26.830852508544922, "TIME_S_1KI": 26.830852508544922, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 724.3529372215271, "W": 23.831275335163394, "J_1KI": 724.3529372215271, "W_1KI": 23.831275335163394, "W_D": 5.2862753351633955, "J_D": 160.67663237214092, "W_D_1KI": 5.2862753351633955, "J_D_1KI": 5.2862753351633955}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 396, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.481398582458496, "TIME_S_1KI": 26.468178238531554, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 323.34643746376037, "W": 22.685982650996596, "J_1KI": 816.5314077367685, "W_1KI": 57.28783497726413, "W_D": 4.214982650996596, "J_D": 60.07672866272925, "W_D_1KI": 10.643895583324738, "J_D_1KI": 26.878524200314995}

64
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.05.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 0.05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250000, "MATRIX_DENSITY": 0.05, "TIME_S": 26.830852508544922}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250000, "MATRIX_DENSITY": 0.05, "TIME_S": 2.6507530212402344}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 229, 490, ..., 1249494,
1249740, 1250000]),
col_indices=tensor([ 18, 28, 58, ..., 4928, 4934, 4938]),
values=tensor([0.5574, 0.5312, 0.0435, ..., 0.2785, 0.4540, 0.5116]),
tensor(crow_indices=tensor([ 0, 267, 531, ..., 1249531,
1249748, 1250000]),
col_indices=tensor([ 12, 24, 45, ..., 4958, 4983, 4986]),
values=tensor([0.7384, 0.2434, 0.0755, ..., 0.4736, 0.1384, 0.4678]),
size=(5000, 5000), nnz=1250000, layout=torch.sparse_csr)
tensor([0.0667, 0.4882, 0.3630, ..., 0.8923, 0.8020, 0.4280])
tensor([0.2921, 0.5624, 0.4015, ..., 0.8005, 0.9400, 0.6114])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -16,16 +16,19 @@ Rows: 5000
Size: 25000000
NNZ: 1250000
Density: 0.05
Time: 26.830852508544922 seconds
Time: 2.6507530212402344 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 396 -ss 5000 -sd 0.05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.481398582458496}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 229, 490, ..., 1249494,
1249740, 1250000]),
col_indices=tensor([ 18, 28, 58, ..., 4928, 4934, 4938]),
values=tensor([0.5574, 0.5312, 0.0435, ..., 0.2785, 0.4540, 0.5116]),
tensor(crow_indices=tensor([ 0, 247, 505, ..., 1249488,
1249757, 1250000]),
col_indices=tensor([ 27, 35, 41, ..., 4930, 4938, 4952]),
values=tensor([0.8294, 0.9821, 0.6691, ..., 0.3905, 0.4873, 0.1672]),
size=(5000, 5000), nnz=1250000, layout=torch.sparse_csr)
tensor([0.0667, 0.4882, 0.3630, ..., 0.8923, 0.8020, 0.4280])
tensor([0.8352, 0.4457, 0.1150, ..., 0.9988, 0.2164, 0.9018])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -33,13 +36,30 @@ Rows: 5000
Size: 25000000
NNZ: 1250000
Density: 0.05
Time: 26.830852508544922 seconds
Time: 10.481398582458496 seconds
[20.52, 20.64, 20.8, 20.76, 20.84, 20.84, 20.8, 20.88, 20.8, 20.88]
[20.64, 20.72, 20.92, 24.36, 26.68, 28.84, 29.6, 27.08, 27.08, 26.84, 23.8, 24.0, 24.32, 24.44, 24.48, 24.52, 24.36, 24.12, 24.24, 24.2, 24.2, 24.28, 24.28, 24.28, 24.2, 24.28, 24.16, 24.12, 24.08, 24.08]
30.395055532455444
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 26.830852508544922, 'TIME_S_1KI': 26.830852508544922, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 724.3529372215271, 'W': 23.831275335163394}
[20.52, 20.64, 20.8, 20.76, 20.84, 20.84, 20.8, 20.88, 20.8, 20.88, 20.44, 20.56, 20.56, 20.6, 20.6, 20.48, 20.32, 20.2, 20.2, 20.2]
370.9
18.544999999999998
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 26.830852508544922, 'TIME_S_1KI': 26.830852508544922, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 724.3529372215271, 'W': 23.831275335163394, 'J_1KI': 724.3529372215271, 'W_1KI': 23.831275335163394, 'W_D': 5.2862753351633955, 'J_D': 160.67663237214092, 'W_D_1KI': 5.2862753351633955, 'J_D_1KI': 5.2862753351633955}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 247, 505, ..., 1249488,
1249757, 1250000]),
col_indices=tensor([ 27, 35, 41, ..., 4930, 4938, 4952]),
values=tensor([0.8294, 0.9821, 0.6691, ..., 0.3905, 0.4873, 0.1672]),
size=(5000, 5000), nnz=1250000, layout=torch.sparse_csr)
tensor([0.8352, 0.4457, 0.1150, ..., 0.9988, 0.2164, 0.9018])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 1250000
Density: 0.05
Time: 10.481398582458496 seconds
[20.72, 20.84, 20.92, 20.68, 20.6, 20.44, 20.48, 20.24, 20.44, 20.24]
[20.24, 20.52, 20.72, 22.36, 24.28, 26.72, 27.0, 27.56, 26.28, 25.88, 24.68, 24.52, 24.36, 24.36]
14.253137826919556
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 396, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.481398582458496, 'TIME_S_1KI': 26.468178238531554, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 323.34643746376037, 'W': 22.685982650996596}
[20.72, 20.84, 20.92, 20.68, 20.6, 20.44, 20.48, 20.24, 20.44, 20.24, 20.24, 20.24, 20.36, 20.48, 20.48, 20.6, 20.68, 20.64, 20.4, 20.6]
369.42
18.471
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 396, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.481398582458496, 'TIME_S_1KI': 26.468178238531554, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 323.34643746376037, 'W': 22.685982650996596, 'J_1KI': 816.5314077367685, 'W_1KI': 57.28783497726413, 'W_D': 4.214982650996596, 'J_D': 60.07672866272925, 'W_D_1KI': 10.643895583324738, 'J_D_1KI': 26.878524200314995}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.1.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.1, "TIME_S": 52.60684037208557, "TIME_S_1KI": 52.60684037208557, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1396.747187700272, "W": 24.19749919495284, "J_1KI": 1396.747187700272, "W_1KI": 24.19749919495284, "W_D": 5.574499194952839, "J_D": 321.77565171742464, "W_D_1KI": 5.574499194952839, "J_D_1KI": 5.574499194952839}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 199, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.1, "TIME_S": 10.477930784225464, "TIME_S_1KI": 52.6529185136958, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 366.7794834327698, "W": 23.985504753820422, "J_1KI": 1843.11298207422, "W_1KI": 120.53017464231368, "W_D": 5.225504753820424, "J_D": 79.90692520141607, "W_D_1KI": 26.25881785839409, "J_D_1KI": 131.95385858489493}

64
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.1.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 0.1 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.1, "TIME_S": 52.60684037208557}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.1 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.1, "TIME_S": 5.267488241195679}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 495, 1009, ..., 2499036,
2499501, 2500000]),
col_indices=tensor([ 3, 5, 7, ..., 4975, 4989, 4996]),
values=tensor([0.3153, 0.9204, 0.9402, ..., 0.8644, 0.1243, 0.2567]),
tensor(crow_indices=tensor([ 0, 489, 972, ..., 2499002,
2499515, 2500000]),
col_indices=tensor([ 0, 4, 21, ..., 4965, 4988, 4998]),
values=tensor([0.4985, 0.2439, 0.0801, ..., 0.3726, 0.6532, 0.2308]),
size=(5000, 5000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.0882, 0.6551, 0.2953, ..., 0.5102, 0.2382, 0.8339])
tensor([0.7620, 0.1310, 0.6898, ..., 0.4324, 0.6267, 0.4614])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -16,16 +16,19 @@ Rows: 5000
Size: 25000000
NNZ: 2500000
Density: 0.1
Time: 52.60684037208557 seconds
Time: 5.267488241195679 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 199 -ss 5000 -sd 0.1 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.1, "TIME_S": 10.477930784225464}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 495, 1009, ..., 2499036,
2499501, 2500000]),
col_indices=tensor([ 3, 5, 7, ..., 4975, 4989, 4996]),
values=tensor([0.3153, 0.9204, 0.9402, ..., 0.8644, 0.1243, 0.2567]),
tensor(crow_indices=tensor([ 0, 479, 980, ..., 2498983,
2499479, 2500000]),
col_indices=tensor([ 7, 13, 23, ..., 4987, 4988, 4998]),
values=tensor([0.4519, 0.3203, 0.6830, ..., 0.2361, 0.6866, 0.7928]),
size=(5000, 5000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.0882, 0.6551, 0.2953, ..., 0.5102, 0.2382, 0.8339])
tensor([0.4502, 0.7188, 0.8112, ..., 0.2797, 0.2285, 0.9848])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -33,13 +36,30 @@ Rows: 5000
Size: 25000000
NNZ: 2500000
Density: 0.1
Time: 52.60684037208557 seconds
Time: 10.477930784225464 seconds
[20.32, 20.32, 20.52, 20.36, 20.68, 20.64, 20.64, 20.84, 20.84, 20.64]
[21.04, 20.8, 23.84, 25.56, 28.32, 28.32, 30.0, 30.76, 27.04, 26.12, 23.96, 24.04, 23.96, 24.4, 24.48, 24.44, 24.4, 24.4, 24.4, 24.44, 24.52, 24.64, 24.44, 24.48, 24.36, 24.52, 24.36, 24.36, 24.28, 24.28, 24.2, 24.2, 24.04, 24.12, 24.04, 24.04, 24.08, 24.16, 24.12, 23.96, 24.0, 24.04, 24.04, 23.92, 23.92, 24.16, 24.32, 24.52, 24.68, 24.64, 24.24, 24.12, 24.2, 24.2, 24.2, 24.36, 24.36]
57.72279095649719
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 52.60684037208557, 'TIME_S_1KI': 52.60684037208557, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1396.747187700272, 'W': 24.19749919495284}
[20.32, 20.32, 20.52, 20.36, 20.68, 20.64, 20.64, 20.84, 20.84, 20.64, 20.48, 20.4, 20.44, 20.76, 20.76, 20.96, 21.12, 21.08, 21.0, 20.76]
372.46000000000004
18.623
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 52.60684037208557, 'TIME_S_1KI': 52.60684037208557, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1396.747187700272, 'W': 24.19749919495284, 'J_1KI': 1396.747187700272, 'W_1KI': 24.19749919495284, 'W_D': 5.574499194952839, 'J_D': 321.77565171742464, 'W_D_1KI': 5.574499194952839, 'J_D_1KI': 5.574499194952839}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 479, 980, ..., 2498983,
2499479, 2500000]),
col_indices=tensor([ 7, 13, 23, ..., 4987, 4988, 4998]),
values=tensor([0.4519, 0.3203, 0.6830, ..., 0.2361, 0.6866, 0.7928]),
size=(5000, 5000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.4502, 0.7188, 0.8112, ..., 0.2797, 0.2285, 0.9848])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 2500000
Density: 0.1
Time: 10.477930784225464 seconds
[20.56, 20.84, 21.36, 21.88, 21.64, 21.6, 21.4, 20.76, 20.76, 20.72]
[21.16, 21.2, 21.4, 25.96, 27.32, 30.6, 31.24, 28.68, 27.4, 26.08, 24.2, 24.2, 24.4, 24.36, 24.2]
15.291714191436768
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 199, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 10.477930784225464, 'TIME_S_1KI': 52.6529185136958, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 366.7794834327698, 'W': 23.985504753820422}
[20.56, 20.84, 21.36, 21.88, 21.64, 21.6, 21.4, 20.76, 20.76, 20.72, 20.44, 20.32, 20.44, 20.64, 20.64, 20.44, 20.4, 20.36, 20.44, 20.84]
375.2
18.759999999999998
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 199, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 10.477930784225464, 'TIME_S_1KI': 52.6529185136958, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 366.7794834327698, 'W': 23.985504753820422, 'J_1KI': 1843.11298207422, 'W_1KI': 120.53017464231368, 'W_D': 5.225504753820424, 'J_D': 79.90692520141607, 'W_D_1KI': 26.25881785839409, 'J_D_1KI': 131.95385858489493}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.2.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.2, "TIME_S": 105.2479407787323, "TIME_S_1KI": 105.2479407787323, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2877.626370677949, "W": 24.097306664358964, "J_1KI": 2877.626370677949, "W_1KI": 24.097306664358964, "W_D": 5.6223066643589625, "J_D": 671.3985984718811, "W_D_1KI": 5.6223066643589625, "J_D_1KI": 5.6223066643589625}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.2, "TIME_S": 10.774195671081543, "TIME_S_1KI": 107.74195671081543, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 357.5530287361145, "W": 23.422029760012947, "J_1KI": 3575.5302873611454, "W_1KI": 234.22029760012947, "W_D": 5.2480297600129475, "J_D": 80.11470204830175, "W_D_1KI": 52.480297600129475, "J_D_1KI": 524.8029760012947}

44
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.2.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 0.2 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.2, "TIME_S": 105.2479407787323}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.2 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.2, "TIME_S": 10.774195671081543}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 987, 1961, ..., 4998062,
4998993, 5000000]),
col_indices=tensor([ 2, 8, 14, ..., 4993, 4994, 4998]),
values=tensor([0.2058, 0.7964, 0.8021, ..., 0.0689, 0.5253, 0.0161]),
tensor(crow_indices=tensor([ 0, 1044, 1986, ..., 4998023,
4998990, 5000000]),
col_indices=tensor([ 2, 11, 17, ..., 4984, 4985, 4991]),
values=tensor([0.4872, 0.8747, 0.2341, ..., 0.7866, 0.4499, 0.5164]),
size=(5000, 5000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.1259, 0.8957, 0.2222, ..., 0.6970, 0.5570, 0.6933])
tensor([0.5529, 0.0016, 0.5040, ..., 0.3915, 0.6771, 0.4202])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -16,16 +16,16 @@ Rows: 5000
Size: 25000000
NNZ: 5000000
Density: 0.2
Time: 105.2479407787323 seconds
Time: 10.774195671081543 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 987, 1961, ..., 4998062,
4998993, 5000000]),
col_indices=tensor([ 2, 8, 14, ..., 4993, 4994, 4998]),
values=tensor([0.2058, 0.7964, 0.8021, ..., 0.0689, 0.5253, 0.0161]),
tensor(crow_indices=tensor([ 0, 1044, 1986, ..., 4998023,
4998990, 5000000]),
col_indices=tensor([ 2, 11, 17, ..., 4984, 4985, 4991]),
values=tensor([0.4872, 0.8747, 0.2341, ..., 0.7866, 0.4499, 0.5164]),
size=(5000, 5000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.1259, 0.8957, 0.2222, ..., 0.6970, 0.5570, 0.6933])
tensor([0.5529, 0.0016, 0.5040, ..., 0.3915, 0.6771, 0.4202])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -33,13 +33,13 @@ Rows: 5000
Size: 25000000
NNZ: 5000000
Density: 0.2
Time: 105.2479407787323 seconds
Time: 10.774195671081543 seconds
[20.6, 20.48, 20.6, 20.6, 20.72, 20.72, 20.68, 20.68, 20.52, 20.76]
[20.8, 20.6, 20.6, 24.4, 25.4, 28.96, 30.96, 31.56, 28.6, 27.8, 25.4, 24.24, 24.24, 24.24, 24.24, 24.24, 24.16, 24.04, 24.12, 24.2, 24.2, 24.28, 24.36, 24.2, 24.32, 24.4, 24.56, 24.56, 24.56, 24.44, 24.44, 24.24, 24.2, 24.16, 24.04, 24.32, 24.24, 24.32, 24.4, 24.4, 24.36, 24.4, 24.6, 24.8, 24.72, 24.88, 24.88, 24.64, 24.48, 24.48, 24.2, 24.12, 24.12, 24.28, 24.48, 24.56, 24.56, 24.6, 24.28, 24.16, 24.16, 24.04, 24.08, 24.24, 24.24, 24.64, 24.72, 24.6, 24.48, 24.12, 24.16, 24.08, 24.16, 24.2, 23.84, 23.92, 23.92, 23.92, 23.76, 24.2, 24.16, 24.28, 24.64, 24.44, 24.36, 24.52, 24.36, 24.4, 24.48, 24.48, 24.56, 24.56, 24.56, 24.36, 24.36, 24.2, 24.32, 24.08, 24.16, 24.24, 24.4, 24.4, 24.44, 24.68, 24.56, 24.4, 24.28, 24.4, 24.32, 24.4, 24.6, 24.48, 24.44, 24.6, 24.6, 24.48, 24.28, 24.24]
119.4169294834137
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 105.2479407787323, 'TIME_S_1KI': 105.2479407787323, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2877.626370677949, 'W': 24.097306664358964}
[20.6, 20.48, 20.6, 20.6, 20.72, 20.72, 20.68, 20.68, 20.52, 20.76, 20.2, 20.32, 20.2, 20.12, 20.32, 20.24, 20.6, 20.72, 20.8, 20.8]
369.5
18.475
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 105.2479407787323, 'TIME_S_1KI': 105.2479407787323, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2877.626370677949, 'W': 24.097306664358964, 'J_1KI': 2877.626370677949, 'W_1KI': 24.097306664358964, 'W_D': 5.6223066643589625, 'J_D': 671.3985984718811, 'W_D_1KI': 5.6223066643589625, 'J_D_1KI': 5.6223066643589625}
[20.36, 20.12, 20.12, 20.04, 20.28, 20.12, 20.08, 20.4, 20.4, 20.72]
[20.8, 21.08, 22.4, 23.32, 25.48, 25.48, 27.96, 29.08, 28.16, 27.04, 25.64, 24.2, 24.2, 24.4, 24.44]
15.265672206878662
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 10.774195671081543, 'TIME_S_1KI': 107.74195671081543, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 357.5530287361145, 'W': 23.422029760012947}
[20.36, 20.12, 20.12, 20.04, 20.28, 20.12, 20.08, 20.4, 20.4, 20.72, 20.32, 20.16, 20.04, 20.08, 20.08, 20.0, 20.16, 20.28, 20.28, 20.28]
363.48
18.174
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 10.774195671081543, 'TIME_S_1KI': 107.74195671081543, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 357.5530287361145, 'W': 23.422029760012947, 'J_1KI': 3575.5302873611454, 'W_1KI': 234.22029760012947, 'W_D': 5.2480297600129475, 'J_D': 80.11470204830175, 'W_D_1KI': 52.480297600129475, 'J_D_1KI': 524.8029760012947}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.3.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 1000, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 7500000, "MATRIX_DENSITY": 0.3, "TIME_S": 171.51510739326477, "TIME_S_1KI": 171.51510739326477, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 4017.952876434326, "W": 24.359705854077962, "J_1KI": 4017.9528764343263, "W_1KI": 24.359705854077962, "W_D": 5.79370585407796, "J_D": 955.6288257770532, "W_D_1KI": 5.79370585407796, "J_D_1KI": 5.79370585407796}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 7500000, "MATRIX_DENSITY": 0.3, "TIME_S": 15.919427633285522, "TIME_S_1KI": 159.19427633285522, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 506.52752479553226, "W": 23.776826253573663, "J_1KI": 5065.275247955323, "W_1KI": 237.76826253573662, "W_D": 5.445826253573667, "J_D": 116.01468014574058, "W_D_1KI": 54.45826253573667, "J_D_1KI": 544.5826253573666}

44
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.3.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 0.3 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 7500000, "MATRIX_DENSITY": 0.3, "TIME_S": 171.51510739326477}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.3 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 7500000, "MATRIX_DENSITY": 0.3, "TIME_S": 15.919427633285522}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1461, 2933, ..., 7497082,
7498527, 7500000]),
col_indices=tensor([ 0, 1, 3, ..., 4992, 4997, 4999]),
values=tensor([0.2348, 0.4295, 0.7390, ..., 0.0266, 0.0621, 0.6356]),
tensor(crow_indices=tensor([ 0, 1464, 2929, ..., 7497018,
7498512, 7500000]),
col_indices=tensor([ 1, 9, 13, ..., 4985, 4989, 4990]),
values=tensor([0.4014, 0.1905, 0.8906, ..., 0.4332, 0.9731, 0.1283]),
size=(5000, 5000), nnz=7500000, layout=torch.sparse_csr)
tensor([0.3584, 0.9157, 0.1902, ..., 0.2272, 0.0135, 0.3908])
tensor([0.5776, 0.8031, 0.5959, ..., 0.3626, 0.0858, 0.0842])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -16,16 +16,16 @@ Rows: 5000
Size: 25000000
NNZ: 7500000
Density: 0.3
Time: 171.51510739326477 seconds
Time: 15.919427633285522 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1461, 2933, ..., 7497082,
7498527, 7500000]),
col_indices=tensor([ 0, 1, 3, ..., 4992, 4997, 4999]),
values=tensor([0.2348, 0.4295, 0.7390, ..., 0.0266, 0.0621, 0.6356]),
tensor(crow_indices=tensor([ 0, 1464, 2929, ..., 7497018,
7498512, 7500000]),
col_indices=tensor([ 1, 9, 13, ..., 4985, 4989, 4990]),
values=tensor([0.4014, 0.1905, 0.8906, ..., 0.4332, 0.9731, 0.1283]),
size=(5000, 5000), nnz=7500000, layout=torch.sparse_csr)
tensor([0.3584, 0.9157, 0.1902, ..., 0.2272, 0.0135, 0.3908])
tensor([0.5776, 0.8031, 0.5959, ..., 0.3626, 0.0858, 0.0842])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -33,13 +33,13 @@ Rows: 5000
Size: 25000000
NNZ: 7500000
Density: 0.3
Time: 171.51510739326477 seconds
Time: 15.919427633285522 seconds
[21.16, 20.96, 20.88, 20.76, 20.72, 20.72, 20.72, 20.6, 20.56, 20.56]
[20.6, 20.48, 20.88, 21.96, 23.44, 26.4, 27.84, 28.8, 28.48, 27.04, 25.6, 24.28, 24.28, 24.48, 24.64, 24.64, 24.6, 24.6, 24.4, 24.24, 24.36, 24.48, 24.44, 24.52, 24.6, 24.6, 24.96, 25.04, 25.04, 25.12, 24.68, 24.56, 24.64, 24.6, 24.68, 24.92, 24.84, 24.84, 24.56, 24.48, 24.8, 24.56, 24.76, 24.76, 24.68, 24.64, 24.72, 24.64, 24.6, 24.6, 24.68, 24.68, 24.56, 24.48, 24.6, 24.48, 24.48, 24.56, 24.64, 24.52, 24.48, 24.6, 24.6, 24.52, 24.24, 24.32, 24.44, 24.32, 24.44, 24.44, 24.64, 24.84, 24.6, 24.76, 24.76, 24.8, 24.92, 25.08, 25.04, 24.92, 24.52, 24.6, 24.56, 24.6, 24.48, 24.6, 24.56, 24.56, 24.4, 24.36, 24.48, 24.48, 24.64, 24.64, 24.52, 24.56, 24.52, 24.44, 24.52, 24.52, 24.52, 24.48, 24.32, 24.52, 24.84, 25.04, 25.04, 25.0, 24.92, 24.68, 24.36, 24.36, 24.36, 24.24, 24.36, 24.36, 24.8, 24.72, 24.84, 24.68, 24.44, 24.56, 24.64, 24.72, 24.72, 25.16, 25.56, 25.68, 25.56, 25.36, 24.88, 24.8, 24.72, 24.52, 24.48, 24.6, 24.6, 24.6, 24.44, 24.44, 24.36, 24.44, 24.56, 24.56, 24.76, 24.64, 24.48, 24.72, 24.72, 24.72, 24.56, 24.92, 24.8, 24.56, 24.72, 24.8, 24.88, 24.88, 24.96, 24.84, 24.6, 24.6, 24.16]
164.9425859451294
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 7500000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 171.51510739326477, 'TIME_S_1KI': 171.51510739326477, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 4017.952876434326, 'W': 24.359705854077962}
[21.16, 20.96, 20.88, 20.76, 20.72, 20.72, 20.72, 20.6, 20.56, 20.56, 20.0, 20.24, 20.64, 20.6, 20.8, 20.76, 20.44, 20.48, 20.4, 20.36]
371.32000000000005
18.566000000000003
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 1000, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 7500000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 171.51510739326477, 'TIME_S_1KI': 171.51510739326477, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 4017.952876434326, 'W': 24.359705854077962, 'J_1KI': 4017.9528764343263, 'W_1KI': 24.359705854077962, 'W_D': 5.79370585407796, 'J_D': 955.6288257770532, 'W_D_1KI': 5.79370585407796, 'J_D_1KI': 5.79370585407796}
[20.2, 20.32, 20.32, 20.32, 20.32, 20.24, 20.52, 20.56, 20.64, 20.76]
[20.76, 20.64, 21.8, 21.8, 22.88, 24.6, 28.28, 30.28, 29.56, 29.28, 26.08, 25.32, 24.68, 24.72, 24.64, 24.44, 24.44, 24.48, 24.24, 24.32, 24.48]
21.303411960601807
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 7500000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 15.919427633285522, 'TIME_S_1KI': 159.19427633285522, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 506.52752479553226, 'W': 23.776826253573663}
[20.2, 20.32, 20.32, 20.32, 20.32, 20.24, 20.52, 20.56, 20.64, 20.76, 20.12, 20.12, 19.96, 20.2, 20.04, 20.4, 20.6, 20.56, 20.64, 20.64]
366.61999999999995
18.330999999999996
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 7500000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 15.919427633285522, 'TIME_S_1KI': 159.19427633285522, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 506.52752479553226, 'W': 23.776826253573663, 'J_1KI': 5065.275247955323, 'W_1KI': 237.76826253573662, 'W_D': 5.445826253573667, 'J_D': 116.01468014574058, 'W_D_1KI': 54.45826253573667, 'J_D_1KI': 544.5826253573666}

1
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.4.json Normal file
View File

@ -0,0 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.4, "TIME_S": 21.475390195846558, "TIME_S_1KI": 214.75390195846558, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 659.3497357940673, "W": 24.06026504427972, "J_1KI": 6593.497357940673, "W_1KI": 240.60265044279723, "W_D": 5.657265044279722, "J_D": 155.03221620368953, "W_D_1KI": 56.57265044279722, "J_D_1KI": 565.7265044279723}

45
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.4.output Normal file
View File

@ -0,0 +1,45 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.4 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.4, "TIME_S": 21.475390195846558}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2112, 4108, ..., 9995977,
9998003, 10000000]),
col_indices=tensor([ 0, 2, 5, ..., 4993, 4997, 4998]),
values=tensor([0.6521, 0.2294, 0.7060, ..., 0.9592, 0.5713, 0.6385]),
size=(5000, 5000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.2067, 0.4320, 0.3905, ..., 0.7782, 0.8244, 0.2696])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 10000000
Density: 0.4
Time: 21.475390195846558 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2112, 4108, ..., 9995977,
9998003, 10000000]),
col_indices=tensor([ 0, 2, 5, ..., 4993, 4997, 4998]),
values=tensor([0.6521, 0.2294, 0.7060, ..., 0.9592, 0.5713, 0.6385]),
size=(5000, 5000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.2067, 0.4320, 0.3905, ..., 0.7782, 0.8244, 0.2696])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 10000000
Density: 0.4
Time: 21.475390195846558 seconds
[20.36, 20.32, 20.12, 20.16, 20.12, 20.28, 20.68, 20.96, 21.08, 21.04]
[20.56, 20.44, 20.44, 23.96, 25.32, 27.12, 30.0, 32.28, 29.04, 28.16, 26.48, 25.72, 24.44, 24.36, 24.24, 24.24, 24.2, 24.16, 24.28, 24.32, 24.36, 24.48, 24.04, 23.92, 23.8, 23.68, 23.88]
27.40409278869629
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.4, 'TIME_S': 21.475390195846558, 'TIME_S_1KI': 214.75390195846558, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 659.3497357940673, 'W': 24.06026504427972}
[20.36, 20.32, 20.12, 20.16, 20.12, 20.28, 20.68, 20.96, 21.08, 21.04, 20.4, 20.4, 20.36, 20.56, 20.6, 20.4, 20.24, 20.36, 20.28, 20.48]
368.06
18.403
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.4, 'TIME_S': 21.475390195846558, 'TIME_S_1KI': 214.75390195846558, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 659.3497357940673, 'W': 24.06026504427972, 'J_1KI': 6593.497357940673, 'W_1KI': 240.60265044279723, 'W_D': 5.657265044279722, 'J_D': 155.03221620368953, 'W_D_1KI': 56.57265044279722, 'J_D_1KI': 565.7265044279723}

1
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.5.json Normal file
View File

@ -0,0 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 100, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 0.5, "TIME_S": 26.60726523399353, "TIME_S_1KI": 266.0726523399353, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 821.7415859985354, "W": 24.511735271206813, "J_1KI": 8217.415859985353, "W_1KI": 245.11735271206814, "W_D": 6.057735271206813, "J_D": 203.08203129005446, "W_D_1KI": 60.57735271206813, "J_D_1KI": 605.7735271206813}

45
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_0.5.output Normal file
View File

@ -0,0 +1,45 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 0.5 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 0.5, "TIME_S": 26.60726523399353}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2547, 5140, ..., 12494977,
12497506, 12500000]),
col_indices=tensor([ 3, 4, 6, ..., 4994, 4995, 4998]),
values=tensor([0.6176, 0.1216, 0.2065, ..., 0.5783, 0.0575, 0.3833]),
size=(5000, 5000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.3583, 0.1424, 0.2491, ..., 0.0607, 0.2583, 0.4693])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 12500000
Density: 0.5
Time: 26.60726523399353 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2547, 5140, ..., 12494977,
12497506, 12500000]),
col_indices=tensor([ 3, 4, 6, ..., 4994, 4995, 4998]),
values=tensor([0.6176, 0.1216, 0.2065, ..., 0.5783, 0.0575, 0.3833]),
size=(5000, 5000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.3583, 0.1424, 0.2491, ..., 0.0607, 0.2583, 0.4693])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 12500000
Density: 0.5
Time: 26.60726523399353 seconds
[20.28, 20.52, 20.68, 20.56, 20.44, 20.52, 20.44, 20.44, 20.72, 21.0]
[20.92, 20.68, 20.68, 23.84, 26.2, 27.84, 31.04, 29.96, 30.76, 29.48, 26.6, 26.6, 26.28, 24.68, 24.6, 24.48, 24.4, 24.24, 24.32, 24.32, 24.48, 24.52, 24.44, 24.52, 24.52, 24.56, 24.4, 24.56, 24.72, 24.68, 24.84, 24.8, 24.64]
33.524415016174316
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 0.5, 'TIME_S': 26.60726523399353, 'TIME_S_1KI': 266.0726523399353, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 821.7415859985354, 'W': 24.511735271206813}
[20.28, 20.52, 20.68, 20.56, 20.44, 20.52, 20.44, 20.44, 20.72, 21.0, 20.24, 20.2, 20.4, 20.6, 20.6, 20.6, 20.44, 20.56, 20.52, 20.16]
369.08000000000004
18.454
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 100, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 0.5, 'TIME_S': 26.60726523399353, 'TIME_S_1KI': 266.0726523399353, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 821.7415859985354, 'W': 24.511735271206813, 'J_1KI': 8217.415859985353, 'W_1KI': 245.11735271206814, 'W_D': 6.057735271206813, 'J_D': 203.08203129005446, 'W_D_1KI': 60.57735271206813, 'J_D_1KI': 605.7735271206813}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 293134, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.89356017112732, "TIME_S_1KI": 0.03716239048055606, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 278.4485914421082, "W": 19.57952781791354, "J_1KI": 0.9499020633638819, "W_1KI": 0.06679377969772711, "W_D": 4.613527817913537, "J_D": 65.610893910408, "W_D_1KI": 0.01573863085794735, "J_D_1KI": 5.3690908792386244e-05}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 284909, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.16942024230957, "TIME_S_1KI": 0.035693573184102885, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 316.9542731094361, "W": 22.304497332617817, "J_1KI": 1.1124754679895548, "W_1KI": 0.07828639085679223, "W_D": 3.6764973326178207, "J_D": 52.24424125194558, "W_D_1KI": 0.012904110900736098, "J_D_1KI": 4.529204377796454e-05}

664
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_1e-05.output
View File

@ -1,75 +1,75 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.04628562927246094}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.012842655181884766}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 250, 250, 250]),
col_indices=tensor([ 683, 1119, 1321, 2450, 3482, 3631, 1761, 3022, 756,
3517, 37, 3468, 4655, 1287, 913, 1692, 3561, 1823,
1971, 4332, 175, 242, 3518, 2634, 2163, 1929, 1347,
4194, 4673, 3242, 1554, 3336, 2363, 4819, 1624, 2276,
4446, 1440, 2278, 2820, 2808, 2194, 1293, 3294, 2532,
2630, 3533, 2517, 2068, 4492, 3196, 31, 2012, 3028,
3263, 1298, 1827, 4518, 2739, 383, 2502, 2163, 2983,
275, 3460, 724, 4585, 3927, 4513, 2645, 242, 1435,
3115, 1351, 1335, 3004, 3671, 2087, 2361, 3470, 3033,
1776, 2762, 2985, 544, 2787, 1009, 4955, 757, 4621,
3559, 4933, 3451, 2535, 2363, 1115, 250, 284, 3453,
4194, 4788, 4427, 434, 2792, 219, 1976, 286, 1619,
3123, 2185, 100, 1443, 2614, 3193, 4750, 1625, 61,
2975, 2813, 3271, 969, 1209, 2770, 2904, 1769, 343,
239, 3167, 403, 2400, 1507, 4176, 1210, 627, 332,
3526, 2019, 4707, 4667, 3689, 1411, 474, 2037, 1559,
3233, 2371, 3442, 4237, 1757, 4685, 2495, 737, 562,
4385, 4537, 1150, 2708, 4099, 4510, 4059, 58, 3153,
2292, 1450, 3200, 4511, 1556, 237, 2082, 3442, 4661,
3624, 407, 1680, 104, 2285, 3192, 1818, 2013, 2874,
4274, 1703, 393, 4638, 3642, 1595, 4200, 2976, 747,
1685, 436, 4175, 3319, 2858, 4687, 1967, 1550, 4498,
5, 3295, 2892, 3076, 2947, 1470, 2928, 4594, 372,
1505, 3795, 2014, 3988, 420, 2057, 4772, 3022, 3131,
376, 1473, 4703, 771, 759, 172, 3505, 2361, 168,
3559, 881, 3500, 894, 4238, 842, 291, 2606, 4128,
2513, 4919, 1689, 1039, 4346, 4963, 184, 2438, 3794,
631, 3050, 4745, 3174, 1910, 3181, 4415]),
values=tensor([0.5133, 0.9500, 0.6089, 0.1299, 0.0389, 0.7021, 0.0545,
0.1504, 0.2775, 0.3654, 0.5414, 0.7066, 0.5062, 0.9276,
0.5403, 0.1473, 0.0619, 0.8013, 0.5229, 0.9618, 0.3595,
0.9768, 0.4894, 0.9436, 0.2586, 0.3228, 0.7550, 0.4654,
0.5557, 0.6099, 0.1466, 0.3234, 0.9559, 0.4861, 0.6590,
0.2645, 0.3128, 0.2881, 0.8916, 0.9625, 0.3287, 0.6208,
0.1989, 0.4749, 0.6654, 0.5023, 0.5464, 0.6484, 0.8692,
0.5946, 0.3095, 0.4520, 0.2934, 0.1142, 0.3825, 0.0692,
0.4451, 0.9095, 0.2024, 0.8392, 0.4692, 0.1054, 0.2753,
0.1688, 0.2684, 0.5848, 0.9464, 0.6200, 0.5357, 0.5307,
0.7002, 0.6351, 0.9452, 0.4196, 0.3107, 0.9700, 0.4879,
0.0926, 0.0442, 0.1064, 0.9432, 0.8436, 0.3680, 0.1497,
0.1266, 0.6045, 0.6916, 0.0824, 0.1706, 0.8211, 0.8262,
0.7835, 0.0310, 0.3323, 0.1890, 0.5250, 0.8324, 0.5975,
0.0174, 0.0556, 0.9553, 0.6279, 0.3153, 0.4085, 0.9318,
0.3588, 0.1032, 0.7200, 0.2145, 0.8631, 0.4178, 0.0372,
0.7636, 0.4317, 0.2105, 0.2684, 0.0231, 0.6996, 0.0880,
0.2381, 0.6281, 0.3203, 0.4143, 0.7477, 0.1347, 0.5900,
0.7586, 0.5291, 0.6348, 0.4495, 0.3601, 0.9398, 0.3999,
0.2033, 0.1346, 0.0706, 0.9911, 0.9515, 0.0420, 0.6637,
0.2691, 0.3435, 0.7224, 0.4624, 0.4390, 0.3084, 0.3677,
0.2556, 0.8927, 0.7015, 0.4402, 0.6275, 0.9141, 0.3633,
0.0870, 0.2460, 0.1945, 0.8036, 0.3884, 0.5353, 0.6776,
0.4646, 0.1680, 0.4783, 0.9893, 0.5596, 0.0460, 0.9167,
0.8564, 0.2217, 0.2454, 0.6476, 0.0091, 0.6634, 0.6906,
0.5109, 0.0619, 0.8391, 0.3721, 0.4015, 0.1086, 0.8568,
0.0263, 0.0960, 0.2106, 0.8204, 0.3496, 0.0650, 0.0530,
0.2300, 0.7920, 0.0833, 0.8839, 0.6947, 0.7490, 0.6930,
0.4034, 0.9770, 0.5568, 0.5813, 0.4457, 0.4409, 0.3165,
0.4290, 0.8018, 0.4890, 0.7248, 0.5066, 0.4197, 0.9251,
0.4526, 0.8257, 0.6029, 0.9210, 0.8099, 0.1966, 0.6605,
0.5583, 0.0851, 0.2553, 0.8703, 0.6237, 0.8267, 0.9769,
0.6623, 0.5390, 0.0172, 0.1684, 0.4788, 0.5289, 0.0477,
0.8018, 0.0914, 0.3275, 0.7127, 0.1031, 0.8096, 0.1163,
0.3143, 0.8185, 0.2797, 0.8908, 0.1307, 0.5822, 0.2044,
0.6227, 0.4853, 0.6034, 0.6732, 0.0321]),
col_indices=tensor([1287, 2037, 612, 4005, 465, 4495, 4486, 1954, 4095,
1514, 3786, 3287, 3358, 3432, 3673, 489, 2823, 505,
4424, 1572, 4277, 474, 3301, 30, 2842, 4780, 2739,
564, 2900, 4485, 4784, 2295, 755, 3717, 1261, 1856,
2818, 3372, 3761, 1939, 4279, 1416, 4196, 1024, 159,
3430, 1464, 630, 4128, 1057, 4758, 4930, 4819, 4211,
3868, 1700, 2760, 4521, 1355, 4737, 4580, 1838, 4056,
1953, 4561, 1726, 3125, 4174, 510, 4743, 502, 2822,
338, 1706, 4412, 4712, 3417, 4607, 4478, 4287, 4365,
4223, 3755, 467, 2870, 999, 1516, 3711, 3345, 4540,
4303, 4477, 4047, 3188, 522, 451, 4048, 1301, 3760,
3807, 142, 526, 3797, 3415, 942, 3041, 1022, 555,
2433, 3440, 4291, 2481, 2516, 1226, 4664, 1242, 2239,
3542, 3300, 3985, 1261, 628, 3797, 3571, 1648, 545,
3417, 523, 297, 1814, 2339, 1387, 4149, 2499, 1698,
4107, 3910, 907, 1601, 3072, 2976, 1955, 76, 3173,
63, 633, 2089, 1360, 1226, 4574, 730, 2472, 4618,
425, 3915, 1299, 1950, 4945, 1796, 628, 1797, 3210,
2055, 2428, 876, 1161, 1529, 1660, 2886, 4614, 2062,
2433, 3539, 1521, 33, 1294, 4198, 863, 2582, 1498,
77, 507, 2697, 2034, 2514, 1935, 4132, 2925, 876,
2808, 4770, 271, 3697, 1635, 2519, 4995, 3590, 3245,
130, 480, 3111, 3121, 3132, 1937, 3910, 1943, 2562,
426, 3962, 1910, 1189, 1897, 1056, 462, 1607, 1444,
118, 191, 2005, 615, 1379, 633, 2360, 3526, 4732,
2267, 3397, 1029, 3432, 2182, 2675, 4099, 3777, 2171,
2640, 3913, 4300, 2946, 3758, 3305, 1103, 4800, 3668,
4286, 3562, 281, 919, 4442, 2167, 2728]),
values=tensor([0.8347, 0.3655, 0.0811, 0.8356, 0.0205, 0.3330, 0.9286,
0.0736, 0.7654, 0.8451, 0.0234, 0.4126, 0.2439, 0.1012,
0.1525, 0.4404, 0.8423, 0.5434, 0.2968, 0.3607, 0.9939,
0.0443, 0.6432, 0.5086, 0.6326, 0.2329, 0.7870, 0.7820,
0.9646, 0.4656, 0.9109, 0.0130, 0.3562, 0.2378, 0.0761,
0.1724, 0.0722, 0.8084, 0.1566, 0.8788, 0.9593, 0.2473,
0.2746, 0.1767, 0.8469, 0.1106, 0.8653, 0.5297, 0.8543,
0.5387, 0.4683, 0.0500, 0.6408, 0.2485, 0.5053, 0.9278,
0.6730, 0.1223, 0.9361, 0.1415, 0.0908, 0.6368, 0.4532,
0.7711, 0.1924, 0.7435, 0.0645, 0.3989, 0.7433, 0.7022,
0.6974, 0.8264, 0.3293, 0.6363, 0.9947, 0.1723, 0.3099,
0.5498, 0.6041, 0.9256, 0.6505, 0.2218, 0.5727, 0.8460,
0.3386, 0.9152, 0.1985, 0.3213, 0.2437, 0.8619, 0.4265,
0.8019, 0.3028, 0.4559, 0.9203, 0.9762, 0.2222, 0.3112,
0.4047, 0.0709, 0.2379, 0.3209, 0.9982, 0.9963, 0.6946,
0.0267, 0.7677, 0.2026, 0.6034, 0.5006, 0.8273, 0.2191,
0.6497, 0.2706, 0.0892, 0.8677, 0.9857, 0.5541, 0.2974,
0.1559, 0.1745, 0.4744, 0.1426, 0.1224, 0.3669, 0.1827,
0.5044, 0.5810, 0.3220, 0.7231, 0.9240, 0.0412, 0.3152,
0.9088, 0.3617, 0.9935, 0.3508, 0.0434, 0.0453, 0.5299,
0.2529, 0.0232, 0.7419, 0.0564, 0.5519, 0.6136, 0.5013,
0.9801, 0.4708, 0.5636, 0.5144, 0.1368, 0.7207, 0.1775,
0.9552, 0.2262, 0.7144, 0.1124, 0.8514, 0.1783, 0.8401,
0.1256, 0.7454, 0.1258, 0.2191, 0.5753, 0.9252, 0.8693,
0.6514, 0.3440, 0.7780, 0.4771, 0.0787, 0.5042, 0.0634,
0.8013, 0.8286, 0.4280, 0.3433, 0.9749, 0.0712, 0.9286,
0.0320, 0.8979, 0.5094, 0.4000, 0.4693, 0.8308, 0.6000,
0.3933, 0.7591, 0.2335, 0.5450, 0.3018, 0.3121, 0.4779,
0.9302, 0.5324, 0.1295, 0.6438, 0.5030, 0.3371, 0.9613,
0.8059, 0.9687, 0.2898, 0.7067, 0.8974, 0.1763, 0.0222,
0.0300, 0.9494, 0.3209, 0.6515, 0.7028, 0.8063, 0.2794,
0.7392, 0.1814, 0.3171, 0.4591, 0.7578, 0.6336, 0.8392,
0.6142, 0.8521, 0.4206, 0.9799, 0.4517, 0.1512, 0.3696,
0.0957, 0.3165, 0.3328, 0.9242, 0.5247, 0.8176, 0.9760,
0.3689, 0.9384, 0.3805, 0.7826, 0.4113, 0.3311, 0.7250,
0.9146, 0.3319, 0.6199, 0.8288, 0.1278]),
size=(5000, 5000), nnz=250, layout=torch.sparse_csr)
tensor([0.7002, 0.3467, 0.9676, ..., 0.8135, 0.6463, 0.9360])
tensor([0.0363, 0.1704, 0.8959, ..., 0.1381, 0.6314, 0.8045])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -77,188 +77,80 @@ Rows: 5000
Size: 25000000
NNZ: 250
Density: 1e-05
Time: 0.04628562927246094 seconds
Time: 0.012842655181884766 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 226852 -ss 5000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 8.58342981338501}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 250, 250, 250]),
col_indices=tensor([2568, 3647, 442, 965, 263, 2383, 651, 4423, 3036,
1922, 4223, 1539, 1097, 1063, 4112, 2393, 2048, 391,
3893, 324, 2099, 2458, 2688, 682, 96, 2079, 4917,
1561, 2320, 1455, 2135, 3126, 2991, 4240, 1021, 4993,
3258, 3975, 1172, 1489, 4782, 364, 2199, 94, 1257,
4686, 607, 1510, 89, 4888, 4165, 3842, 3018, 4662,
3670, 3231, 4131, 746, 49, 680, 3901, 1594, 359,
3311, 2321, 3005, 4317, 2855, 829, 3097, 2418, 1365,
3858, 1930, 3446, 1588, 4464, 2454, 3676, 2837, 1569,
2885, 1556, 3076, 4363, 2721, 3030, 172, 2121, 2698,
3156, 442, 947, 2541, 828, 1038, 4897, 3795, 2214,
609, 3658, 77, 3238, 2356, 89, 2253, 2806, 2065,
2259, 579, 2660, 1688, 2237, 2605, 1390, 4025, 2509,
2831, 635, 2338, 2347, 3405, 393, 82, 2030, 4203,
4365, 3211, 1439, 3151, 1397, 476, 3123, 1758, 2491,
252, 1078, 102, 4624, 527, 163, 2201, 1415, 53,
3597, 2281, 1819, 1693, 3944, 4697, 560, 1457, 1677,
2072, 2996, 1150, 4324, 2498, 4491, 2244, 3104, 2934,
632, 2182, 4187, 1162, 422, 1444, 3294, 1160, 1691,
2846, 266, 3519, 3656, 2923, 1457, 1651, 1147, 1014,
671, 3331, 4535, 2766, 1343, 680, 3907, 3255, 700,
2823, 4644, 4966, 3493, 4426, 2084, 2312, 2388, 1167,
2294, 2501, 1866, 3421, 4059, 858, 4657, 794, 658,
2225, 1411, 1995, 2476, 795, 2719, 1100, 685, 3038,
1607, 4350, 2782, 410, 2489, 2516, 1183, 2789, 4067,
4708, 2699, 3392, 4757, 4834, 2136, 1271, 2790, 3056,
2835, 3630, 2085, 603, 3829, 4234, 710, 378, 2071,
1558, 4206, 4361, 1063, 3780, 352, 168]),
values=tensor([6.8562e-01, 9.9314e-01, 3.4074e-01, 1.7233e-01,
1.4522e-02, 6.3720e-01, 5.5464e-02, 7.3826e-01,
1.5940e-01, 1.2632e-01, 2.2414e-01, 7.6966e-01,
6.9475e-01, 9.2958e-01, 3.8229e-01, 7.5368e-01,
7.6972e-01, 6.6374e-01, 5.6166e-01, 6.7113e-01,
2.6640e-01, 3.1404e-01, 8.1747e-01, 7.0390e-01,
3.3211e-02, 4.2381e-01, 1.8457e-01, 3.9280e-01,
7.9738e-01, 4.8542e-01, 5.6000e-01, 2.0755e-01,
7.0598e-01, 8.6707e-01, 1.7337e-01, 7.0748e-01,
9.7389e-01, 7.9562e-01, 6.7701e-01, 4.6490e-01,
5.4665e-01, 4.9560e-02, 5.8946e-01, 3.8658e-01,
3.0672e-01, 2.5947e-01, 8.6455e-01, 8.5056e-02,
3.3869e-01, 3.9093e-01, 5.9721e-01, 6.2207e-01,
8.8265e-01, 8.1640e-01, 1.7680e-01, 2.4072e-01,
3.6980e-01, 2.2490e-01, 6.0225e-01, 7.0554e-01,
8.5790e-01, 7.4936e-01, 1.7010e-01, 2.0063e-01,
1.1246e-01, 6.8727e-01, 6.8037e-01, 8.9757e-01,
3.8505e-01, 6.5721e-01, 9.3013e-01, 4.9507e-01,
7.9582e-01, 3.6413e-01, 6.2028e-01, 2.8858e-01,
2.8115e-01, 4.5974e-01, 9.8822e-01, 1.1635e-01,
5.8307e-01, 5.1420e-02, 1.1202e-01, 5.4531e-01,
7.6023e-01, 9.0514e-01, 5.3398e-01, 1.7667e-01,
9.2343e-01, 9.0805e-01, 9.6041e-01, 5.0364e-01,
2.4720e-01, 1.5194e-01, 2.2205e-01, 3.0452e-01,
6.8304e-02, 7.0941e-02, 2.3679e-01, 2.9428e-01,
2.6988e-01, 2.9905e-01, 9.7067e-01, 3.9498e-01,
4.5558e-01, 6.9955e-01, 5.3969e-02, 3.5860e-01,
7.2397e-01, 7.1675e-01, 8.0095e-01, 4.8315e-01,
4.1035e-01, 3.9824e-01, 5.0060e-01, 5.6947e-01,
2.5338e-01, 1.2799e-01, 9.1108e-01, 7.6016e-02,
8.5394e-01, 4.5257e-01, 4.8350e-01, 1.3291e-01,
2.2106e-01, 8.0845e-01, 6.7657e-01, 4.4898e-01,
6.6830e-01, 4.0859e-01, 8.4227e-01, 7.7311e-01,
5.4753e-01, 3.9804e-01, 9.4899e-01, 8.2056e-01,
7.7146e-01, 6.3508e-01, 6.2972e-01, 7.4169e-01,
7.8963e-01, 1.0699e-01, 5.7796e-01, 7.2429e-01,
6.3979e-02, 4.5238e-02, 6.3144e-01, 9.8512e-01,
5.1816e-01, 3.2546e-01, 8.7580e-01, 9.7697e-01,
4.6167e-01, 2.4042e-01, 1.1377e-01, 9.7747e-01,
7.4258e-01, 6.3887e-01, 7.3930e-01, 2.3402e-01,
4.1461e-01, 4.8691e-01, 2.7849e-01, 5.9673e-01,
8.6946e-02, 9.5615e-01, 9.7242e-01, 8.9092e-01,
4.1164e-01, 3.3893e-01, 9.4485e-01, 3.2960e-01,
7.1004e-01, 4.1240e-01, 1.1151e-01, 7.6114e-01,
5.5779e-01, 9.3723e-01, 2.2015e-01, 9.0422e-01,
2.5683e-01, 4.6041e-01, 3.3427e-01, 4.3355e-02,
3.1777e-01, 6.8533e-01, 4.9880e-01, 7.0528e-01,
6.2605e-01, 9.9580e-01, 3.8253e-02, 1.0464e-02,
6.2010e-02, 4.9009e-02, 8.8508e-01, 8.3043e-01,
9.5592e-01, 8.5708e-01, 5.1611e-01, 1.7460e-01,
4.5394e-01, 4.2516e-01, 8.0836e-01, 5.2242e-01,
8.0860e-01, 5.1184e-01, 7.3172e-01, 9.2625e-01,
3.8652e-01, 8.6518e-01, 6.9408e-01, 2.5732e-01,
6.0297e-01, 2.2091e-01, 1.2658e-02, 7.6721e-01,
9.6888e-02, 6.6146e-01, 4.4139e-01, 1.9043e-01,
1.1703e-04, 3.3229e-01, 3.7446e-01, 3.2871e-01,
5.5144e-01, 4.6404e-01, 7.9360e-01, 3.2754e-01,
9.8665e-01, 3.2413e-01, 4.6510e-01, 6.8652e-01,
9.6619e-01, 4.0817e-01, 5.0618e-01, 8.0048e-01,
3.4373e-01, 9.9556e-01, 1.4700e-01, 1.2820e-01,
8.0477e-01, 2.3035e-01, 5.4135e-01, 3.8689e-01,
1.8548e-01, 9.7019e-01, 2.2577e-01, 3.2056e-01,
4.1451e-02, 1.3423e-01]), size=(5000, 5000), nnz=250,
layout=torch.sparse_csr)
tensor([0.0595, 0.8939, 0.2592, ..., 0.5348, 0.8468, 0.6804])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
Rows: 5000
Size: 25000000
NNZ: 250
Density: 1e-05
Time: 8.58342981338501 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 277505 -ss 5000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 9.940149784088135}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 81758 -ss 5000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 3.01309871673584}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 250, 250, 250]),
col_indices=tensor([3014, 4957, 1583, 2867, 2783, 475, 2139, 2382, 3400,
1371, 4277, 2356, 2363, 809, 3070, 166, 954, 1491,
2451, 1189, 312, 609, 4247, 23, 459, 898, 2311,
4831, 338, 2271, 1779, 2454, 4584, 3113, 487, 1534,
2828, 4851, 633, 1451, 3532, 1285, 3207, 3942, 1871,
1291, 465, 1879, 4867, 2362, 2141, 1675, 4085, 954,
3823, 3407, 284, 572, 14, 2939, 2313, 3750, 1562,
2613, 2778, 2860, 3224, 2726, 239, 3475, 2082, 2253,
3516, 1146, 3276, 4995, 2558, 345, 3127, 2150, 75,
826, 1135, 4736, 4690, 2556, 910, 1899, 2387, 947,
695, 304, 2013, 2, 897, 3875, 3772, 1882, 451,
3308, 1440, 3959, 2068, 783, 1822, 1945, 4659, 2440,
4920, 2894, 4923, 1763, 2739, 4990, 4910, 2298, 1281,
1642, 4403, 354, 879, 3935, 4111, 1373, 3061, 948,
4840, 4778, 2992, 2315, 2233, 3168, 3973, 2138, 2299,
4743, 1438, 4906, 254, 4427, 953, 1389, 2612, 1867,
4913, 4975, 2438, 2961, 96, 3956, 1648, 4671, 3511,
2332, 4616, 3807, 1099, 2689, 951, 1859, 3672, 4327,
4946, 755, 3445, 807, 2050, 4470, 819, 3494, 4764,
1487, 2681, 1451, 1828, 600, 2998, 2378, 1446, 2079,
873, 270, 4942, 3757, 4929, 2560, 3562, 3539, 1466,
871, 1762, 750, 1346, 533, 2678, 341, 1486, 2504,
3221, 679, 2068, 2145, 3144, 834, 1808, 3153, 3407,
2103, 1634, 1022, 1783, 3740, 3527, 3470, 3178, 4350,
3648, 2120, 4578, 1596, 135, 2530, 1745, 608, 4825,
4913, 4142, 3012, 1856, 2018, 3602, 264, 275, 4814,
1938, 4047, 1223, 3103, 4868, 3533, 4726, 3018, 1931,
379, 2338, 475, 3665, 4431, 938, 707]),
values=tensor([0.8419, 0.4424, 0.5698, 0.2999, 0.9295, 0.4679, 0.3442,
0.3474, 0.7467, 0.0757, 0.0276, 0.8208, 0.7200, 0.1976,
0.3319, 0.9583, 0.8463, 0.9566, 0.3073, 0.6760, 0.4346,
0.2886, 0.9486, 0.0795, 0.8036, 0.5111, 0.4404, 0.5873,
0.2286, 0.4238, 0.6160, 0.9372, 0.8314, 0.1765, 0.9714,
0.5934, 0.0764, 0.5254, 0.7722, 0.8765, 0.7821, 0.7165,
0.7425, 0.1690, 0.9418, 0.7089, 0.3090, 0.3146, 0.3776,
0.3970, 0.7107, 0.4232, 0.2742, 0.1785, 0.3661, 0.7381,
0.7677, 0.2922, 0.0118, 0.5142, 0.3228, 0.6287, 0.6950,
0.5212, 0.9233, 0.5583, 0.3402, 0.9655, 0.1707, 0.5180,
0.7601, 0.0519, 0.3853, 0.1663, 0.4842, 0.9445, 0.1159,
0.1236, 0.2320, 0.4008, 0.3127, 0.0194, 0.2149, 0.2742,
0.3828, 0.5264, 0.2515, 0.8214, 0.1769, 0.1933, 0.8188,
0.5274, 0.2875, 0.2494, 0.8088, 0.9923, 0.5445, 0.3175,
0.6285, 0.6236, 0.2042, 0.2625, 0.5051, 0.4802, 0.6055,
0.2595, 0.3970, 0.4291, 0.2183, 0.7748, 0.7343, 0.0474,
0.5801, 0.6534, 0.2948, 0.0363, 0.3237, 0.2880, 0.2211,
0.1790, 0.3192, 0.9079, 0.1088, 0.8037, 0.5242, 0.3090,
0.6078, 0.5167, 0.1361, 0.1093, 0.6079, 0.0095, 0.5118,
0.3018, 0.1316, 0.6571, 0.0073, 0.3654, 0.4280, 0.8191,
0.3184, 0.2360, 0.6869, 0.0155, 0.5085, 0.4025, 0.0799,
0.7194, 0.4048, 0.5539, 0.2632, 0.0734, 0.9784, 0.3601,
0.9418, 0.0499, 0.8840, 0.6116, 0.9865, 0.6081, 0.4861,
0.7266, 0.7795, 0.1224, 0.6387, 0.9470, 0.4315, 0.0825,
0.8006, 0.5528, 0.3202, 0.1662, 0.3257, 0.8268, 0.0860,
0.4786, 0.2279, 0.0058, 0.4003, 0.0577, 0.1538, 0.9729,
0.3529, 0.3205, 0.9176, 0.5843, 0.6548, 0.1570, 0.8380,
0.7278, 0.2116, 0.1503, 0.0103, 0.8089, 0.9813, 0.7760,
0.2123, 0.9690, 0.9240, 0.5892, 0.4778, 0.5100, 0.0404,
0.6261, 0.6426, 0.9521, 0.0053, 0.5755, 0.0743, 0.7500,
0.3281, 0.4225, 0.6900, 0.0916, 0.8990, 0.2711, 0.5755,
0.5712, 0.7556, 0.0051, 0.6971, 0.0437, 0.5565, 0.4256,
0.2960, 0.6043, 0.6836, 0.9303, 0.4472, 0.6016, 0.6132,
0.9503, 0.8339, 0.2697, 0.0658, 0.7983, 0.4874, 0.1771,
0.9875, 0.2001, 0.2752, 0.5608, 0.4997, 0.6797, 0.1612,
0.9007, 0.9904, 0.7264, 0.3981, 0.6661]),
col_indices=tensor([3901, 3663, 229, 3286, 2751, 3483, 2128, 1446, 4503,
3960, 1594, 4088, 4861, 755, 2490, 2051, 2278, 4729,
1103, 1042, 1554, 2825, 4031, 3744, 2602, 531, 4883,
202, 3192, 4676, 1858, 1406, 3540, 1868, 4258, 1296,
40, 2235, 3617, 4799, 680, 649, 3963, 1933, 1725,
575, 4509, 1266, 2706, 2278, 2363, 503, 1376, 3736,
4576, 1431, 3018, 4449, 379, 171, 699, 4269, 423,
15, 756, 1388, 2467, 4283, 3498, 330, 1828, 2936,
727, 2300, 2833, 2863, 109, 1560, 2849, 1474, 443,
230, 4619, 4963, 1262, 1748, 3663, 2453, 3127, 315,
2799, 4404, 3122, 4451, 3711, 4611, 77, 2376, 1019,
1276, 1245, 139, 3620, 243, 636, 3728, 523, 3275,
1319, 407, 4648, 783, 1431, 764, 2964, 886, 1325,
4241, 4394, 4361, 2591, 57, 2666, 3482, 767, 690,
1227, 174, 1857, 1992, 525, 1315, 1715, 350, 2942,
4827, 2788, 3414, 2521, 2538, 2474, 540, 1078, 638,
3541, 2163, 4818, 623, 1593, 2829, 4885, 1974, 2036,
2239, 4847, 1633, 3181, 2785, 2293, 3813, 4130, 3145,
586, 2131, 1377, 3635, 1416, 96, 1290, 4122, 4111,
3371, 3627, 1730, 4198, 1682, 3780, 3696, 967, 3018,
2372, 3866, 4983, 4085, 4264, 3039, 2899, 4144, 4646,
2984, 172, 4147, 1120, 2707, 124, 2849, 2323, 1520,
3250, 2548, 2660, 4708, 2235, 593, 199, 448, 1975,
4340, 1665, 3502, 4596, 1936, 3151, 759, 4834, 1142,
2548, 2586, 1859, 3767, 1222, 4845, 1226, 3446, 4199,
4381, 2057, 3436, 396, 3566, 1282, 173, 1462, 3915,
4165, 1821, 3059, 730, 1341, 2913, 4066, 2638, 3070,
430, 826, 3316, 2761, 4646, 4961, 2657]),
values=tensor([0.2777, 0.7994, 0.6055, 0.8858, 0.5401, 0.6289, 0.2315,
0.7108, 0.2919, 0.9440, 0.4352, 0.9935, 0.7547, 0.9487,
0.1198, 0.7138, 0.2381, 0.9158, 0.3679, 0.3919, 0.7781,
0.8318, 0.6054, 0.4010, 0.3469, 0.3425, 0.7408, 0.3350,
0.1090, 0.6705, 0.8875, 0.7480, 0.9795, 0.0427, 0.1593,
0.2017, 0.8682, 0.9704, 0.5078, 0.7021, 0.2351, 0.1990,
0.1799, 0.1614, 0.9052, 0.7818, 0.9469, 0.9144, 0.6416,
0.0978, 0.2154, 0.3794, 0.0722, 0.4288, 0.4423, 0.0392,
0.2361, 0.1773, 0.5238, 0.6266, 0.9495, 0.4070, 0.9779,
0.1080, 0.9756, 0.0212, 0.7575, 0.6901, 0.5589, 0.8829,
0.5870, 0.3829, 0.0909, 0.7414, 0.5878, 0.3480, 0.1168,
0.3972, 0.2804, 0.8860, 0.5903, 0.9778, 0.2522, 0.2229,
0.0973, 0.3159, 0.6835, 0.0134, 0.3067, 0.7266, 0.6764,
0.3082, 0.0327, 0.3921, 0.8622, 0.8074, 0.6252, 0.9606,
0.3313, 0.3455, 0.4533, 0.6697, 0.2711, 0.3754, 0.8727,
0.6651, 0.0380, 0.1210, 0.0259, 0.0087, 0.3017, 0.7186,
0.9688, 0.5810, 0.6939, 0.8057, 0.2727, 0.5144, 0.0126,
0.0636, 0.8543, 0.9756, 0.4583, 0.3014, 0.5014, 0.4285,
0.3361, 0.3583, 0.8660, 0.8752, 0.5050, 0.1837, 0.7102,
0.1957, 0.9064, 0.7982, 0.5015, 0.4099, 0.5809, 0.8801,
0.0073, 0.5658, 0.8433, 0.7251, 0.8971, 0.9752, 0.6676,
0.2814, 0.9394, 0.9811, 0.1778, 0.5627, 0.3569, 0.2951,
0.4362, 0.7414, 0.7224, 0.6917, 0.2922, 0.7465, 0.6523,
0.5621, 0.0779, 0.8744, 0.6553, 0.5271, 0.0990, 0.8629,
0.6483, 0.0044, 0.2027, 0.6359, 0.0842, 0.9816, 0.4377,
0.7291, 0.7757, 0.4150, 0.9512, 0.9053, 0.6628, 0.9162,
0.6353, 0.3725, 0.8919, 0.1505, 0.1975, 0.7728, 0.1846,
0.5340, 0.4217, 0.7643, 0.3438, 0.6005, 0.7795, 0.2067,
0.6674, 0.9142, 0.4620, 0.8140, 0.1036, 0.3590, 0.3372,
0.0756, 0.4219, 0.7019, 0.2017, 0.1876, 0.8857, 0.9443,
0.7034, 0.3858, 0.6463, 0.0872, 0.7101, 0.2546, 0.8101,
0.3637, 0.4495, 0.8137, 0.4469, 0.4204, 0.1055, 0.8379,
0.1725, 0.3312, 0.1791, 0.6141, 0.0562, 0.4774, 0.5212,
0.7724, 0.9039, 0.5626, 0.1051, 0.2569, 0.5243, 0.3982,
0.0444, 0.0991, 0.8125, 0.2081, 0.2559, 0.6572, 0.3238,
0.3534, 0.8270, 0.9704, 0.5262, 0.1397]),
size=(5000, 5000), nnz=250, layout=torch.sparse_csr)
tensor([0.5019, 0.1367, 0.6742, ..., 0.0249, 0.2703, 0.5698])
tensor([0.2610, 0.0051, 0.8611, ..., 0.6706, 0.7457, 0.2823])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -266,80 +158,80 @@ Rows: 5000
Size: 25000000
NNZ: 250
Density: 1e-05
Time: 9.940149784088135 seconds
Time: 3.01309871673584 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 293134 -ss 5000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.89356017112732}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 284909 -ss 5000 -sd 1e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 250, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.16942024230957}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 250, 250, 250]),
col_indices=tensor([3878, 1793, 196, 4523, 2590, 2367, 223, 4753, 811,
1344, 3831, 3891, 1471, 342, 75, 2706, 3424, 2402,
4777, 1498, 772, 4383, 2094, 2326, 2946, 1126, 2886,
1923, 932, 4969, 3541, 896, 863, 2668, 4869, 4410,
1937, 3764, 94, 3879, 1282, 4972, 2002, 1481, 4954,
2173, 3425, 2770, 4498, 897, 827, 829, 4189, 4991,
1934, 1179, 1128, 2114, 2395, 2017, 3561, 4691, 2694,
1416, 4554, 620, 4382, 4973, 3582, 230, 4891, 3107,
1979, 140, 933, 4466, 308, 3556, 3173, 4882, 4797,
1238, 3925, 3354, 291, 1290, 414, 2445, 3042, 2325,
2660, 1731, 777, 3782, 2858, 3541, 2112, 2412, 4004,
942, 1750, 1450, 45, 960, 1184, 3190, 4107, 4960,
4877, 4158, 1493, 3953, 2210, 1987, 3321, 3595, 1029,
1281, 3168, 3718, 1747, 2899, 1694, 1765, 2658, 2297,
2811, 4794, 2166, 3385, 3144, 590, 756, 1857, 2472,
2864, 4092, 2483, 1698, 3039, 4797, 2893, 4815, 1160,
92, 3486, 728, 546, 1233, 1206, 3098, 284, 3217,
1908, 3538, 556, 1018, 3457, 4789, 3509, 986, 1353,
3714, 674, 3739, 3917, 378, 4295, 1240, 678, 634,
156, 182, 2959, 787, 2431, 894, 4155, 1278, 4710,
115, 3800, 3528, 4651, 4055, 3457, 4797, 3790, 2898,
2898, 1677, 2106, 3532, 1869, 3926, 1788, 4954, 1802,
3662, 3116, 1672, 1899, 2743, 4402, 201, 2790, 4915,
3309, 2448, 4340, 71, 1083, 3547, 1833, 4517, 4811,
4522, 4837, 4905, 1773, 2748, 2712, 2488, 4842, 2297,
377, 2695, 2905, 534, 1022, 2504, 1436, 3486, 3980,
4913, 361, 4684, 2741, 722, 1718, 3274, 513, 1785,
4555, 575, 662, 3842, 1584, 2198, 215]),
values=tensor([0.8076, 0.3370, 0.6780, 0.9299, 0.5410, 0.0897, 0.3343,
0.8017, 0.5673, 0.5602, 0.9800, 0.8553, 0.3447, 0.6119,
0.3490, 0.5758, 0.1388, 0.7568, 0.7228, 0.2456, 0.6799,
0.7488, 0.1368, 0.7230, 0.3714, 0.8061, 0.2178, 0.6691,
0.7090, 0.6240, 0.5615, 0.6385, 0.8034, 0.6963, 0.9896,
0.1078, 0.2316, 0.3754, 0.7350, 0.4907, 0.3665, 0.2209,
0.4611, 0.7569, 0.4815, 0.7270, 0.4688, 0.5127, 0.0439,
0.4951, 0.3454, 0.1899, 0.8750, 0.1915, 0.8080, 0.6042,
0.7305, 0.2510, 0.4960, 0.3143, 0.3207, 0.3323, 0.5478,
0.3218, 0.1649, 0.9155, 0.2697, 0.4415, 0.6177, 0.1457,
0.9256, 0.6524, 0.8106, 0.1943, 0.2636, 0.7375, 0.5837,
0.4529, 0.4107, 0.4337, 0.4074, 0.1673, 0.9988, 0.7338,
0.1243, 0.9778, 0.4221, 0.2348, 0.5442, 0.1259, 0.4222,
0.6127, 0.0857, 0.6974, 0.0596, 0.3553, 0.4614, 0.5799,
0.4404, 0.3360, 0.3314, 0.9445, 0.7231, 0.9851, 0.8853,
0.4987, 0.3871, 0.5069, 0.6349, 0.9384, 0.3450, 0.4613,
0.2127, 0.4994, 0.0034, 0.9538, 0.3203, 0.8248, 0.5140,
0.0568, 0.3913, 0.0456, 0.0790, 0.1457, 0.8710, 0.7025,
0.5191, 0.7160, 0.2410, 0.7547, 0.7169, 0.9282, 0.0473,
0.4454, 0.5093, 0.4795, 0.3417, 0.5014, 0.0605, 0.9341,
0.8068, 0.8325, 0.0916, 0.8219, 0.9882, 0.3617, 0.8114,
0.3412, 0.2133, 0.4138, 0.2870, 0.1987, 0.5576, 0.8136,
0.4067, 0.6195, 0.5018, 0.5513, 0.5252, 0.0402, 0.7889,
0.3122, 0.6215, 0.9385, 0.7669, 0.1080, 0.2818, 0.0494,
0.0251, 0.3317, 0.4666, 0.5981, 0.5539, 0.1688, 0.7416,
0.8841, 0.4123, 0.1102, 0.1371, 0.7232, 0.6598, 0.7427,
0.8150, 0.1180, 0.3866, 0.1447, 0.4442, 0.5099, 0.1417,
0.2917, 0.8599, 0.3553, 0.2307, 0.1388, 0.1482, 0.8529,
0.3988, 0.9926, 0.3184, 0.2404, 0.4847, 0.5288, 0.0738,
0.0517, 0.1797, 0.1796, 0.7215, 0.5955, 0.6432, 0.0017,
0.6486, 0.6664, 0.4487, 0.7630, 0.7774, 0.9276, 0.9518,
0.4507, 0.3399, 0.7495, 0.4581, 0.6140, 0.0659, 0.8137,
0.4343, 0.4836, 0.5681, 0.7947, 0.1417, 0.8229, 0.0824,
0.2070, 0.8783, 0.3511, 0.9580, 0.1053, 0.3375, 0.2396,
0.0513, 0.5334, 0.3977, 0.6765, 0.1035, 0.8974, 0.8093,
0.7238, 0.8002, 0.6243, 0.9654, 0.2803]),
col_indices=tensor([1669, 2388, 3410, 214, 4888, 2047, 1859, 3824, 1130,
3331, 4650, 808, 1845, 4600, 2980, 4756, 2639, 4242,
120, 4542, 2175, 1322, 104, 704, 854, 2110, 1063,
1256, 2794, 2665, 1239, 4623, 2397, 2905, 1669, 3634,
691, 1001, 4550, 1274, 2606, 2628, 4848, 3423, 4205,
4849, 3844, 4805, 2751, 822, 856, 3866, 2362, 4396,
3513, 3731, 4108, 1129, 2401, 2429, 238, 3568, 2538,
4839, 3438, 2131, 3982, 1035, 620, 3061, 2659, 870,
31, 582, 3725, 2164, 3897, 4881, 3537, 2824, 936,
4420, 341, 4499, 2690, 351, 3823, 4169, 4790, 4554,
2495, 1376, 3626, 221, 4721, 2833, 4128, 83, 287,
4091, 4135, 4551, 3973, 764, 392, 4740, 2858, 4378,
2517, 4820, 3243, 3784, 1749, 2694, 3058, 661, 4273,
2427, 4542, 135, 3704, 3578, 4193, 3743, 3465, 2179,
4188, 2714, 3316, 1323, 3063, 3972, 3355, 1842, 1656,
2481, 1669, 1106, 4204, 1040, 565, 3967, 2999, 776,
1132, 4335, 252, 3480, 3592, 4417, 2743, 508, 1998,
2250, 4747, 3247, 3778, 2520, 4340, 4333, 889, 3347,
1306, 252, 3840, 4251, 3753, 922, 1530, 732, 4724,
4652, 2305, 676, 3763, 2577, 479, 3149, 3237, 682,
2204, 1170, 4037, 1115, 902, 2463, 2133, 49, 3338,
846, 2596, 1254, 611, 336, 2556, 4596, 3162, 2347,
1052, 1946, 3013, 1910, 3262, 793, 681, 3061, 4097,
649, 4096, 3982, 4856, 2244, 770, 1157, 3683, 1150,
4034, 4307, 4867, 947, 1680, 3888, 190, 677, 2841,
816, 454, 4546, 1683, 1115, 4528, 4055, 324, 2442,
1530, 1512, 2880, 1124, 741, 2337, 2820, 1096, 969,
4662, 1861, 4067, 2109, 3996, 1635, 499]),
values=tensor([0.8132, 0.1702, 0.5583, 0.1261, 0.6291, 0.5508, 0.1330,
0.9627, 0.2059, 0.3644, 0.3622, 0.4731, 0.3091, 0.8919,
0.7060, 0.5289, 0.7945, 0.3422, 0.4040, 0.9747, 0.1778,
0.1060, 0.3373, 0.1041, 0.0936, 0.4036, 0.4021, 0.5444,
0.4938, 0.5992, 0.1894, 0.3036, 0.6677, 0.4744, 0.8443,
0.2067, 0.1390, 0.7860, 0.2069, 0.5019, 0.5539, 0.4807,
0.6194, 0.5176, 0.2767, 0.7631, 0.4453, 0.0999, 0.7181,
0.2470, 0.2255, 0.5250, 0.2866, 0.8997, 0.0544, 0.1824,
0.2628, 0.9339, 0.2590, 0.5943, 0.2439, 0.4256, 0.8224,
0.2204, 0.5000, 0.2703, 0.2122, 0.2501, 0.5794, 0.6155,
0.5183, 0.5021, 0.6112, 0.1537, 0.1024, 0.3154, 0.0744,
0.5354, 0.3979, 0.6342, 0.7319, 0.0847, 0.3194, 0.5800,
0.2467, 0.5775, 0.6339, 0.2050, 0.2286, 0.7874, 0.1733,
0.7255, 0.0573, 0.6716, 0.4231, 0.6554, 0.3477, 0.4703,
0.1981, 0.3923, 0.3520, 0.4289, 0.4033, 0.1353, 0.5197,
0.9189, 0.6985, 0.9291, 0.8051, 0.5530, 0.0423, 0.9594,
0.8487, 0.2554, 0.0395, 0.4103, 0.1345, 0.0607, 0.2812,
0.7571, 0.9906, 0.2249, 0.3326, 0.1389, 0.8069, 0.2156,
0.3462, 0.2324, 0.0457, 0.8244, 0.5205, 0.0833, 0.1781,
0.3837, 0.9227, 0.2976, 0.9031, 0.2499, 0.3484, 0.3298,
0.6568, 0.3816, 0.5687, 0.3523, 0.3593, 0.7242, 0.1034,
0.3478, 0.4454, 0.7734, 0.2847, 0.4512, 0.5866, 0.1633,
0.7139, 0.4511, 0.5642, 0.2230, 0.1384, 0.2467, 0.5114,
0.5149, 0.4901, 0.7340, 0.5840, 0.0495, 0.1493, 0.4501,
0.5299, 0.1752, 0.0737, 0.0887, 0.7004, 0.7171, 0.6451,
0.1099, 0.6191, 0.3209, 0.2667, 0.2735, 0.3592, 0.7035,
0.1766, 0.2292, 0.6138, 0.2492, 0.8422, 0.5205, 0.0949,
0.6311, 0.1200, 0.6842, 0.3167, 0.3418, 0.7978, 0.1885,
0.9433, 0.6390, 0.5217, 0.8313, 0.4066, 0.8623, 0.9330,
0.7999, 0.0688, 0.3315, 0.2496, 0.2006, 0.0199, 0.1239,
0.0030, 0.9251, 0.8374, 0.2492, 0.6001, 0.0171, 0.3645,
0.9564, 0.7314, 0.8427, 0.8917, 0.1465, 0.2355, 0.6975,
0.9025, 0.0358, 0.2860, 0.4051, 0.9734, 0.8626, 0.4028,
0.9642, 0.0743, 0.8714, 0.6919, 0.3640, 0.9239, 0.1573,
0.9549, 0.3068, 0.2789, 0.0169, 0.6253, 0.7318, 0.1857,
0.1394, 0.2220, 0.2355, 0.9726, 0.9750]),
size=(5000, 5000), nnz=250, layout=torch.sparse_csr)
tensor([0.4617, 0.6014, 0.4133, ..., 0.3579, 0.3877, 0.5185])
tensor([0.9060, 0.0911, 0.6185, ..., 0.7353, 0.0547, 0.2301])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -347,77 +239,77 @@ Rows: 5000
Size: 25000000
NNZ: 250
Density: 1e-05
Time: 10.89356017112732 seconds
Time: 10.16942024230957 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 250, 250, 250]),
col_indices=tensor([3878, 1793, 196, 4523, 2590, 2367, 223, 4753, 811,
1344, 3831, 3891, 1471, 342, 75, 2706, 3424, 2402,
4777, 1498, 772, 4383, 2094, 2326, 2946, 1126, 2886,
1923, 932, 4969, 3541, 896, 863, 2668, 4869, 4410,
1937, 3764, 94, 3879, 1282, 4972, 2002, 1481, 4954,
2173, 3425, 2770, 4498, 897, 827, 829, 4189, 4991,
1934, 1179, 1128, 2114, 2395, 2017, 3561, 4691, 2694,
1416, 4554, 620, 4382, 4973, 3582, 230, 4891, 3107,
1979, 140, 933, 4466, 308, 3556, 3173, 4882, 4797,
1238, 3925, 3354, 291, 1290, 414, 2445, 3042, 2325,
2660, 1731, 777, 3782, 2858, 3541, 2112, 2412, 4004,
942, 1750, 1450, 45, 960, 1184, 3190, 4107, 4960,
4877, 4158, 1493, 3953, 2210, 1987, 3321, 3595, 1029,
1281, 3168, 3718, 1747, 2899, 1694, 1765, 2658, 2297,
2811, 4794, 2166, 3385, 3144, 590, 756, 1857, 2472,
2864, 4092, 2483, 1698, 3039, 4797, 2893, 4815, 1160,
92, 3486, 728, 546, 1233, 1206, 3098, 284, 3217,
1908, 3538, 556, 1018, 3457, 4789, 3509, 986, 1353,
3714, 674, 3739, 3917, 378, 4295, 1240, 678, 634,
156, 182, 2959, 787, 2431, 894, 4155, 1278, 4710,
115, 3800, 3528, 4651, 4055, 3457, 4797, 3790, 2898,
2898, 1677, 2106, 3532, 1869, 3926, 1788, 4954, 1802,
3662, 3116, 1672, 1899, 2743, 4402, 201, 2790, 4915,
3309, 2448, 4340, 71, 1083, 3547, 1833, 4517, 4811,
4522, 4837, 4905, 1773, 2748, 2712, 2488, 4842, 2297,
377, 2695, 2905, 534, 1022, 2504, 1436, 3486, 3980,
4913, 361, 4684, 2741, 722, 1718, 3274, 513, 1785,
4555, 575, 662, 3842, 1584, 2198, 215]),
values=tensor([0.8076, 0.3370, 0.6780, 0.9299, 0.5410, 0.0897, 0.3343,
0.8017, 0.5673, 0.5602, 0.9800, 0.8553, 0.3447, 0.6119,
0.3490, 0.5758, 0.1388, 0.7568, 0.7228, 0.2456, 0.6799,
0.7488, 0.1368, 0.7230, 0.3714, 0.8061, 0.2178, 0.6691,
0.7090, 0.6240, 0.5615, 0.6385, 0.8034, 0.6963, 0.9896,
0.1078, 0.2316, 0.3754, 0.7350, 0.4907, 0.3665, 0.2209,
0.4611, 0.7569, 0.4815, 0.7270, 0.4688, 0.5127, 0.0439,
0.4951, 0.3454, 0.1899, 0.8750, 0.1915, 0.8080, 0.6042,
0.7305, 0.2510, 0.4960, 0.3143, 0.3207, 0.3323, 0.5478,
0.3218, 0.1649, 0.9155, 0.2697, 0.4415, 0.6177, 0.1457,
0.9256, 0.6524, 0.8106, 0.1943, 0.2636, 0.7375, 0.5837,
0.4529, 0.4107, 0.4337, 0.4074, 0.1673, 0.9988, 0.7338,
0.1243, 0.9778, 0.4221, 0.2348, 0.5442, 0.1259, 0.4222,
0.6127, 0.0857, 0.6974, 0.0596, 0.3553, 0.4614, 0.5799,
0.4404, 0.3360, 0.3314, 0.9445, 0.7231, 0.9851, 0.8853,
0.4987, 0.3871, 0.5069, 0.6349, 0.9384, 0.3450, 0.4613,
0.2127, 0.4994, 0.0034, 0.9538, 0.3203, 0.8248, 0.5140,
0.0568, 0.3913, 0.0456, 0.0790, 0.1457, 0.8710, 0.7025,
0.5191, 0.7160, 0.2410, 0.7547, 0.7169, 0.9282, 0.0473,
0.4454, 0.5093, 0.4795, 0.3417, 0.5014, 0.0605, 0.9341,
0.8068, 0.8325, 0.0916, 0.8219, 0.9882, 0.3617, 0.8114,
0.3412, 0.2133, 0.4138, 0.2870, 0.1987, 0.5576, 0.8136,
0.4067, 0.6195, 0.5018, 0.5513, 0.5252, 0.0402, 0.7889,
0.3122, 0.6215, 0.9385, 0.7669, 0.1080, 0.2818, 0.0494,
0.0251, 0.3317, 0.4666, 0.5981, 0.5539, 0.1688, 0.7416,
0.8841, 0.4123, 0.1102, 0.1371, 0.7232, 0.6598, 0.7427,
0.8150, 0.1180, 0.3866, 0.1447, 0.4442, 0.5099, 0.1417,
0.2917, 0.8599, 0.3553, 0.2307, 0.1388, 0.1482, 0.8529,
0.3988, 0.9926, 0.3184, 0.2404, 0.4847, 0.5288, 0.0738,
0.0517, 0.1797, 0.1796, 0.7215, 0.5955, 0.6432, 0.0017,
0.6486, 0.6664, 0.4487, 0.7630, 0.7774, 0.9276, 0.9518,
0.4507, 0.3399, 0.7495, 0.4581, 0.6140, 0.0659, 0.8137,
0.4343, 0.4836, 0.5681, 0.7947, 0.1417, 0.8229, 0.0824,
0.2070, 0.8783, 0.3511, 0.9580, 0.1053, 0.3375, 0.2396,
0.0513, 0.5334, 0.3977, 0.6765, 0.1035, 0.8974, 0.8093,
0.7238, 0.8002, 0.6243, 0.9654, 0.2803]),
col_indices=tensor([1669, 2388, 3410, 214, 4888, 2047, 1859, 3824, 1130,
3331, 4650, 808, 1845, 4600, 2980, 4756, 2639, 4242,
120, 4542, 2175, 1322, 104, 704, 854, 2110, 1063,
1256, 2794, 2665, 1239, 4623, 2397, 2905, 1669, 3634,
691, 1001, 4550, 1274, 2606, 2628, 4848, 3423, 4205,
4849, 3844, 4805, 2751, 822, 856, 3866, 2362, 4396,
3513, 3731, 4108, 1129, 2401, 2429, 238, 3568, 2538,
4839, 3438, 2131, 3982, 1035, 620, 3061, 2659, 870,
31, 582, 3725, 2164, 3897, 4881, 3537, 2824, 936,
4420, 341, 4499, 2690, 351, 3823, 4169, 4790, 4554,
2495, 1376, 3626, 221, 4721, 2833, 4128, 83, 287,
4091, 4135, 4551, 3973, 764, 392, 4740, 2858, 4378,
2517, 4820, 3243, 3784, 1749, 2694, 3058, 661, 4273,
2427, 4542, 135, 3704, 3578, 4193, 3743, 3465, 2179,
4188, 2714, 3316, 1323, 3063, 3972, 3355, 1842, 1656,
2481, 1669, 1106, 4204, 1040, 565, 3967, 2999, 776,
1132, 4335, 252, 3480, 3592, 4417, 2743, 508, 1998,
2250, 4747, 3247, 3778, 2520, 4340, 4333, 889, 3347,
1306, 252, 3840, 4251, 3753, 922, 1530, 732, 4724,
4652, 2305, 676, 3763, 2577, 479, 3149, 3237, 682,
2204, 1170, 4037, 1115, 902, 2463, 2133, 49, 3338,
846, 2596, 1254, 611, 336, 2556, 4596, 3162, 2347,
1052, 1946, 3013, 1910, 3262, 793, 681, 3061, 4097,
649, 4096, 3982, 4856, 2244, 770, 1157, 3683, 1150,
4034, 4307, 4867, 947, 1680, 3888, 190, 677, 2841,
816, 454, 4546, 1683, 1115, 4528, 4055, 324, 2442,
1530, 1512, 2880, 1124, 741, 2337, 2820, 1096, 969,
4662, 1861, 4067, 2109, 3996, 1635, 499]),
values=tensor([0.8132, 0.1702, 0.5583, 0.1261, 0.6291, 0.5508, 0.1330,
0.9627, 0.2059, 0.3644, 0.3622, 0.4731, 0.3091, 0.8919,
0.7060, 0.5289, 0.7945, 0.3422, 0.4040, 0.9747, 0.1778,
0.1060, 0.3373, 0.1041, 0.0936, 0.4036, 0.4021, 0.5444,
0.4938, 0.5992, 0.1894, 0.3036, 0.6677, 0.4744, 0.8443,
0.2067, 0.1390, 0.7860, 0.2069, 0.5019, 0.5539, 0.4807,
0.6194, 0.5176, 0.2767, 0.7631, 0.4453, 0.0999, 0.7181,
0.2470, 0.2255, 0.5250, 0.2866, 0.8997, 0.0544, 0.1824,
0.2628, 0.9339, 0.2590, 0.5943, 0.2439, 0.4256, 0.8224,
0.2204, 0.5000, 0.2703, 0.2122, 0.2501, 0.5794, 0.6155,
0.5183, 0.5021, 0.6112, 0.1537, 0.1024, 0.3154, 0.0744,
0.5354, 0.3979, 0.6342, 0.7319, 0.0847, 0.3194, 0.5800,
0.2467, 0.5775, 0.6339, 0.2050, 0.2286, 0.7874, 0.1733,
0.7255, 0.0573, 0.6716, 0.4231, 0.6554, 0.3477, 0.4703,
0.1981, 0.3923, 0.3520, 0.4289, 0.4033, 0.1353, 0.5197,
0.9189, 0.6985, 0.9291, 0.8051, 0.5530, 0.0423, 0.9594,
0.8487, 0.2554, 0.0395, 0.4103, 0.1345, 0.0607, 0.2812,
0.7571, 0.9906, 0.2249, 0.3326, 0.1389, 0.8069, 0.2156,
0.3462, 0.2324, 0.0457, 0.8244, 0.5205, 0.0833, 0.1781,
0.3837, 0.9227, 0.2976, 0.9031, 0.2499, 0.3484, 0.3298,
0.6568, 0.3816, 0.5687, 0.3523, 0.3593, 0.7242, 0.1034,
0.3478, 0.4454, 0.7734, 0.2847, 0.4512, 0.5866, 0.1633,
0.7139, 0.4511, 0.5642, 0.2230, 0.1384, 0.2467, 0.5114,
0.5149, 0.4901, 0.7340, 0.5840, 0.0495, 0.1493, 0.4501,
0.5299, 0.1752, 0.0737, 0.0887, 0.7004, 0.7171, 0.6451,
0.1099, 0.6191, 0.3209, 0.2667, 0.2735, 0.3592, 0.7035,
0.1766, 0.2292, 0.6138, 0.2492, 0.8422, 0.5205, 0.0949,
0.6311, 0.1200, 0.6842, 0.3167, 0.3418, 0.7978, 0.1885,
0.9433, 0.6390, 0.5217, 0.8313, 0.4066, 0.8623, 0.9330,
0.7999, 0.0688, 0.3315, 0.2496, 0.2006, 0.0199, 0.1239,
0.0030, 0.9251, 0.8374, 0.2492, 0.6001, 0.0171, 0.3645,
0.9564, 0.7314, 0.8427, 0.8917, 0.1465, 0.2355, 0.6975,
0.9025, 0.0358, 0.2860, 0.4051, 0.9734, 0.8626, 0.4028,
0.9642, 0.0743, 0.8714, 0.6919, 0.3640, 0.9239, 0.1573,
0.9549, 0.3068, 0.2789, 0.0169, 0.6253, 0.7318, 0.1857,
0.1394, 0.2220, 0.2355, 0.9726, 0.9750]),
size=(5000, 5000), nnz=250, layout=torch.sparse_csr)
tensor([0.4617, 0.6014, 0.4133, ..., 0.3579, 0.3877, 0.5185])
tensor([0.9060, 0.0911, 0.6185, ..., 0.7353, 0.0547, 0.2301])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -425,13 +317,13 @@ Rows: 5000
Size: 25000000
NNZ: 250
Density: 1e-05
Time: 10.89356017112732 seconds
Time: 10.16942024230957 seconds
[16.36, 16.36, 16.4, 16.6, 16.6, 16.96, 16.96, 16.88, 16.88, 16.48]
[16.32, 16.28, 19.04, 20.36, 23.52, 24.24, 24.96, 24.96, 22.16, 21.36, 19.68, 19.76, 19.8, 19.64]
14.221415042877197
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 293134, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.89356017112732, 'TIME_S_1KI': 0.03716239048055606, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 278.4485914421082, 'W': 19.57952781791354}
[16.36, 16.36, 16.4, 16.6, 16.6, 16.96, 16.96, 16.88, 16.88, 16.48, 16.6, 16.64, 16.68, 16.76, 16.68, 16.56, 16.4, 16.4, 16.52, 16.64]
299.32000000000005
14.966000000000003
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 293134, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.89356017112732, 'TIME_S_1KI': 0.03716239048055606, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 278.4485914421082, 'W': 19.57952781791354, 'J_1KI': 0.9499020633638819, 'W_1KI': 0.06679377969772711, 'W_D': 4.613527817913537, 'J_D': 65.610893910408, 'W_D_1KI': 0.01573863085794735, 'J_D_1KI': 5.3690908792386244e-05}
[20.28, 20.4, 20.4, 20.48, 20.76, 21.28, 21.16, 20.96, 20.76, 20.4]
[20.12, 20.44, 21.32, 22.44, 25.12, 25.12, 25.76, 26.28, 25.96, 25.44, 23.88, 24.04, 24.16, 23.84]
14.210330247879028
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 284909, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.16942024230957, 'TIME_S_1KI': 0.035693573184102885, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 316.9542731094361, 'W': 22.304497332617817}
[20.28, 20.4, 20.4, 20.48, 20.76, 21.28, 21.16, 20.96, 20.76, 20.4, 20.8, 20.88, 20.72, 20.76, 20.84, 20.84, 20.44, 20.52, 20.32, 20.6]
372.55999999999995
18.627999999999997
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 284909, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 250, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.16942024230957, 'TIME_S_1KI': 0.035693573184102885, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 316.9542731094361, 'W': 22.304497332617817, 'J_1KI': 1.1124754679895548, 'W_1KI': 0.07828639085679223, 'W_D': 3.6764973326178207, 'J_D': 52.24424125194558, 'W_D_1KI': 0.012904110900736098, 'J_D_1KI': 4.529204377796454e-05}

2
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 151147, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.619585275650024, "TIME_S_1KI": 0.07025998051995755, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 328.34527337074286, "W": 23.06683265939357, "J_1KI": 2.1723571977660345, "W_1KI": 0.15261191197571614, "W_D": 4.6928326593935665, "J_D": 66.80021679544454, "W_D_1KI": 0.031048136313612352, "J_D_1KI": 0.00020541682146263144}
{"CPU": "Altra", "CORES": 16, "ITERATIONS": 154432, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.621034145355225, "TIME_S_1KI": 0.06877482740206191, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 316.4244461250305, "W": 22.262246477486542, "J_1KI": 2.0489564735613763, "W_1KI": 0.144155657360434, "W_D": 3.7392464774865424, "J_D": 53.14778078484536, "W_D_1KI": 0.024212899382812774, "J_D_1KI": 0.00015678680184685024}

68
pytorch/output_synthetic_16core/altra_16_csr_10_10_10_synthetic_5000_5e-05.output
View File

@ -1,13 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 1000 -ss 5000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.07715368270874023}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 100 -ss 5000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.014879941940307617}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 1250, 1250, 1250]),
col_indices=tensor([4186, 604, 2911, ..., 3524, 2664, 807]),
values=tensor([0.1303, 0.5472, 0.9541, ..., 0.4453, 0.4813, 0.2933]),
tensor(crow_indices=tensor([ 0, 0, 0, ..., 1248, 1250, 1250]),
col_indices=tensor([1397, 3608, 621, ..., 1983, 2722, 4972]),
values=tensor([0.7898, 0.8890, 0.9853, ..., 0.2806, 0.4332, 0.7785]),
size=(5000, 5000), nnz=1250, layout=torch.sparse_csr)
tensor([0.2363, 0.5745, 0.8536, ..., 0.3028, 0.7626, 0.7945])
tensor([0.8515, 0.1205, 0.1290, ..., 0.0596, 0.1294, 0.2178])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -15,18 +15,18 @@ Rows: 5000
Size: 25000000
NNZ: 1250
Density: 5e-05
Time: 0.07715368270874023 seconds
Time: 0.014879941940307617 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 136092 -ss 5000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 9.454103946685791}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 70564 -ss 5000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 4.797720670700073}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 1250, 1250, 1250]),
col_indices=tensor([2642, 3295, 3317, ..., 552, 1688, 3754]),
values=tensor([0.5853, 0.8410, 0.7758, ..., 0.7543, 0.4171, 0.3907]),
tensor(crow_indices=tensor([ 0, 1, 2, ..., 1249, 1250, 1250]),
col_indices=tensor([4236, 1927, 389, ..., 3900, 4084, 4178]),
values=tensor([0.5819, 0.5926, 0.4032, ..., 0.1422, 0.8129, 0.9187]),
size=(5000, 5000), nnz=1250, layout=torch.sparse_csr)
tensor([0.4145, 0.1634, 0.4401, ..., 0.9903, 0.7928, 0.8495])
tensor([0.4782, 0.7587, 0.6755, ..., 0.4641, 0.3230, 0.1517])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -34,18 +34,18 @@ Rows: 5000
Size: 25000000
NNZ: 1250
Density: 5e-05
Time: 9.454103946685791 seconds
Time: 4.797720670700073 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 151147 -ss 5000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.619585275650024}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-altra.sif', '-c', 'numactl --cpunodebind=0 --membind=0 python3 spmv.py synthetic csr 154432 -ss 5000 -sd 5e-05 -c 16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [5000, 5000], "MATRIX_ROWS": 5000, "MATRIX_SIZE": 25000000, "MATRIX_NNZ": 1250, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.621034145355225}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 1250, 1250, 1250]),
col_indices=tensor([2120, 4070, 3230, ..., 2901, 3405, 168]),
values=tensor([0.1519, 0.7232, 0.6282, ..., 0.2528, 0.5199, 0.9452]),
tensor(crow_indices=tensor([ 0, 1, 1, ..., 1249, 1250, 1250]),
col_indices=tensor([ 91, 2944, 3974, ..., 4430, 70, 3263]),
values=tensor([0.2553, 0.0855, 0.4739, ..., 0.3797, 0.6721, 0.4378]),
size=(5000, 5000), nnz=1250, layout=torch.sparse_csr)
tensor([0.2553, 0.4766, 0.2217, ..., 0.4056, 0.3500, 0.9553])
tensor([0.8009, 0.9874, 0.1682, ..., 0.8612, 0.3697, 0.0752])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -53,15 +53,15 @@ Rows: 5000
Size: 25000000
NNZ: 1250
Density: 5e-05
Time: 10.619585275650024 seconds
Time: 10.621034145355225 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at /space/jenkins/workspace/Releases/pytorch-dls/pytorch-dls/aten/src/ATen/SparseCsrTensorImpl.cpp:55.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 1250, 1250, 1250]),
col_indices=tensor([2120, 4070, 3230, ..., 2901, 3405, 168]),
values=tensor([0.1519, 0.7232, 0.6282, ..., 0.2528, 0.5199, 0.9452]),
tensor(crow_indices=tensor([ 0, 1, 1, ..., 1249, 1250, 1250]),
col_indices=tensor([ 91, 2944, 3974, ..., 4430, 70, 3263]),
values=tensor([0.2553, 0.0855, 0.4739, ..., 0.3797, 0.6721, 0.4378]),
size=(5000, 5000), nnz=1250, layout=torch.sparse_csr)
tensor([0.2553, 0.4766, 0.2217, ..., 0.4056, 0.3500, 0.9553])
tensor([0.8009, 0.9874, 0.1682, ..., 0.8612, 0.3697, 0.0752])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([5000, 5000])
@ -69,13 +69,13 @@ Rows: 5000
Size: 25000000
NNZ: 1250
Density: 5e-05
Time: 10.619585275650024 seconds
Time: 10.621034145355225 seconds
[20.44, 20.36, 20.28, 20.16, 20.04, 20.16, 20.28, 20.32, 20.56, 20.56]
[20.72, 20.76, 20.8, 24.32, 26.28, 28.48, 29.36, 29.52, 26.16, 23.92, 23.8, 23.64, 23.64, 23.56]
14.234519243240356
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 151147, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.619585275650024, 'TIME_S_1KI': 0.07025998051995755, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 328.34527337074286, 'W': 23.06683265939357}
[20.44, 20.36, 20.28, 20.16, 20.04, 20.16, 20.28, 20.32, 20.56, 20.56, 20.24, 20.2, 20.36, 20.68, 20.96, 20.8, 20.68, 20.4, 20.44, 20.36]
367.48
18.374000000000002
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 151147, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.619585275650024, 'TIME_S_1KI': 0.07025998051995755, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 328.34527337074286, 'W': 23.06683265939357, 'J_1KI': 2.1723571977660345, 'W_1KI': 0.15261191197571614, 'W_D': 4.6928326593935665, 'J_D': 66.80021679544454, 'W_D_1KI': 0.031048136313612352, 'J_D_1KI': 0.00020541682146263144}
[20.4, 20.68, 20.44, 20.16, 20.16, 20.48, 20.44, 20.68, 20.64, 20.88]
[20.92, 20.96, 21.52, 22.88, 25.2, 25.52, 26.2, 26.2, 25.6, 24.92, 23.32, 23.36, 23.56, 23.44]
14.213500261306763
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 154432, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.621034145355225, 'TIME_S_1KI': 0.06877482740206191, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 316.4244461250305, 'W': 22.262246477486542}
[20.4, 20.68, 20.44, 20.16, 20.16, 20.48, 20.44, 20.68, 20.64, 20.88, 20.72, 20.64, 20.68, 20.52, 20.64, 20.8, 20.76, 20.76, 20.68, 20.6]
370.46
18.523
{'CPU': 'Altra', 'CORES': 16, 'ITERATIONS': 154432, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [5000, 5000], 'MATRIX_ROWS': 5000, 'MATRIX_SIZE': 25000000, 'MATRIX_NNZ': 1250, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.621034145355225, 'TIME_S_1KI': 0.06877482740206191, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 316.4244461250305, 'W': 22.262246477486542, 'J_1KI': 2.0489564735613763, 'W_1KI': 0.144155657360434, 'W_D': 3.7392464774865424, 'J_D': 53.14778078484536, 'W_D_1KI': 0.024212899382812774, 'J_D_1KI': 0.00015678680184685024}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 63031, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.264819622039795, "TIME_S_1KI": 0.16285351052719765, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1907.2002365589144, "W": 143.8, "J_1KI": 30.25813070646054, "W_1KI": 2.2814170804842067, "W_D": 106.65425000000002, "J_D": 1414.5411045202616, "W_D_1KI": 1.6920919864828419, "J_D_1KI": 0.026845393322061237}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 65446, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.748796463012695, "TIME_S_1KI": 0.16423916607604278, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1866.1015530323982, "W": 143.13, "J_1KI": 28.51360744785622, "W_1KI": 2.1869938575314, "W_D": 106.99974999999999, "J_D": 1395.0422668139338, "W_D_1KI": 1.634931852214039, "J_D_1KI": 0.024981386978792274}

120
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_0.0001.output
View File

@ -1,14 +1,54 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '100000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.20868682861328125}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '100000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.060246944427490234}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 11, 23, ..., 999975,
tensor(crow_indices=tensor([ 0, 16, 25, ..., 999980,
999989, 1000000]),
col_indices=tensor([ 4573, 4595, 4948, ..., 71788, 92544, 99741]),
values=tensor([0.3512, 0.1040, 0.2729, ..., 0.2513, 0.9554, 0.9408]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.1257, 0.5794, 0.5612, ..., 0.8235, 0.1474, 0.3975])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 0.060246944427490234 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '17428', '-ss', '100000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 2.7960927486419678}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 8, 17, ..., 999980,
999989, 1000000]),
col_indices=tensor([11836, 34889, 39226, ..., 79566, 86668, 94364]),
values=tensor([0.7886, 0.3777, 0.4340, ..., 0.5250, 0.8836, 0.4934]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.9435, 0.7532, 0.3829, ..., 0.0561, 0.6547, 0.0145])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 2.7960927486419678 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '65446', '-ss', '100000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.748796463012695}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 8, 17, ..., 999978,
999990, 1000000]),
col_indices=tensor([ 1102, 1885, 5689, ..., 70464, 82505, 82637]),
values=tensor([0.9145, 0.6563, 0.0210, ..., 0.3467, 0.9517, 0.4307]),
col_indices=tensor([ 6624, 6694, 37331, ..., 71444, 97628, 99166]),
values=tensor([0.8094, 0.0427, 0.0622, ..., 0.4502, 0.4633, 0.1157]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.3954, 0.8531, 0.4592, ..., 0.1653, 0.9288, 0.8508])
tensor([0.2357, 0.1643, 0.3206, ..., 0.7759, 0.8620, 0.1771])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,39 +56,16 @@ Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 0.20868682861328125 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '50314', '-ss', '100000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 8.38151502609253}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 9, 15, ..., 999976,
999987, 1000000]),
col_indices=tensor([ 9326, 16949, 19479, ..., 70135, 76689, 93251]),
values=tensor([0.2491, 0.4486, 0.5526, ..., 0.3620, 0.8491, 0.1510]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.1294, 0.2549, 0.0676, ..., 0.6377, 0.6452, 0.0657])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 8.38151502609253 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '63031', '-ss', '100000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.264819622039795}
Time: 10.748796463012695 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 8, 17, ..., 999978,
999986, 1000000]),
col_indices=tensor([ 1906, 11602, 20474, ..., 94634, 95193, 99629]),
values=tensor([0.7595, 0.5479, 0.3671, ..., 0.3196, 0.4186, 0.5082]),
999990, 1000000]),
col_indices=tensor([ 6624, 6694, 37331, ..., 71444, 97628, 99166]),
values=tensor([0.8094, 0.0427, 0.0622, ..., 0.4502, 0.4633, 0.1157]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.5397, 0.2720, 0.7091, ..., 0.7919, 0.2241, 0.5973])
tensor([0.2357, 0.1643, 0.3206, ..., 0.7759, 0.8620, 0.1771])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -56,30 +73,13 @@ Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 10.264819622039795 seconds
Time: 10.748796463012695 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 8, 17, ..., 999978,
999986, 1000000]),
col_indices=tensor([ 1906, 11602, 20474, ..., 94634, 95193, 99629]),
values=tensor([0.7595, 0.5479, 0.3671, ..., 0.3196, 0.4186, 0.5082]),
size=(100000, 100000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.5397, 0.2720, 0.7091, ..., 0.7919, 0.2241, 0.5973])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 1000000
Density: 0.0001
Time: 10.264819622039795 seconds
[40.75, 39.66, 39.67, 39.19, 39.32, 45.15, 39.35, 39.19, 39.95, 39.54]
[143.8]
13.262866735458374
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 63031, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.264819622039795, 'TIME_S_1KI': 0.16285351052719765, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1907.2002365589144, 'W': 143.8}
[40.75, 39.66, 39.67, 39.19, 39.32, 45.15, 39.35, 39.19, 39.95, 39.54, 40.42, 44.44, 57.71, 39.25, 40.02, 40.75, 39.74, 39.58, 39.68, 39.82]
742.915
37.14575
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 63031, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.264819622039795, 'TIME_S_1KI': 0.16285351052719765, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1907.2002365589144, 'W': 143.8, 'J_1KI': 30.25813070646054, 'W_1KI': 2.2814170804842067, 'W_D': 106.65425000000002, 'J_D': 1414.5411045202616, 'W_D_1KI': 1.6920919864828419, 'J_D_1KI': 0.026845393322061237}
[40.51, 40.31, 40.08, 39.71, 39.8, 39.59, 39.58, 41.48, 39.56, 40.06]
[143.13]
13.037808656692505
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 65446, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.748796463012695, 'TIME_S_1KI': 0.16423916607604278, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1866.1015530323982, 'W': 143.13}
[40.51, 40.31, 40.08, 39.71, 39.8, 39.59, 39.58, 41.48, 39.56, 40.06, 41.48, 39.67, 40.17, 39.98, 41.44, 39.98, 40.5, 40.23, 39.59, 39.82]
722.605
36.130250000000004
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 65446, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.748796463012695, 'TIME_S_1KI': 0.16423916607604278, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1866.1015530323982, 'W': 143.13, 'J_1KI': 28.51360744785622, 'W_1KI': 2.1869938575314, 'W_D': 106.99974999999999, 'J_D': 1395.0422668139338, 'W_D_1KI': 1.634931852214039, 'J_D_1KI': 0.024981386978792274}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_0.001.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 4290, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.725477457046509, "TIME_S_1KI": 2.500111295348837, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2014.465433692932, "W": 126.69, "J_1KI": 469.57236216618463, "W_1KI": 29.53146853146853, "W_D": 91.17699999999999, "J_D": 1449.7822625923156, "W_D_1KI": 21.25337995337995, "J_D_1KI": 4.954167821300688}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 4693, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 12.009309530258179, "TIME_S_1KI": 2.558983492490556, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2204.305100774765, "W": 128.95, "J_1KI": 469.7006394150362, "W_1KI": 27.477093543575535, "W_D": 92.7465, "J_D": 1585.4329820008277, "W_D_1KI": 19.76273172810569, "J_D_1KI": 4.211108401471487}

80
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_0.001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '100000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 2.4475483894348145}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '100000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.27472805976867676}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 108, 211, ..., 9999795,
9999912, 10000000]),
col_indices=tensor([ 147, 1138, 2699, ..., 95915, 96101, 99505]),
values=tensor([0.5370, 0.7637, 0.8320, ..., 0.1671, 0.6910, 0.1145]),
tensor(crow_indices=tensor([ 0, 103, 224, ..., 9999788,
9999890, 10000000]),
col_indices=tensor([ 311, 3365, 5161, ..., 98602, 99530, 99576]),
values=tensor([0.9917, 0.0583, 0.3712, ..., 0.9136, 0.4986, 0.7909]),
size=(100000, 100000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.0022, 0.6683, 0.3307, ..., 0.4747, 0.3475, 0.4636])
tensor([0.4323, 0.4083, 0.9080, ..., 0.7530, 0.1922, 0.7136])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,19 +16,19 @@ Rows: 100000
Size: 10000000000
NNZ: 10000000
Density: 0.001
Time: 2.4475483894348145 seconds
Time: 0.27472805976867676 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '4290', '-ss', '100000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.725477457046509}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '3821', '-ss', '100000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 8.548935651779175}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 113, 209, ..., 9999816,
9999914, 10000000]),
col_indices=tensor([ 524, 3053, 3097, ..., 98248, 99944, 99996]),
values=tensor([0.2951, 0.6504, 0.4617, ..., 0.6241, 0.9747, 0.0943]),
tensor(crow_indices=tensor([ 0, 86, 193, ..., 9999790,
9999889, 10000000]),
col_indices=tensor([ 598, 3163, 6325, ..., 93333, 94869, 95502]),
values=tensor([0.3479, 0.2007, 0.7107, ..., 0.5121, 0.1193, 0.0296]),
size=(100000, 100000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.1529, 0.0141, 0.4287, ..., 0.1937, 0.2308, 0.9820])
tensor([0.9967, 0.6546, 0.0107, ..., 0.1473, 0.4856, 0.1261])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -36,16 +36,19 @@ Rows: 100000
Size: 10000000000
NNZ: 10000000
Density: 0.001
Time: 10.725477457046509 seconds
Time: 8.548935651779175 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '4693', '-ss', '100000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.001, "TIME_S": 12.009309530258179}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 113, 209, ..., 9999816,
9999914, 10000000]),
col_indices=tensor([ 524, 3053, 3097, ..., 98248, 99944, 99996]),
values=tensor([0.2951, 0.6504, 0.4617, ..., 0.6241, 0.9747, 0.0943]),
tensor(crow_indices=tensor([ 0, 80, 177, ..., 9999782,
9999890, 10000000]),
col_indices=tensor([ 1894, 3295, 3747, ..., 98404, 98823, 99018]),
values=tensor([0.1540, 0.7163, 0.3077, ..., 0.3211, 0.5255, 0.5012]),
size=(100000, 100000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.1529, 0.0141, 0.4287, ..., 0.1937, 0.2308, 0.9820])
tensor([0.8104, 0.7178, 0.6885, ..., 0.8661, 0.7147, 0.1559])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -53,13 +56,30 @@ Rows: 100000
Size: 10000000000
NNZ: 10000000
Density: 0.001
Time: 10.725477457046509 seconds
Time: 12.009309530258179 seconds
[39.92, 39.19, 39.59, 39.28, 39.81, 39.63, 39.63, 39.45, 39.41, 39.18]
[126.69]
15.900745391845703
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4290, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.725477457046509, 'TIME_S_1KI': 2.500111295348837, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2014.465433692932, 'W': 126.69}
[39.92, 39.19, 39.59, 39.28, 39.81, 39.63, 39.63, 39.45, 39.41, 39.18, 40.85, 39.15, 39.3, 39.24, 39.74, 39.48, 39.55, 39.09, 39.1, 39.29]
710.26
35.513
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4290, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.725477457046509, 'TIME_S_1KI': 2.500111295348837, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2014.465433692932, 'W': 126.69, 'J_1KI': 469.57236216618463, 'W_1KI': 29.53146853146853, 'W_D': 91.17699999999999, 'J_D': 1449.7822625923156, 'W_D_1KI': 21.25337995337995, 'J_D_1KI': 4.954167821300688}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 80, 177, ..., 9999782,
9999890, 10000000]),
col_indices=tensor([ 1894, 3295, 3747, ..., 98404, 98823, 99018]),
values=tensor([0.1540, 0.7163, 0.3077, ..., 0.3211, 0.5255, 0.5012]),
size=(100000, 100000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.8104, 0.7178, 0.6885, ..., 0.8661, 0.7147, 0.1559])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
Rows: 100000
Size: 10000000000
NNZ: 10000000
Density: 0.001
Time: 12.009309530258179 seconds
[41.32, 39.94, 39.97, 39.81, 39.83, 40.24, 40.5, 40.22, 40.21, 41.33]
[128.95]
17.09426212310791
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4693, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 12.009309530258179, 'TIME_S_1KI': 2.558983492490556, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2204.305100774765, 'W': 128.95}
[41.32, 39.94, 39.97, 39.81, 39.83, 40.24, 40.5, 40.22, 40.21, 41.33, 40.63, 40.93, 40.28, 39.66, 39.87, 41.7, 39.67, 40.03, 39.68, 39.78]
724.0699999999999
36.2035
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4693, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 12.009309530258179, 'TIME_S_1KI': 2.558983492490556, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2204.305100774765, 'W': 128.95, 'J_1KI': 469.7006394150362, 'W_1KI': 27.477093543575535, 'W_D': 92.7465, 'J_D': 1585.4329820008277, 'W_D_1KI': 19.76273172810569, 'J_D_1KI': 4.211108401471487}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 102924, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.60866904258728, "TIME_S_1KI": 0.103072840567674, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1542.0244372987747, "W": 115.47, "J_1KI": 14.982165843717448, "W_1KI": 1.121895767750962, "W_D": 79.97325000000001, "J_D": 1067.989138565898, "W_D_1KI": 0.7770126501107615, "J_D_1KI": 0.00754938255519375}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 99857, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.432250738143921, "TIME_S_1KI": 0.10447190220158749, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1536.9044136524199, "W": 115.09999999999998, "J_1KI": 15.391053342804408, "W_1KI": 1.152648287050482, "W_D": 79.15799999999997, "J_D": 1056.9789711198803, "W_D_1KI": 0.7927135804200004, "J_D_1KI": 0.007938487841813797}

68
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_1e-05.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '100000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.12978029251098633}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '100000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.043670654296875}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 3, 4, ..., 99999, 100000,
tensor(crow_indices=tensor([ 0, 1, 1, ..., 99996, 99998,
100000]),
col_indices=tensor([21616, 77637, 85619, ..., 53732, 81470, 6094]),
values=tensor([0.4857, 0.1991, 0.9153, ..., 0.9203, 0.8308, 0.8562]),
col_indices=tensor([ 6609, 19255, 81333, ..., 81128, 51531, 76130]),
values=tensor([0.9876, 0.0139, 0.8085, ..., 0.3685, 0.4758, 0.0266]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.0197, 0.8164, 0.2872, ..., 0.9903, 0.3891, 0.9778])
tensor([0.1735, 0.8240, 0.8190, ..., 0.4288, 0.7745, 0.1715])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,19 +16,19 @@ Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 0.12978029251098633 seconds
Time: 0.043670654296875 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '80905', '-ss', '100000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 8.253613233566284}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '24043', '-ss', '100000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 2.5281074047088623}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2, 3, ..., 99999, 99999,
tensor(crow_indices=tensor([ 0, 0, 0, ..., 99997, 99999,
100000]),
col_indices=tensor([18950, 61338, 17160, ..., 57514, 79997, 96494]),
values=tensor([0.7220, 0.1840, 0.6067, ..., 0.9597, 0.4652, 0.5228]),
col_indices=tensor([69039, 75318, 84133, ..., 16483, 23976, 47642]),
values=tensor([0.3961, 0.2517, 0.3876, ..., 0.3761, 0.7912, 0.1675]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.0221, 0.6414, 0.1516, ..., 0.3018, 0.8902, 0.3461])
tensor([0.9918, 0.3750, 0.7737, ..., 0.5214, 0.0832, 0.2225])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -36,19 +36,19 @@ Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 8.253613233566284 seconds
Time: 2.5281074047088623 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '102924', '-ss', '100000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.60866904258728}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '99857', '-ss', '100000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.432250738143921}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 99999, 99999,
tensor(crow_indices=tensor([ 0, 1, 1, ..., 99998, 100000,
100000]),
col_indices=tensor([ 4611, 80501, 8771, ..., 95435, 27789, 45343]),
values=tensor([0.8274, 0.0201, 0.6109, ..., 0.4116, 0.6491, 0.0785]),
col_indices=tensor([18969, 38131, 43029, ..., 81495, 1519, 27704]),
values=tensor([0.3850, 0.3770, 0.8820, ..., 0.3865, 0.0804, 0.8829]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.0461, 0.3256, 0.3375, ..., 0.6234, 0.9526, 0.7301])
tensor([0.4374, 0.1348, 0.8967, ..., 0.5157, 0.0353, 0.0014])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -56,16 +56,16 @@ Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 10.60866904258728 seconds
Time: 10.432250738143921 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 99999, 99999,
tensor(crow_indices=tensor([ 0, 1, 1, ..., 99998, 100000,
100000]),
col_indices=tensor([ 4611, 80501, 8771, ..., 95435, 27789, 45343]),
values=tensor([0.8274, 0.0201, 0.6109, ..., 0.4116, 0.6491, 0.0785]),
col_indices=tensor([18969, 38131, 43029, ..., 81495, 1519, 27704]),
values=tensor([0.3850, 0.3770, 0.8820, ..., 0.3865, 0.0804, 0.8829]),
size=(100000, 100000), nnz=100000, layout=torch.sparse_csr)
tensor([0.0461, 0.3256, 0.3375, ..., 0.6234, 0.9526, 0.7301])
tensor([0.4374, 0.1348, 0.8967, ..., 0.5157, 0.0353, 0.0014])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -73,13 +73,13 @@ Rows: 100000
Size: 10000000000
NNZ: 100000
Density: 1e-05
Time: 10.60866904258728 seconds
Time: 10.432250738143921 seconds
[39.93, 39.22, 39.43, 40.14, 39.51, 39.36, 39.3, 39.3, 39.26, 39.1]
[115.47]
13.354329586029053
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 102924, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.60866904258728, 'TIME_S_1KI': 0.103072840567674, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1542.0244372987747, 'W': 115.47}
[39.93, 39.22, 39.43, 40.14, 39.51, 39.36, 39.3, 39.3, 39.26, 39.1, 41.76, 39.08, 39.78, 39.45, 39.66, 39.16, 39.27, 39.06, 39.08, 38.96]
709.935
35.49675
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 102924, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.60866904258728, 'TIME_S_1KI': 0.103072840567674, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1542.0244372987747, 'W': 115.47, 'J_1KI': 14.982165843717448, 'W_1KI': 1.121895767750962, 'W_D': 79.97325000000001, 'J_D': 1067.989138565898, 'W_D_1KI': 0.7770126501107615, 'J_D_1KI': 0.00754938255519375}
[40.36, 39.67, 39.74, 39.69, 39.75, 39.62, 40.16, 41.41, 40.17, 40.09]
[115.1]
13.35277509689331
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 99857, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.432250738143921, 'TIME_S_1KI': 0.10447190220158749, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1536.9044136524199, 'W': 115.09999999999998}
[40.36, 39.67, 39.74, 39.69, 39.75, 39.62, 40.16, 41.41, 40.17, 40.09, 40.33, 39.61, 39.56, 41.58, 39.51, 39.51, 39.84, 39.34, 39.38, 39.82]
718.8400000000001
35.94200000000001
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 99857, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.432250738143921, 'TIME_S_1KI': 0.10447190220158749, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1536.9044136524199, 'W': 115.09999999999998, 'J_1KI': 15.391053342804408, 'W_1KI': 1.152648287050482, 'W_D': 79.15799999999997, 'J_D': 1056.9789711198803, 'W_D_1KI': 0.7927135804200004, 'J_D_1KI': 0.007938487841813797}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 85448, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.79839825630188, "TIME_S_1KI": 0.12637391461826936, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1847.2158340501787, "W": 132.18, "J_1KI": 21.618011352520583, "W_1KI": 1.5469057204381613, "W_D": 96.35400000000001, "J_D": 1346.5473935093883, "W_D_1KI": 1.127633180413819, "J_D_1KI": 0.013196718242835633}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 81276, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.078009605407715, "TIME_S_1KI": 0.12399736214144047, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1722.4750410318375, "W": 132.75, "J_1KI": 21.192911081153568, "W_1KI": 1.6333234903292484, "W_D": 96.411, "J_D": 1250.9645286698342, "W_D_1KI": 1.1862173335301935, "J_D_1KI": 0.014594927574317062}

68
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_100000_5e-05.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '100000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.157515287399292}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '100000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.04085874557495117}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2, 8, ..., 499988, 499996,
tensor(crow_indices=tensor([ 0, 4, 9, ..., 499987, 499993,
500000]),
col_indices=tensor([50162, 75153, 30191, ..., 32389, 47580, 60210]),
values=tensor([0.9007, 0.9447, 0.0410, ..., 0.6472, 0.2952, 0.4267]),
col_indices=tensor([ 4658, 51132, 55767, ..., 77897, 84680, 91168]),
values=tensor([0.8716, 0.7460, 0.9968, ..., 0.7762, 0.8585, 0.9878]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.3259, 0.8902, 0.7186, ..., 0.8330, 0.5312, 0.8917])
tensor([0.7678, 0.5187, 0.4774, ..., 0.8664, 0.3724, 0.0254])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -16,19 +16,19 @@ Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 0.157515287399292 seconds
Time: 0.04085874557495117 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '66660', '-ss', '100000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 8.191283702850342}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '25698', '-ss', '100000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 3.3198819160461426}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 4, 8, ..., 499990, 499997,
tensor(crow_indices=tensor([ 0, 4, 4, ..., 499992, 499998,
500000]),
col_indices=tensor([ 3937, 41482, 51345, ..., 57028, 62776, 96568]),
values=tensor([0.3669, 0.7790, 0.6636, ..., 0.0088, 0.3191, 0.1015]),
col_indices=tensor([33478, 35089, 63624, ..., 93258, 3464, 77760]),
values=tensor([0.8303, 0.5286, 0.9064, ..., 0.8655, 0.5788, 0.5903]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.1888, 0.6317, 0.9833, ..., 0.5078, 0.6417, 0.5906])
tensor([0.0892, 0.6340, 0.1475, ..., 0.5230, 0.0009, 0.8265])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -36,19 +36,19 @@ Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 8.191283702850342 seconds
Time: 3.3198819160461426 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '85448', '-ss', '100000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.79839825630188}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '81276', '-ss', '100000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [100000, 100000], "MATRIX_ROWS": 100000, "MATRIX_SIZE": 10000000000, "MATRIX_NNZ": 500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.078009605407715}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 8, 13, ..., 499988, 499995,
tensor(crow_indices=tensor([ 0, 3, 9, ..., 499997, 500000,
500000]),
col_indices=tensor([ 3698, 26087, 35796, ..., 95832, 96289, 98226]),
values=tensor([0.4478, 0.6896, 0.5878, ..., 0.3885, 0.0788, 0.0500]),
col_indices=tensor([38450, 44227, 69625, ..., 8507, 39094, 82179]),
values=tensor([0.2677, 0.9845, 0.1042, ..., 0.9974, 0.0756, 0.3422]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.6913, 0.6407, 0.8664, ..., 0.8625, 0.1823, 0.9429])
tensor([0.8400, 0.1962, 0.3075, ..., 0.6034, 0.5737, 0.0994])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -56,16 +56,16 @@ Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 10.79839825630188 seconds
Time: 10.078009605407715 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 8, 13, ..., 499988, 499995,
tensor(crow_indices=tensor([ 0, 3, 9, ..., 499997, 500000,
500000]),
col_indices=tensor([ 3698, 26087, 35796, ..., 95832, 96289, 98226]),
values=tensor([0.4478, 0.6896, 0.5878, ..., 0.3885, 0.0788, 0.0500]),
col_indices=tensor([38450, 44227, 69625, ..., 8507, 39094, 82179]),
values=tensor([0.2677, 0.9845, 0.1042, ..., 0.9974, 0.0756, 0.3422]),
size=(100000, 100000), nnz=500000, layout=torch.sparse_csr)
tensor([0.6913, 0.6407, 0.8664, ..., 0.8625, 0.1823, 0.9429])
tensor([0.8400, 0.1962, 0.3075, ..., 0.6034, 0.5737, 0.0994])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([100000, 100000])
@ -73,13 +73,13 @@ Rows: 100000
Size: 10000000000
NNZ: 500000
Density: 5e-05
Time: 10.79839825630188 seconds
Time: 10.078009605407715 seconds
[40.27, 39.58, 39.76, 40.26, 40.31, 39.92, 40.36, 39.5, 39.65, 39.54]
[132.18]
13.975002527236938
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 85448, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.79839825630188, 'TIME_S_1KI': 0.12637391461826936, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1847.2158340501787, 'W': 132.18}
[40.27, 39.58, 39.76, 40.26, 40.31, 39.92, 40.36, 39.5, 39.65, 39.54, 40.41, 39.45, 40.31, 39.36, 39.58, 39.39, 39.62, 39.75, 39.86, 39.5]
716.52
35.826
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 85448, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.79839825630188, 'TIME_S_1KI': 0.12637391461826936, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1847.2158340501787, 'W': 132.18, 'J_1KI': 21.618011352520583, 'W_1KI': 1.5469057204381613, 'W_D': 96.35400000000001, 'J_D': 1346.5473935093883, 'W_D_1KI': 1.127633180413819, 'J_D_1KI': 0.013196718242835633}
[41.19, 39.74, 39.62, 40.23, 39.69, 39.5, 44.76, 39.57, 39.56, 39.98]
[132.75]
12.975329875946045
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 81276, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.078009605407715, 'TIME_S_1KI': 0.12399736214144047, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1722.4750410318375, 'W': 132.75}
[41.19, 39.74, 39.62, 40.23, 39.69, 39.5, 44.76, 39.57, 39.56, 39.98, 40.29, 39.57, 39.62, 45.06, 39.69, 39.48, 40.04, 39.95, 40.05, 39.84]
726.78
36.339
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 81276, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [100000, 100000], 'MATRIX_ROWS': 100000, 'MATRIX_SIZE': 10000000000, 'MATRIX_NNZ': 500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.078009605407715, 'TIME_S_1KI': 0.12399736214144047, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1722.4750410318375, 'W': 132.75, 'J_1KI': 21.192911081153568, 'W_1KI': 1.6333234903292484, 'W_D': 96.411, 'J_D': 1250.9645286698342, 'W_D_1KI': 1.1862173335301935, 'J_D_1KI': 0.014594927574317062}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 278690, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.3841392993927, "TIME_S_1KI": 0.0372605378714439, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1292.3170569992064, "W": 98.52, "J_1KI": 4.63711312569237, "W_1KI": 0.3535110696472783, "W_D": 63.16824999999999, "J_D": 828.5973095390796, "W_D_1KI": 0.2266613441458251, "J_D_1KI": 0.0008133099291177477}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 280711, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.449347734451294, "TIME_S_1KI": 0.03722457521953644, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1289.5745413303375, "W": 98.5, "J_1KI": 4.593957989998032, "W_1KI": 0.35089469240606885, "W_D": 62.83475, "J_D": 822.6405473183394, "W_D_1KI": 0.22384142409809377, "J_D_1KI": 0.0007974088086968226}

68
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.0001.output
View File

@ -1,13 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.05305743217468262}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.019724130630493164}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 9999, 9999, 10000]),
col_indices=tensor([2207, 830, 7633, ..., 2513, 8541, 2972]),
values=tensor([0.9417, 0.1071, 0.2127, ..., 0.2034, 0.4535, 0.3737]),
tensor(crow_indices=tensor([ 0, 0, 0, ..., 9997, 9999, 10000]),
col_indices=tensor([ 730, 4220, 7544, ..., 4458, 7562, 5619]),
values=tensor([0.0181, 0.7832, 0.5914, ..., 0.2469, 0.2734, 0.2796]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.2095, 0.5712, 0.5435, ..., 0.2564, 0.5818, 0.1577])
tensor([0.6994, 0.7339, 0.7582, ..., 0.9456, 0.1186, 0.3856])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -15,18 +15,18 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 0.05305743217468262 seconds
Time: 0.019724130630493164 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '197898', '-ss', '10000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 7.456049680709839}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '53234', '-ss', '10000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 1.991217851638794}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 2, ..., 10000, 10000, 10000]),
col_indices=tensor([7930, 9951, 4041, ..., 9045, 6420, 8503]),
values=tensor([0.2418, 0.2435, 0.4116, ..., 0.5201, 0.9725, 0.0713]),
tensor(crow_indices=tensor([ 0, 5, 5, ..., 9999, 9999, 10000]),
col_indices=tensor([2031, 5960, 7493, ..., 3747, 8534, 6060]),
values=tensor([0.1847, 0.1000, 0.1920, ..., 0.9911, 0.4392, 0.2330]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.5895, 0.0291, 0.5304, ..., 0.4324, 0.9976, 0.6205])
tensor([0.7239, 0.0636, 0.4781, ..., 0.2276, 0.2279, 0.8613])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -34,18 +34,18 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 7.456049680709839 seconds
Time: 1.991217851638794 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '278690', '-ss', '10000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.3841392993927}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '280711', '-ss', '10000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.449347734451294}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 9996, 9998, 10000]),
col_indices=tensor([9574, 4944, 2003, ..., 2641, 7523, 8416]),
values=tensor([0.4157, 0.2537, 0.8916, ..., 0.3966, 0.1591, 0.0732]),
tensor(crow_indices=tensor([ 0, 0, 1, ..., 9997, 9999, 10000]),
col_indices=tensor([8732, 42, 2512, ..., 1373, 9550, 9690]),
values=tensor([0.4706, 0.1126, 0.6045, ..., 0.0102, 0.1178, 0.6557]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.4368, 0.0363, 0.2687, ..., 0.3029, 0.2331, 0.6830])
tensor([0.4976, 0.6299, 0.3127, ..., 0.9623, 0.9434, 0.7070])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -53,15 +53,15 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 10.3841392993927 seconds
Time: 10.449347734451294 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 9996, 9998, 10000]),
col_indices=tensor([9574, 4944, 2003, ..., 2641, 7523, 8416]),
values=tensor([0.4157, 0.2537, 0.8916, ..., 0.3966, 0.1591, 0.0732]),
tensor(crow_indices=tensor([ 0, 0, 1, ..., 9997, 9999, 10000]),
col_indices=tensor([8732, 42, 2512, ..., 1373, 9550, 9690]),
values=tensor([0.4706, 0.1126, 0.6045, ..., 0.0102, 0.1178, 0.6557]),
size=(10000, 10000), nnz=10000, layout=torch.sparse_csr)
tensor([0.4368, 0.0363, 0.2687, ..., 0.3029, 0.2331, 0.6830])
tensor([0.4976, 0.6299, 0.3127, ..., 0.9623, 0.9434, 0.7070])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -69,13 +69,13 @@ Rows: 10000
Size: 100000000
NNZ: 10000
Density: 0.0001
Time: 10.3841392993927 seconds
Time: 10.449347734451294 seconds
[40.74, 38.88, 38.93, 39.04, 38.99, 38.97, 39.42, 39.32, 39.26, 38.71]
[98.52]
13.11730670928955
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 278690, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.3841392993927, 'TIME_S_1KI': 0.0372605378714439, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1292.3170569992064, 'W': 98.52}
[40.74, 38.88, 38.93, 39.04, 38.99, 38.97, 39.42, 39.32, 39.26, 38.71, 39.37, 39.1, 39.73, 39.04, 39.15, 38.73, 39.2, 38.61, 38.78, 44.95]
707.0350000000001
35.35175
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 278690, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.3841392993927, 'TIME_S_1KI': 0.0372605378714439, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1292.3170569992064, 'W': 98.52, 'J_1KI': 4.63711312569237, 'W_1KI': 0.3535110696472783, 'W_D': 63.16824999999999, 'J_D': 828.5973095390796, 'W_D_1KI': 0.2266613441458251, 'J_D_1KI': 0.0008133099291177477}
[40.09, 39.2, 39.41, 39.39, 39.38, 42.02, 40.51, 39.25, 39.25, 39.51]
[98.5]
13.092127323150635
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 280711, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.449347734451294, 'TIME_S_1KI': 0.03722457521953644, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1289.5745413303375, 'W': 98.5}
[40.09, 39.2, 39.41, 39.39, 39.38, 42.02, 40.51, 39.25, 39.25, 39.51, 40.09, 39.17, 39.95, 40.04, 39.24, 39.23, 39.51, 39.16, 39.2, 39.1]
713.3050000000001
35.66525
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 280711, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.449347734451294, 'TIME_S_1KI': 0.03722457521953644, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1289.5745413303375, 'W': 98.5, 'J_1KI': 4.593957989998032, 'W_1KI': 0.35089469240606885, 'W_D': 62.83475, 'J_D': 822.6405473183394, 'W_D_1KI': 0.22384142409809377, 'J_D_1KI': 0.0007974088086968226}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.001.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 181643, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.058825254440308, "TIME_S_1KI": 0.05537689453730839, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1334.3994569778442, "W": 108.0, "J_1KI": 7.346275149484671, "W_1KI": 0.5945728709611711, "W_D": 72.86524999999999, "J_D": 900.2902780792116, "W_D_1KI": 0.4011453785722543, "J_D_1KI": 0.002208427401949177}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 193546, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.657205820083618, "TIME_S_1KI": 0.05506290917964524, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1424.268603067398, "W": 108.67, "J_1KI": 7.358811874528009, "W_1KI": 0.5614685914459612, "W_D": 72.91375, "J_D": 955.6341663467883, "W_D_1KI": 0.3767256879501513, "J_D_1KI": 0.001946440060503195}

89
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.07387351989746094}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.025844097137451172}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 11, 20, ..., 99983, 99988,
tensor(crow_indices=tensor([ 0, 12, 28, ..., 99969, 99983,
100000]),
col_indices=tensor([2080, 2520, 2867, ..., 8307, 8901, 9286]),
values=tensor([0.8261, 0.1055, 0.9939, ..., 0.1447, 0.1951, 0.2617]),
col_indices=tensor([1079, 2122, 3254, ..., 9373, 9823, 9958]),
values=tensor([0.1589, 0.8596, 0.7837, ..., 0.1493, 0.1272, 0.2084]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.7373, 0.8108, 0.8070, ..., 0.3032, 0.8916, 0.0356])
tensor([0.0719, 0.4122, 0.7875, ..., 0.0407, 0.8322, 0.6511])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,20 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 0.07387351989746094 seconds
Time: 0.025844097137451172 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '142134', '-ss', '10000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 8.216149806976318}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '40628', '-ss', '10000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 2.4707634449005127}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 8, 16, ..., 99977, 99988,
tensor(crow_indices=tensor([ 0, 12, 18, ..., 99974, 99987,
100000]),
col_indices=tensor([ 929, 1145, 1167, ..., 7253, 9439, 9881]),
values=tensor([3.5267e-01, 8.9746e-01, 4.0379e-01, ...,
8.5718e-04, 5.6681e-01, 4.6851e-01]),
col_indices=tensor([ 792, 1032, 1238, ..., 8561, 8731, 9370]),
values=tensor([0.4488, 0.9659, 0.1268, ..., 0.7863, 0.6709, 0.3638]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.4055, 0.0658, 0.7904, ..., 0.2959, 0.0826, 0.7426])
tensor([0.8213, 0.7389, 0.9585, ..., 0.8858, 0.0787, 0.3979])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -37,19 +36,19 @@ Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 8.216149806976318 seconds
Time: 2.4707634449005127 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '181643', '-ss', '10000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.058825254440308}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '172656', '-ss', '10000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 9.366683721542358}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 10, 23, ..., 99984, 99996,
tensor(crow_indices=tensor([ 0, 10, 24, ..., 99973, 99986,
100000]),
col_indices=tensor([2026, 2065, 2399, ..., 4623, 7297, 9355]),
values=tensor([0.4157, 0.6883, 0.2119, ..., 0.3441, 0.2622, 0.5721]),
col_indices=tensor([ 684, 3301, 3344, ..., 8499, 8709, 9229]),
values=tensor([0.0104, 0.6771, 0.5927, ..., 0.6883, 0.2524, 0.4550]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.4888, 0.3451, 0.6891, ..., 0.9797, 0.8702, 0.1612])
tensor([0.4786, 0.6837, 0.1379, ..., 0.3005, 0.2266, 0.1673])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -57,16 +56,19 @@ Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 10.058825254440308 seconds
Time: 9.366683721542358 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '193546', '-ss', '10000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 100000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.657205820083618}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 10, 23, ..., 99984, 99996,
tensor(crow_indices=tensor([ 0, 13, 24, ..., 99982, 99990,
100000]),
col_indices=tensor([2026, 2065, 2399, ..., 4623, 7297, 9355]),
values=tensor([0.4157, 0.6883, 0.2119, ..., 0.3441, 0.2622, 0.5721]),
col_indices=tensor([ 667, 823, 2535, ..., 7218, 8112, 8309]),
values=tensor([0.9044, 0.9079, 0.6825, ..., 0.1587, 0.6143, 0.0618]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.4888, 0.3451, 0.6891, ..., 0.9797, 0.8702, 0.1612])
tensor([0.5914, 0.6686, 0.5823, ..., 0.5362, 0.3609, 0.2297])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -74,13 +76,30 @@ Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 10.058825254440308 seconds
Time: 10.657205820083618 seconds
[40.01, 39.91, 39.26, 39.32, 39.03, 38.69, 39.03, 38.68, 38.9, 38.67]
[108.0]
12.355550527572632
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 181643, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.058825254440308, 'TIME_S_1KI': 0.05537689453730839, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1334.3994569778442, 'W': 108.0}
[40.01, 39.91, 39.26, 39.32, 39.03, 38.69, 39.03, 38.68, 38.9, 38.67, 40.03, 39.23, 39.15, 39.23, 38.85, 38.69, 38.72, 38.72, 38.62, 38.62]
702.6950000000002
35.13475000000001
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 181643, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.058825254440308, 'TIME_S_1KI': 0.05537689453730839, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1334.3994569778442, 'W': 108.0, 'J_1KI': 7.346275149484671, 'W_1KI': 0.5945728709611711, 'W_D': 72.86524999999999, 'J_D': 900.2902780792116, 'W_D_1KI': 0.4011453785722543, 'J_D_1KI': 0.002208427401949177}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 13, 24, ..., 99982, 99990,
100000]),
col_indices=tensor([ 667, 823, 2535, ..., 7218, 8112, 8309]),
values=tensor([0.9044, 0.9079, 0.6825, ..., 0.1587, 0.6143, 0.0618]),
size=(10000, 10000), nnz=100000, layout=torch.sparse_csr)
tensor([0.5914, 0.6686, 0.5823, ..., 0.5362, 0.3609, 0.2297])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
Rows: 10000
Size: 100000000
NNZ: 100000
Density: 0.001
Time: 10.657205820083618 seconds
[40.9, 39.63, 39.53, 39.38, 39.44, 39.45, 39.92, 39.64, 39.85, 39.99]
[108.67]
13.106364250183105
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 193546, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.657205820083618, 'TIME_S_1KI': 0.05506290917964524, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1424.268603067398, 'W': 108.67}
[40.9, 39.63, 39.53, 39.38, 39.44, 39.45, 39.92, 39.64, 39.85, 39.99, 39.96, 39.48, 39.45, 39.41, 39.63, 39.38, 40.94, 39.79, 39.91, 39.74]
715.125
35.75625
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 193546, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 100000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.657205820083618, 'TIME_S_1KI': 0.05506290917964524, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1424.268603067398, 'W': 108.67, 'J_1KI': 7.358811874528009, 'W_1KI': 0.5614685914459612, 'W_D': 72.91375, 'J_D': 955.6341663467883, 'W_D_1KI': 0.3767256879501513, 'J_D_1KI': 0.001946440060503195}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.01.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 104114, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.469164609909058, "TIME_S_1KI": 0.10055482077250953, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1774.261237783432, "W": 135.86, "J_1KI": 17.041524077294426, "W_1KI": 1.304915765410992, "W_D": 100.35275000000001, "J_D": 1310.5549420725108, "W_D_1KI": 0.9638737345601938, "J_D_1KI": 0.009257868630157269}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 102691, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.217256307601929, "TIME_S_1KI": 0.09949514862648069, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1713.9247414016722, "W": 132.92, "J_1KI": 16.690116382172462, "W_1KI": 1.294368542520766, "W_D": 96.75025, "J_D": 1247.5372194688318, "W_D_1KI": 0.9421492633239523, "J_D_1KI": 0.009174604038561825}

96
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.01.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '0.01', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 0.14159297943115234}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '0.01', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 0.027860403060913086}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 124, 236, ..., 999773,
999882, 1000000]),
col_indices=tensor([ 35, 69, 144, ..., 9773, 9862, 9873]),
values=tensor([0.1838, 0.7773, 0.5109, ..., 0.8192, 0.8376, 0.6812]),
tensor(crow_indices=tensor([ 0, 86, 184, ..., 999787,
999901, 1000000]),
col_indices=tensor([ 81, 93, 211, ..., 9891, 9936, 9983]),
values=tensor([0.0273, 0.9948, 0.2764, ..., 0.0318, 0.5538, 0.8532]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.0358, 0.2032, 0.7087, ..., 0.4931, 0.1706, 0.1726])
tensor([0.8459, 0.7440, 0.9932, ..., 0.5464, 0.7654, 0.2266])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,19 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 0.14159297943115234 seconds
Time: 0.027860403060913086 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '74156', '-ss', '10000', '-sd', '0.01', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 7.9134438037872314}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '37687', '-ss', '10000', '-sd', '0.01', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 3.853411912918091}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 93, 199, ..., 999798,
999892, 1000000]),
col_indices=tensor([ 57, 323, 325, ..., 9719, 9779, 9889]),
values=tensor([0.3339, 0.1610, 0.8675, ..., 0.7107, 0.3615, 0.1870]),
tensor(crow_indices=tensor([ 0, 90, 178, ..., 999807,
999899, 1000000]),
col_indices=tensor([ 9, 87, 435, ..., 9776, 9821, 9947]),
values=tensor([0.6051, 0.3509, 0.6551, ..., 0.3060, 0.1178, 0.2325]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.9536, 0.3002, 0.1616, ..., 0.3121, 0.8413, 0.9505])
tensor([0.6802, 0.0969, 0.8232, ..., 0.8757, 0.6573, 0.4893])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -36,19 +36,19 @@ Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 7.9134438037872314 seconds
Time: 3.853411912918091 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '98394', '-ss', '10000', '-sd', '0.01', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 9.923112392425537}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '102691', '-ss', '10000', '-sd', '0.01', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.217256307601929}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 97, 191, ..., 999779,
999891, 1000000]),
col_indices=tensor([ 18, 52, 269, ..., 9883, 9995, 9999]),
values=tensor([0.5511, 0.2767, 0.8168, ..., 0.6887, 0.5827, 0.0686]),
tensor(crow_indices=tensor([ 0, 98, 191, ..., 999800,
999904, 1000000]),
col_indices=tensor([ 18, 19, 89, ..., 9675, 9719, 9959]),
values=tensor([0.5811, 0.2000, 0.4195, ..., 0.8918, 0.7545, 0.5786]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.2767, 0.4380, 0.7945, ..., 0.2102, 0.5547, 0.8740])
tensor([0.3032, 0.6522, 0.8844, ..., 0.7793, 0.6874, 0.5546])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -56,19 +56,16 @@ Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 9.923112392425537 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '104114', '-ss', '10000', '-sd', '0.01', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000000, "MATRIX_DENSITY": 0.01, "TIME_S": 10.469164609909058}
Time: 10.217256307601929 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 112, 221, ..., 999805,
999915, 1000000]),
col_indices=tensor([ 402, 501, 665, ..., 9291, 9326, 9607]),
values=tensor([0.0486, 0.5637, 0.4384, ..., 0.7973, 0.3634, 0.8351]),
tensor(crow_indices=tensor([ 0, 98, 191, ..., 999800,
999904, 1000000]),
col_indices=tensor([ 18, 19, 89, ..., 9675, 9719, 9959]),
values=tensor([0.5811, 0.2000, 0.4195, ..., 0.8918, 0.7545, 0.5786]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.7936, 0.9785, 0.9590, ..., 0.6005, 0.0137, 0.6516])
tensor([0.3032, 0.6522, 0.8844, ..., 0.7793, 0.6874, 0.5546])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -76,30 +73,13 @@ Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 10.469164609909058 seconds
Time: 10.217256307601929 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 112, 221, ..., 999805,
999915, 1000000]),
col_indices=tensor([ 402, 501, 665, ..., 9291, 9326, 9607]),
values=tensor([0.0486, 0.5637, 0.4384, ..., 0.7973, 0.3634, 0.8351]),
size=(10000, 10000), nnz=1000000, layout=torch.sparse_csr)
tensor([0.7936, 0.9785, 0.9590, ..., 0.6005, 0.0137, 0.6516])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
Rows: 10000
Size: 100000000
NNZ: 1000000
Density: 0.01
Time: 10.469164609909058 seconds
[42.04, 39.8, 39.79, 39.52, 39.3, 39.73, 39.12, 39.49, 39.14, 38.98]
[135.86]
13.059482097625732
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 104114, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.469164609909058, 'TIME_S_1KI': 0.10055482077250953, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1774.261237783432, 'W': 135.86}
[42.04, 39.8, 39.79, 39.52, 39.3, 39.73, 39.12, 39.49, 39.14, 38.98, 39.91, 39.55, 39.06, 39.42, 39.43, 39.34, 38.96, 39.36, 38.94, 39.46]
710.145
35.50725
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 104114, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.469164609909058, 'TIME_S_1KI': 0.10055482077250953, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1774.261237783432, 'W': 135.86, 'J_1KI': 17.041524077294426, 'W_1KI': 1.304915765410992, 'W_D': 100.35275000000001, 'J_D': 1310.5549420725108, 'W_D_1KI': 0.9638737345601938, 'J_D_1KI': 0.009257868630157269}
[40.93, 39.91, 40.43, 39.4, 39.51, 40.07, 39.51, 39.35, 39.39, 44.7]
[132.92]
12.894408226013184
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 102691, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.217256307601929, 'TIME_S_1KI': 0.09949514862648069, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1713.9247414016722, 'W': 132.92}
[40.93, 39.91, 40.43, 39.4, 39.51, 40.07, 39.51, 39.35, 39.39, 44.7, 40.04, 39.9, 39.44, 39.42, 39.81, 39.87, 45.14, 39.82, 39.78, 39.62]
723.395
36.16975
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 102691, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 1000000, 'MATRIX_DENSITY': 0.01, 'TIME_S': 10.217256307601929, 'TIME_S_1KI': 0.09949514862648069, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1713.9247414016722, 'W': 132.92, 'J_1KI': 16.690116382172462, 'W_1KI': 1.294368542520766, 'W_D': 96.75025, 'J_D': 1247.5372194688318, 'W_D_1KI': 0.9421492633239523, 'J_D_1KI': 0.009174604038561825}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.05.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 27894, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.822905540466309, "TIME_S_1KI": 0.3880012024258374, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2101.6757108688353, "W": 151.95, "J_1KI": 75.34508176915593, "W_1KI": 5.447408044740804, "W_D": 116.03349999999999, "J_D": 1604.90153732872, "W_D_1KI": 4.159801390980138, "J_D_1KI": 0.14912889477952743}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 27775, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.346900463104248, "TIME_S_1KI": 0.3725256692386768, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2138.1740951538086, "W": 151.25, "J_1KI": 76.98196562209931, "W_1KI": 5.445544554455445, "W_D": 115.164, "J_D": 1628.037563598633, "W_D_1KI": 4.146318631863187, "J_D_1KI": 0.1492823989869734}

76
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.05.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '0.05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 0.4579291343688965}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '0.05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 0.07912850379943848}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 485, 1001, ..., 4998993,
4999541, 5000000]),
col_indices=tensor([ 5, 20, 61, ..., 9897, 9942, 9998]),
values=tensor([0.7241, 0.0945, 0.6836, ..., 0.9220, 0.2796, 0.2745]),
tensor(crow_indices=tensor([ 0, 485, 973, ..., 4998984,
4999512, 5000000]),
col_indices=tensor([ 23, 33, 35, ..., 9878, 9920, 9946]),
values=tensor([0.8956, 0.5440, 0.5650, ..., 0.6571, 0.0981, 0.4530]),
size=(10000, 10000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.6528, 0.9454, 0.7224, ..., 0.5670, 0.2826, 0.8750])
tensor([0.3320, 0.0557, 0.6993, ..., 0.8374, 0.3528, 0.6849])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,19 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 5000000
Density: 0.05
Time: 0.4579291343688965 seconds
Time: 0.07912850379943848 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '22929', '-ss', '10000', '-sd', '0.05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 8.630967855453491}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '13269', '-ss', '10000', '-sd', '0.05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 5.016182899475098}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 483, 987, ..., 4998961,
4999465, 5000000]),
col_indices=tensor([ 47, 67, 96, ..., 9993, 9994, 9997]),
values=tensor([0.9705, 0.3882, 0.2458, ..., 0.5796, 0.8899, 0.6056]),
tensor(crow_indices=tensor([ 0, 490, 975, ..., 4999017,
4999511, 5000000]),
col_indices=tensor([ 5, 7, 17, ..., 9925, 9927, 9956]),
values=tensor([0.3061, 0.0982, 0.7519, ..., 0.4711, 0.1343, 0.2753]),
size=(10000, 10000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.2409, 0.7584, 0.7571, ..., 0.5444, 0.5564, 0.6333])
tensor([0.4300, 0.5593, 0.7816, ..., 0.7590, 0.1985, 0.5681])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -36,19 +36,19 @@ Rows: 10000
Size: 100000000
NNZ: 5000000
Density: 0.05
Time: 8.630967855453491 seconds
Time: 5.016182899475098 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '27894', '-ss', '10000', '-sd', '0.05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.822905540466309}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '27775', '-ss', '10000', '-sd', '0.05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000000, "MATRIX_DENSITY": 0.05, "TIME_S": 10.346900463104248}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 505, 994, ..., 4999000,
4999548, 5000000]),
col_indices=tensor([ 26, 89, 94, ..., 9963, 9967, 9969]),
values=tensor([0.5738, 0.3729, 0.9664, ..., 0.7914, 0.9130, 0.5932]),
tensor(crow_indices=tensor([ 0, 516, 985, ..., 4998986,
4999503, 5000000]),
col_indices=tensor([ 0, 38, 62, ..., 9969, 9984, 9993]),
values=tensor([0.4538, 0.1922, 0.3497, ..., 0.8541, 0.7038, 0.0561]),
size=(10000, 10000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.2560, 0.0745, 0.3692, ..., 0.2024, 0.4618, 0.6540])
tensor([0.3516, 0.9610, 0.6827, ..., 0.5287, 0.4040, 0.0575])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -56,16 +56,16 @@ Rows: 10000
Size: 100000000
NNZ: 5000000
Density: 0.05
Time: 10.822905540466309 seconds
Time: 10.346900463104248 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 505, 994, ..., 4999000,
4999548, 5000000]),
col_indices=tensor([ 26, 89, 94, ..., 9963, 9967, 9969]),
values=tensor([0.5738, 0.3729, 0.9664, ..., 0.7914, 0.9130, 0.5932]),
tensor(crow_indices=tensor([ 0, 516, 985, ..., 4998986,
4999503, 5000000]),
col_indices=tensor([ 0, 38, 62, ..., 9969, 9984, 9993]),
values=tensor([0.4538, 0.1922, 0.3497, ..., 0.8541, 0.7038, 0.0561]),
size=(10000, 10000), nnz=5000000, layout=torch.sparse_csr)
tensor([0.2560, 0.0745, 0.3692, ..., 0.2024, 0.4618, 0.6540])
tensor([0.3516, 0.9610, 0.6827, ..., 0.5287, 0.4040, 0.0575])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -73,13 +73,13 @@ Rows: 10000
Size: 100000000
NNZ: 5000000
Density: 0.05
Time: 10.822905540466309 seconds
Time: 10.346900463104248 seconds
[41.38, 40.1, 39.88, 40.17, 39.72, 39.62, 39.69, 39.64, 39.9, 39.73]
[151.95]
13.831363677978516
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 27894, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.822905540466309, 'TIME_S_1KI': 0.3880012024258374, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2101.6757108688353, 'W': 151.95}
[41.38, 40.1, 39.88, 40.17, 39.72, 39.62, 39.69, 39.64, 39.9, 39.73, 40.26, 39.63, 39.55, 39.93, 41.22, 40.33, 39.48, 39.56, 39.48, 39.49]
718.3299999999999
35.9165
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 27894, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.822905540466309, 'TIME_S_1KI': 0.3880012024258374, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2101.6757108688353, 'W': 151.95, 'J_1KI': 75.34508176915593, 'W_1KI': 5.447408044740804, 'W_D': 116.03349999999999, 'J_D': 1604.90153732872, 'W_D_1KI': 4.159801390980138, 'J_D_1KI': 0.14912889477952743}
[41.25, 39.91, 39.68, 39.83, 39.58, 41.43, 40.59, 39.78, 40.26, 39.69]
[151.25]
14.136688232421875
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 27775, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.346900463104248, 'TIME_S_1KI': 0.3725256692386768, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2138.1740951538086, 'W': 151.25}
[41.25, 39.91, 39.68, 39.83, 39.58, 41.43, 40.59, 39.78, 40.26, 39.69, 40.82, 40.72, 40.11, 39.71, 39.6, 39.85, 39.65, 39.99, 39.77, 40.76]
721.72
36.086
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 27775, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000000, 'MATRIX_DENSITY': 0.05, 'TIME_S': 10.346900463104248, 'TIME_S_1KI': 0.3725256692386768, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2138.1740951538086, 'W': 151.25, 'J_1KI': 76.98196562209931, 'W_1KI': 5.445544554455445, 'W_D': 115.164, 'J_D': 1628.037563598633, 'W_D_1KI': 4.146318631863187, 'J_D_1KI': 0.1492823989869734}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.1.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 4679, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 11.427535057067871, "TIME_S_1KI": 2.4423028546843066, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1873.982270450592, "W": 124.72, "J_1KI": 400.5091409383612, "W_1KI": 26.655268219705064, "W_D": 88.71, "J_D": 1332.9134638524054, "W_D_1KI": 18.959179311818765, "J_D_1KI": 4.051972496648593}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 4427, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 10.78234076499939, "TIME_S_1KI": 2.4355863485428935, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1855.6078462219239, "W": 121.94, "J_1KI": 419.15695645401485, "W_1KI": 27.544612604472555, "W_D": 85.48649999999999, "J_D": 1300.8809262428283, "W_D_1KI": 19.310255251863563, "J_D_1KI": 4.361927999065633}

78
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.1.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '0.1', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 2.382111072540283}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '0.1', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 0.23713254928588867}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1024, 2032, ..., 9997994,
tensor(crow_indices=tensor([ 0, 967, 1927, ..., 9997983,
9998974, 10000000]),
col_indices=tensor([ 25, 59, 80, ..., 9969, 9975, 9986]),
values=tensor([0.6759, 0.5147, 0.7066, ..., 0.5276, 0.4088, 0.2550]),
col_indices=tensor([ 2, 7, 17, ..., 9977, 9981, 9986]),
values=tensor([0.0113, 0.4578, 0.3712, ..., 0.8300, 0.4518, 0.5288]),
size=(10000, 10000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.6863, 0.6243, 0.0191, ..., 0.9166, 0.1487, 0.8503])
tensor([0.6464, 0.5946, 0.9135, ..., 0.7384, 0.8851, 0.3138])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,19 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 10000000
Density: 0.1
Time: 2.382111072540283 seconds
Time: 0.23713254928588867 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '4407', '-ss', '10000', '-sd', '0.1', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 9.887728929519653}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '4427', '-ss', '10000', '-sd', '0.1', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 10.78234076499939}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 980, 2002, ..., 9998027,
9998983, 10000000]),
col_indices=tensor([ 0, 5, 25, ..., 9979, 9984, 9986]),
values=tensor([0.6732, 0.9055, 0.4649, ..., 0.1468, 0.7629, 0.6148]),
tensor(crow_indices=tensor([ 0, 1000, 1968, ..., 9997976,
9998957, 10000000]),
col_indices=tensor([ 18, 35, 37, ..., 9972, 9974, 9993]),
values=tensor([0.5495, 0.5155, 0.6909, ..., 0.5748, 0.2988, 0.6189]),
size=(10000, 10000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.4732, 0.0327, 0.4956, ..., 0.7189, 0.9869, 0.4026])
tensor([0.2327, 0.3005, 0.5005, ..., 0.5867, 0.2890, 0.0524])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -36,19 +36,16 @@ Rows: 10000
Size: 100000000
NNZ: 10000000
Density: 0.1
Time: 9.887728929519653 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '4679', '-ss', '10000', '-sd', '0.1', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 10000000, "MATRIX_DENSITY": 0.1, "TIME_S": 11.427535057067871}
Time: 10.78234076499939 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1053, 2097, ..., 9998095,
9999109, 10000000]),
col_indices=tensor([ 10, 15, 24, ..., 9977, 9991, 9995]),
values=tensor([0.4155, 0.1468, 0.0149, ..., 0.0226, 0.9383, 0.4708]),
tensor(crow_indices=tensor([ 0, 1000, 1968, ..., 9997976,
9998957, 10000000]),
col_indices=tensor([ 18, 35, 37, ..., 9972, 9974, 9993]),
values=tensor([0.5495, 0.5155, 0.6909, ..., 0.5748, 0.2988, 0.6189]),
size=(10000, 10000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.3975, 0.8045, 0.4645, ..., 0.1781, 0.4097, 0.1046])
tensor([0.2327, 0.3005, 0.5005, ..., 0.5867, 0.2890, 0.0524])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -56,30 +53,13 @@ Rows: 10000
Size: 100000000
NNZ: 10000000
Density: 0.1
Time: 11.427535057067871 seconds
Time: 10.78234076499939 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1053, 2097, ..., 9998095,
9999109, 10000000]),
col_indices=tensor([ 10, 15, 24, ..., 9977, 9991, 9995]),
values=tensor([0.4155, 0.1468, 0.0149, ..., 0.0226, 0.9383, 0.4708]),
size=(10000, 10000), nnz=10000000, layout=torch.sparse_csr)
tensor([0.3975, 0.8045, 0.4645, ..., 0.1781, 0.4097, 0.1046])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
Rows: 10000
Size: 100000000
NNZ: 10000000
Density: 0.1
Time: 11.427535057067871 seconds
[40.62, 41.67, 40.6, 39.62, 39.77, 39.59, 39.62, 40.85, 40.24, 40.01]
[124.72]
15.02551531791687
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4679, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 11.427535057067871, 'TIME_S_1KI': 2.4423028546843066, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1873.982270450592, 'W': 124.72}
[40.62, 41.67, 40.6, 39.62, 39.77, 39.59, 39.62, 40.85, 40.24, 40.01, 40.22, 39.64, 40.11, 39.47, 39.58, 39.84, 39.97, 39.93, 39.51, 39.53]
720.2
36.010000000000005
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4679, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 11.427535057067871, 'TIME_S_1KI': 2.4423028546843066, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1873.982270450592, 'W': 124.72, 'J_1KI': 400.5091409383612, 'W_1KI': 26.655268219705064, 'W_D': 88.71, 'J_D': 1332.9134638524054, 'W_D_1KI': 18.959179311818765, 'J_D_1KI': 4.051972496648593}
[40.15, 39.73, 39.55, 39.93, 39.94, 39.96, 44.96, 39.82, 40.15, 39.43]
[121.94]
15.217384338378906
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4427, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 10.78234076499939, 'TIME_S_1KI': 2.4355863485428935, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1855.6078462219239, 'W': 121.94}
[40.15, 39.73, 39.55, 39.93, 39.94, 39.96, 44.96, 39.82, 40.15, 39.43, 40.99, 45.07, 39.68, 40.13, 40.01, 39.56, 39.72, 40.75, 40.01, 39.63]
729.07
36.453500000000005
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 4427, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 10000000, 'MATRIX_DENSITY': 0.1, 'TIME_S': 10.78234076499939, 'TIME_S_1KI': 2.4355863485428935, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1855.6078462219239, 'W': 121.94, 'J_1KI': 419.15695645401485, 'W_1KI': 27.544612604472555, 'W_D': 85.48649999999999, 'J_D': 1300.8809262428283, 'W_D_1KI': 19.310255251863563, 'J_D_1KI': 4.361927999065633}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.2.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 2251, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 10.887785911560059, "TIME_S_1KI": 4.83686624236342, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2027.23151884079, "W": 120.43, "J_1KI": 900.591523252239, "W_1KI": 53.500666370501996, "W_D": 84.27900000000001, "J_D": 1418.6917311000825, "W_D_1KI": 37.44069302532208, "J_D_1KI": 16.632915604319006}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 2210, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 10.475984573364258, "TIME_S_1KI": 4.740264512834506, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2065.6901198005676, "W": 119.44, "J_1KI": 934.701411674465, "W_1KI": 54.04524886877828, "W_D": 83.048, "J_D": 1436.2979995746614, "W_D_1KI": 37.57828054298643, "J_D_1KI": 17.00374685203006}

76
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.2.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '0.2', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 4.959461212158203}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '0.2', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 0.5103092193603516}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1995, 4040, ..., 19996024,
19998015, 20000000]),
col_indices=tensor([ 0, 11, 12, ..., 9985, 9992, 9995]),
values=tensor([0.1207, 0.1695, 0.9340, ..., 0.6555, 0.7804, 0.2569]),
tensor(crow_indices=tensor([ 0, 2001, 3993, ..., 19996027,
19997998, 20000000]),
col_indices=tensor([ 4, 8, 12, ..., 9988, 9991, 9998]),
values=tensor([0.1397, 0.5991, 0.8904, ..., 0.1163, 0.3047, 0.7503]),
size=(10000, 10000), nnz=20000000, layout=torch.sparse_csr)
tensor([0.9596, 0.9534, 0.3471, ..., 0.1162, 0.8421, 0.0589])
tensor([0.7325, 0.0863, 0.4494, ..., 0.5445, 0.3494, 0.7015])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,19 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 20000000
Density: 0.2
Time: 4.959461212158203 seconds
Time: 0.5103092193603516 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '2117', '-ss', '10000', '-sd', '0.2', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 9.870691061019897}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '2057', '-ss', '10000', '-sd', '0.2', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 9.769307136535645}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2060, 4088, ..., 19995982,
19997996, 20000000]),
col_indices=tensor([ 3, 8, 18, ..., 9972, 9995, 9999]),
values=tensor([0.9088, 0.2769, 0.7723, ..., 0.9463, 0.8275, 0.8743]),
tensor(crow_indices=tensor([ 0, 1965, 3996, ..., 19995929,
19997992, 20000000]),
col_indices=tensor([ 4, 9, 15, ..., 9975, 9986, 9992]),
values=tensor([0.0708, 0.7889, 0.9973, ..., 0.4384, 0.2830, 0.3299]),
size=(10000, 10000), nnz=20000000, layout=torch.sparse_csr)
tensor([0.1663, 0.5238, 0.4734, ..., 0.4751, 0.9551, 0.4862])
tensor([0.8359, 0.1884, 0.2769, ..., 0.8252, 0.8191, 0.5472])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -36,19 +36,19 @@ Rows: 10000
Size: 100000000
NNZ: 20000000
Density: 0.2
Time: 9.870691061019897 seconds
Time: 9.769307136535645 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '2251', '-ss', '10000', '-sd', '0.2', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 10.887785911560059}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '2210', '-ss', '10000', '-sd', '0.2', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 20000000, "MATRIX_DENSITY": 0.2, "TIME_S": 10.475984573364258}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1987, 4056, ..., 19996028,
19998022, 20000000]),
col_indices=tensor([ 5, 19, 20, ..., 9991, 9993, 9995]),
values=tensor([0.9673, 0.5215, 0.1097, ..., 0.0767, 0.9506, 0.8361]),
tensor(crow_indices=tensor([ 0, 2066, 4070, ..., 19995990,
19998002, 20000000]),
col_indices=tensor([ 1, 2, 8, ..., 9986, 9990, 9993]),
values=tensor([0.6258, 0.8376, 0.0180, ..., 0.7990, 0.4511, 0.0511]),
size=(10000, 10000), nnz=20000000, layout=torch.sparse_csr)
tensor([0.1308, 0.6150, 0.9184, ..., 0.0574, 0.4962, 0.6674])
tensor([0.7373, 0.4078, 0.5568, ..., 0.6016, 0.2858, 0.4434])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -56,16 +56,16 @@ Rows: 10000
Size: 100000000
NNZ: 20000000
Density: 0.2
Time: 10.887785911560059 seconds
Time: 10.475984573364258 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1987, 4056, ..., 19996028,
19998022, 20000000]),
col_indices=tensor([ 5, 19, 20, ..., 9991, 9993, 9995]),
values=tensor([0.9673, 0.5215, 0.1097, ..., 0.0767, 0.9506, 0.8361]),
tensor(crow_indices=tensor([ 0, 2066, 4070, ..., 19995990,
19998002, 20000000]),
col_indices=tensor([ 1, 2, 8, ..., 9986, 9990, 9993]),
values=tensor([0.6258, 0.8376, 0.0180, ..., 0.7990, 0.4511, 0.0511]),
size=(10000, 10000), nnz=20000000, layout=torch.sparse_csr)
tensor([0.1308, 0.6150, 0.9184, ..., 0.0574, 0.4962, 0.6674])
tensor([0.7373, 0.4078, 0.5568, ..., 0.6016, 0.2858, 0.4434])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -73,13 +73,13 @@ Rows: 10000
Size: 100000000
NNZ: 20000000
Density: 0.2
Time: 10.887785911560059 seconds
Time: 10.475984573364258 seconds
[40.74, 39.87, 40.58, 40.31, 40.56, 40.27, 40.45, 40.39, 39.9, 40.0]
[120.43]
16.833276748657227
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2251, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 10.887785911560059, 'TIME_S_1KI': 4.83686624236342, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2027.23151884079, 'W': 120.43}
[40.74, 39.87, 40.58, 40.31, 40.56, 40.27, 40.45, 40.39, 39.9, 40.0, 40.84, 39.74, 40.34, 40.47, 40.2, 40.2, 39.72, 39.73, 39.66, 39.68]
723.02
36.150999999999996
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2251, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 10.887785911560059, 'TIME_S_1KI': 4.83686624236342, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2027.23151884079, 'W': 120.43, 'J_1KI': 900.591523252239, 'W_1KI': 53.500666370501996, 'W_D': 84.27900000000001, 'J_D': 1418.6917311000825, 'W_D_1KI': 37.44069302532208, 'J_D_1KI': 16.632915604319006}
[40.41, 45.21, 40.37, 40.41, 40.26, 39.65, 40.47, 41.87, 39.92, 39.65]
[119.44]
17.294793367385864
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2210, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 10.475984573364258, 'TIME_S_1KI': 4.740264512834506, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2065.6901198005676, 'W': 119.44}
[40.41, 45.21, 40.37, 40.41, 40.26, 39.65, 40.47, 41.87, 39.92, 39.65, 40.39, 39.85, 39.69, 40.06, 40.15, 39.58, 40.59, 39.58, 39.95, 40.01]
727.8399999999999
36.391999999999996
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2210, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 20000000, 'MATRIX_DENSITY': 0.2, 'TIME_S': 10.475984573364258, 'TIME_S_1KI': 4.740264512834506, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2065.6901198005676, 'W': 119.44, 'J_1KI': 934.701411674465, 'W_1KI': 54.04524886877828, 'W_D': 83.048, 'J_D': 1436.2979995746614, 'W_D_1KI': 37.57828054298643, 'J_D_1KI': 17.00374685203006}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.3.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 1475, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 10.550647735595703, "TIME_S_1KI": 7.152981515658104, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2246.3171599555017, "W": 116.09, "J_1KI": 1522.9268881054247, "W_1KI": 78.7050847457627, "W_D": 80.02975, "J_D": 1548.5588830385805, "W_D_1KI": 54.25745762711865, "J_D_1KI": 36.78471703533467}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 1434, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 10.430570602416992, "TIME_S_1KI": 7.273759136971403, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2171.006755914688, "W": 115.95999999999998, "J_1KI": 1513.9517126322787, "W_1KI": 80.8647140864714, "W_D": 79.27299999999998, "J_D": 1484.1515915973182, "W_D_1KI": 55.281032078103195, "J_D_1KI": 38.55023157468842}

60
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_0.3.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '0.3', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 7.117977619171143}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '0.3', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 0.7320935726165771}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2996, 5947, ..., 29993941,
29997016, 30000000]),
col_indices=tensor([ 2, 4, 5, ..., 9994, 9995, 9997]),
values=tensor([0.7643, 0.7440, 0.4862, ..., 0.6436, 0.3641, 0.9418]),
tensor(crow_indices=tensor([ 0, 2976, 5887, ..., 29993981,
29996974, 30000000]),
col_indices=tensor([ 2, 12, 13, ..., 9995, 9997, 9999]),
values=tensor([0.2872, 0.6919, 0.0045, ..., 0.7234, 0.8152, 0.1470]),
size=(10000, 10000), nnz=30000000, layout=torch.sparse_csr)
tensor([0.8566, 0.8595, 0.2293, ..., 0.0057, 0.7338, 0.0583])
tensor([0.3759, 0.5048, 0.7452, ..., 0.9323, 0.0206, 0.6020])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -16,19 +16,19 @@ Rows: 10000
Size: 100000000
NNZ: 30000000
Density: 0.3
Time: 7.117977619171143 seconds
Time: 0.7320935726165771 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1475', '-ss', '10000', '-sd', '0.3', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 10.550647735595703}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1434', '-ss', '10000', '-sd', '0.3', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 30000000, "MATRIX_DENSITY": 0.3, "TIME_S": 10.430570602416992}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2975, 5983, ..., 29994011,
29997029, 30000000]),
col_indices=tensor([ 7, 9, 13, ..., 9994, 9995, 9998]),
values=tensor([0.9780, 0.9139, 0.8008, ..., 0.7340, 0.6785, 0.0713]),
tensor(crow_indices=tensor([ 0, 2946, 5856, ..., 29993956,
29997054, 30000000]),
col_indices=tensor([ 1, 3, 10, ..., 9992, 9994, 9995]),
values=tensor([0.6658, 0.8893, 0.2640, ..., 0.2436, 0.9944, 0.7745]),
size=(10000, 10000), nnz=30000000, layout=torch.sparse_csr)
tensor([0.9158, 0.5873, 0.3730, ..., 0.5358, 0.1305, 0.5051])
tensor([0.7478, 0.4417, 0.0487, ..., 0.7713, 0.8445, 0.5646])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -36,16 +36,16 @@ Rows: 10000
Size: 100000000
NNZ: 30000000
Density: 0.3
Time: 10.550647735595703 seconds
Time: 10.430570602416992 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 2975, 5983, ..., 29994011,
29997029, 30000000]),
col_indices=tensor([ 7, 9, 13, ..., 9994, 9995, 9998]),
values=tensor([0.9780, 0.9139, 0.8008, ..., 0.7340, 0.6785, 0.0713]),
tensor(crow_indices=tensor([ 0, 2946, 5856, ..., 29993956,
29997054, 30000000]),
col_indices=tensor([ 1, 3, 10, ..., 9992, 9994, 9995]),
values=tensor([0.6658, 0.8893, 0.2640, ..., 0.2436, 0.9944, 0.7745]),
size=(10000, 10000), nnz=30000000, layout=torch.sparse_csr)
tensor([0.9158, 0.5873, 0.3730, ..., 0.5358, 0.1305, 0.5051])
tensor([0.7478, 0.4417, 0.0487, ..., 0.7713, 0.8445, 0.5646])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -53,13 +53,13 @@ Rows: 10000
Size: 100000000
NNZ: 30000000
Density: 0.3
Time: 10.550647735595703 seconds
Time: 10.430570602416992 seconds
[41.39, 39.87, 39.86, 40.02, 39.88, 39.87, 40.22, 40.42, 40.2, 40.31]
[116.09]
19.349790334701538
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1475, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 10.550647735595703, 'TIME_S_1KI': 7.152981515658104, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2246.3171599555017, 'W': 116.09}
[41.39, 39.87, 39.86, 40.02, 39.88, 39.87, 40.22, 40.42, 40.2, 40.31, 41.51, 39.76, 39.79, 39.7, 39.79, 40.38, 40.29, 39.67, 39.96, 39.84]
721.2049999999999
36.060249999999996
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1475, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 10.550647735595703, 'TIME_S_1KI': 7.152981515658104, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2246.3171599555017, 'W': 116.09, 'J_1KI': 1522.9268881054247, 'W_1KI': 78.7050847457627, 'W_D': 80.02975, 'J_D': 1548.5588830385805, 'W_D_1KI': 54.25745762711865, 'J_D_1KI': 36.78471703533467}
[40.39, 40.27, 40.31, 40.01, 39.78, 40.12, 45.36, 39.94, 40.02, 52.21]
[115.96]
18.722031354904175
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1434, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 10.430570602416992, 'TIME_S_1KI': 7.273759136971403, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2171.006755914688, 'W': 115.95999999999998}
[40.39, 40.27, 40.31, 40.01, 39.78, 40.12, 45.36, 39.94, 40.02, 52.21, 41.63, 40.16, 40.0, 40.35, 41.93, 39.55, 39.78, 39.55, 39.63, 39.73]
733.74
36.687
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1434, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 30000000, 'MATRIX_DENSITY': 0.3, 'TIME_S': 10.430570602416992, 'TIME_S_1KI': 7.273759136971403, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2171.006755914688, 'W': 115.95999999999998, 'J_1KI': 1513.9517126322787, 'W_1KI': 80.8647140864714, 'W_D': 79.27299999999998, 'J_D': 1484.1515915973182, 'W_D_1KI': 55.281032078103195, 'J_D_1KI': 38.55023157468842}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 355068, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.454679489135742, "TIME_S_1KI": 0.029444161369472165, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1283.4049695920944, "W": 97.06, "J_1KI": 3.614532905224054, "W_1KI": 0.27335608953777873, "W_D": 61.48125, "J_D": 812.9542735084892, "W_D_1KI": 0.17315345229646154, "J_D_1KI": 0.0004876627921875853}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 349456, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 1000, "MATRIX_DENSITY": 1e-05, "TIME_S": 11.173964023590088, "TIME_S_1KI": 0.03197531026392475, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1252.9686733055114, "W": 97.63, "J_1KI": 3.585483360724988, "W_1KI": 0.2793770889611282, "W_D": 61.798, "J_D": 793.1061976127625, "W_D_1KI": 0.17684057506524428, "J_D_1KI": 0.0005060453249200021}

2623
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_1e-05.output
View File

File diff suppressed because it is too large Load Diff

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 307566, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.194913387298584, "TIME_S_1KI": 0.033147075383165185, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1254.457565395832, "W": 97.91, "J_1KI": 4.078661378032136, "W_1KI": 0.3183381778219959, "W_D": 62.19175, "J_D": 796.8227075141073, "W_D_1KI": 0.2022061931422849, "J_D_1KI": 0.0006574400068352317}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 308023, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.28289532661438, "TIME_S_1KI": 0.03338353086170311, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1226.4741826629638, "W": 98.08, "J_1KI": 3.9817616952726382, "W_1KI": 0.3184177804904179, "W_D": 61.967, "J_D": 774.8870888772011, "W_D_1KI": 0.20117653551845155, "J_D_1KI": 0.0006531217977828004}

68
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_10000_5e-05.output
View File

@ -1,13 +1,13 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '10000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.04550504684448242}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '10000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.019933462142944336}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 0, ..., 4998, 5000, 5000]),
col_indices=tensor([9281, 526, 5110, ..., 4172, 680, 4833]),
values=tensor([0.9710, 0.4177, 0.1273, ..., 0.7621, 0.2431, 0.8030]),
tensor(crow_indices=tensor([ 0, 2, 5, ..., 4999, 5000, 5000]),
col_indices=tensor([4080, 6557, 3158, ..., 4357, 6307, 2550]),
values=tensor([0.9910, 0.3414, 0.4855, ..., 0.2598, 0.6108, 0.2815]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.6244, 0.3231, 0.3638, ..., 0.2586, 0.1943, 0.4038])
tensor([0.2787, 0.7388, 0.8319, ..., 0.5413, 0.0496, 0.2437])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -15,18 +15,18 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 0.04550504684448242 seconds
Time: 0.019933462142944336 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '230743', '-ss', '10000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 7.877320289611816}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '52675', '-ss', '10000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 1.7956035137176514}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 1, 1, ..., 4996, 4999, 5000]),
col_indices=tensor([5149, 830, 3827, ..., 6947, 7825, 8143]),
values=tensor([0.7974, 0.8672, 0.6352, ..., 0.0945, 0.9729, 0.8206]),
tensor(crow_indices=tensor([ 0, 0, 0, ..., 5000, 5000, 5000]),
col_indices=tensor([ 408, 476, 3837, ..., 3097, 8388, 8856]),
values=tensor([0.3698, 0.9808, 0.6496, ..., 0.7839, 0.4021, 0.3346]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.8724, 0.1762, 0.3345, ..., 0.8958, 0.7321, 0.5036])
tensor([0.1775, 0.8809, 0.7204, ..., 0.4994, 0.6943, 0.3851])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -34,18 +34,18 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 7.877320289611816 seconds
Time: 1.7956035137176514 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '307566', '-ss', '10000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.194913387298584}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '308023', '-ss', '10000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [10000, 10000], "MATRIX_ROWS": 10000, "MATRIX_SIZE": 100000000, "MATRIX_NNZ": 5000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.28289532661438}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 1, ..., 5000, 5000, 5000]),
col_indices=tensor([6013, 2562, 3841, ..., 8262, 3270, 1424]),
values=tensor([0.2685, 0.3885, 0.6243, ..., 0.1198, 0.9466, 0.9233]),
tensor(crow_indices=tensor([ 0, 0, 2, ..., 4999, 5000, 5000]),
col_indices=tensor([2483, 6584, 3017, ..., 870, 3138, 2052]),
values=tensor([0.7385, 0.7043, 0.9061, ..., 0.4377, 0.8515, 0.3180]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.2690, 0.9597, 0.8597, ..., 0.6996, 0.9887, 0.3737])
tensor([0.1916, 0.9837, 0.2990, ..., 0.4110, 0.2807, 0.4933])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -53,15 +53,15 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 10.194913387298584 seconds
Time: 10.28289532661438 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 0, 1, ..., 5000, 5000, 5000]),
col_indices=tensor([6013, 2562, 3841, ..., 8262, 3270, 1424]),
values=tensor([0.2685, 0.3885, 0.6243, ..., 0.1198, 0.9466, 0.9233]),
tensor(crow_indices=tensor([ 0, 0, 2, ..., 4999, 5000, 5000]),
col_indices=tensor([2483, 6584, 3017, ..., 870, 3138, 2052]),
values=tensor([0.7385, 0.7043, 0.9061, ..., 0.4377, 0.8515, 0.3180]),
size=(10000, 10000), nnz=5000, layout=torch.sparse_csr)
tensor([0.2690, 0.9597, 0.8597, ..., 0.6996, 0.9887, 0.3737])
tensor([0.1916, 0.9837, 0.2990, ..., 0.4110, 0.2807, 0.4933])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([10000, 10000])
@ -69,13 +69,13 @@ Rows: 10000
Size: 100000000
NNZ: 5000
Density: 5e-05
Time: 10.194913387298584 seconds
Time: 10.28289532661438 seconds
[40.9, 39.57, 39.65, 39.48, 39.52, 39.37, 39.42, 39.88, 40.21, 39.84]
[97.91]
12.81235384941101
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 307566, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.194913387298584, 'TIME_S_1KI': 0.033147075383165185, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1254.457565395832, 'W': 97.91}
[40.9, 39.57, 39.65, 39.48, 39.52, 39.37, 39.42, 39.88, 40.21, 39.84, 41.61, 39.93, 39.26, 40.39, 39.17, 39.17, 39.39, 39.38, 39.35, 40.1]
714.365
35.71825
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 307566, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.194913387298584, 'TIME_S_1KI': 0.033147075383165185, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1254.457565395832, 'W': 97.91, 'J_1KI': 4.078661378032136, 'W_1KI': 0.3183381778219959, 'W_D': 62.19175, 'J_D': 796.8227075141073, 'W_D_1KI': 0.2022061931422849, 'J_D_1KI': 0.0006574400068352317}
[40.25, 39.88, 39.44, 44.89, 40.01, 39.55, 39.45, 39.48, 39.91, 41.29]
[98.08]
12.504834651947021
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 308023, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.28289532661438, 'TIME_S_1KI': 0.03338353086170311, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1226.4741826629638, 'W': 98.08}
[40.25, 39.88, 39.44, 44.89, 40.01, 39.55, 39.45, 39.48, 39.91, 41.29, 44.96, 39.78, 39.79, 40.18, 39.36, 39.46, 39.39, 39.37, 39.43, 39.28]
722.26
36.113
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 308023, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [10000, 10000], 'MATRIX_ROWS': 10000, 'MATRIX_SIZE': 100000000, 'MATRIX_NNZ': 5000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.28289532661438, 'TIME_S_1KI': 0.03338353086170311, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1226.4741826629638, 'W': 98.08, 'J_1KI': 3.9817616952726382, 'W_1KI': 0.3184177804904179, 'W_D': 61.967, 'J_D': 774.8870888772011, 'W_D_1KI': 0.20117653551845155, 'J_D_1KI': 0.0006531217977828004}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_500000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 1316, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.480877876281738, "TIME_S_1KI": 7.964192915107703, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2349.371359462738, "W": 119.72, "J_1KI": 1785.2365953364272, "W_1KI": 90.9726443768997, "W_D": 83.6515, "J_D": 1641.5673093559742, "W_D_1KI": 63.564969604863215, "J_D_1KI": 48.3016486359143}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 1275, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.370937824249268, "TIME_S_1KI": 8.134068881764131, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2317.954887828827, "W": 118.94, "J_1KI": 1818.0038335912368, "W_1KI": 93.28627450980392, "W_D": 82.976, "J_D": 1617.072681793213, "W_D_1KI": 65.07921568627451, "J_D_1KI": 51.04252210688197}

87
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_500000_0.0001.output
View File

@ -1,15 +1,15 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '500000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 8.476217031478882}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '500000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.8229217529296875}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 58, 110, ..., 24999904,
24999948, 25000000]),
col_indices=tensor([ 6107, 36475, 44542, ..., 455197, 482838,
484709]),
values=tensor([0.8741, 0.1087, 0.7265, ..., 0.9387, 0.2139, 0.8984]),
tensor(crow_indices=tensor([ 0, 48, 92, ..., 24999906,
24999954, 25000000]),
col_indices=tensor([ 13687, 16103, 22085, ..., 466250, 497468,
498839]),
values=tensor([0.1763, 0.0612, 0.1831, ..., 0.7206, 0.9735, 0.4201]),
size=(500000, 500000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.2008, 0.1464, 0.5363, ..., 0.5258, 0.1478, 0.0153])
tensor([0.0392, 0.3068, 0.8540, ..., 0.0771, 0.2433, 0.8939])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -17,20 +17,20 @@ Rows: 500000
Size: 250000000000
NNZ: 25000000
Density: 0.0001
Time: 8.476217031478882 seconds
Time: 0.8229217529296875 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1238', '-ss', '500000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 9.872223138809204}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1275', '-ss', '500000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.370937824249268}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 42, 103, ..., 24999895,
24999951, 25000000]),
col_indices=tensor([ 1782, 32597, 35292, ..., 490788, 494408,
495086]),
values=tensor([0.7532, 0.4055, 0.7849, ..., 0.0826, 0.2837, 0.9366]),
tensor(crow_indices=tensor([ 0, 45, 89, ..., 24999893,
24999957, 25000000]),
col_indices=tensor([ 25264, 35882, 38786, ..., 487781, 491680,
492236]),
values=tensor([0.0901, 0.4292, 0.0295, ..., 0.7641, 0.5758, 0.3435]),
size=(500000, 500000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.6389, 0.8135, 0.6286, ..., 0.0387, 0.4513, 0.2151])
tensor([0.7878, 0.6485, 0.9023, ..., 0.5055, 0.2764, 0.4227])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -38,20 +38,17 @@ Rows: 500000
Size: 250000000000
NNZ: 25000000
Density: 0.0001
Time: 9.872223138809204 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1316', '-ss', '500000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 25000000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.480877876281738}
Time: 10.370937824249268 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 49, 93, ..., 24999907,
24999956, 25000000]),
col_indices=tensor([ 1315, 9605, 36829, ..., 467295, 483577,
490282]),
values=tensor([0.4504, 0.6377, 0.8962, ..., 0.8735, 0.2973, 0.1824]),
tensor(crow_indices=tensor([ 0, 45, 89, ..., 24999893,
24999957, 25000000]),
col_indices=tensor([ 25264, 35882, 38786, ..., 487781, 491680,
492236]),
values=tensor([0.0901, 0.4292, 0.0295, ..., 0.7641, 0.5758, 0.3435]),
size=(500000, 500000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.1673, 0.3950, 0.6065, ..., 0.2020, 0.8580, 0.9739])
tensor([0.7878, 0.6485, 0.9023, ..., 0.5055, 0.2764, 0.4227])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -59,31 +56,13 @@ Rows: 500000
Size: 250000000000
NNZ: 25000000
Density: 0.0001
Time: 10.480877876281738 seconds
Time: 10.370937824249268 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 49, 93, ..., 24999907,
24999956, 25000000]),
col_indices=tensor([ 1315, 9605, 36829, ..., 467295, 483577,
490282]),
values=tensor([0.4504, 0.6377, 0.8962, ..., 0.8735, 0.2973, 0.1824]),
size=(500000, 500000), nnz=25000000, layout=torch.sparse_csr)
tensor([0.1673, 0.3950, 0.6065, ..., 0.2020, 0.8580, 0.9739])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
Rows: 500000
Size: 250000000000
NNZ: 25000000
Density: 0.0001
Time: 10.480877876281738 seconds
[40.78, 40.23, 40.39, 40.14, 40.33, 40.17, 39.83, 39.85, 39.79, 39.68]
[119.72]
19.623883724212646
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1316, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.480877876281738, 'TIME_S_1KI': 7.964192915107703, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2349.371359462738, 'W': 119.72}
[40.78, 40.23, 40.39, 40.14, 40.33, 40.17, 39.83, 39.85, 39.79, 39.68, 41.19, 39.75, 40.31, 41.27, 39.67, 39.76, 39.72, 39.58, 39.67, 40.17]
721.3699999999999
36.06849999999999
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1316, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.480877876281738, 'TIME_S_1KI': 7.964192915107703, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2349.371359462738, 'W': 119.72, 'J_1KI': 1785.2365953364272, 'W_1KI': 90.9726443768997, 'W_D': 83.6515, 'J_D': 1641.5673093559742, 'W_D_1KI': 63.564969604863215, 'J_D_1KI': 48.3016486359143}
[40.75, 40.02, 39.59, 39.62, 39.74, 39.84, 39.97, 39.74, 40.73, 39.5]
[118.94]
19.488438606262207
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1275, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.370937824249268, 'TIME_S_1KI': 8.134068881764131, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2317.954887828827, 'W': 118.94}
[40.75, 40.02, 39.59, 39.62, 39.74, 39.84, 39.97, 39.74, 40.73, 39.5, 41.45, 40.23, 40.12, 39.57, 39.77, 39.94, 39.75, 39.69, 40.29, 39.64]
719.28
35.964
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 1275, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 25000000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.370937824249268, 'TIME_S_1KI': 8.134068881764131, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2317.954887828827, 'W': 118.94, 'J_1KI': 1818.0038335912368, 'W_1KI': 93.28627450980392, 'W_D': 82.976, 'J_D': 1617.072681793213, 'W_D_1KI': 65.07921568627451, 'J_D_1KI': 51.04252210688197}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_500000_1e-05.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 21497, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.492619514465332, "TIME_S_1KI": 0.488096921173435, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2075.744820713997, "W": 155.25, "J_1KI": 96.55974418356035, "W_1KI": 7.221937944829511, "W_D": 119.6075, "J_D": 1599.1925838553905, "W_D_1KI": 5.5639158952411965, "J_D_1KI": 0.25882290064851826}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 20602, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.051029920578003, "TIME_S_1KI": 0.4878667081146492, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2012.3513662075998, "W": 152.11, "J_1KI": 97.67747627451702, "W_1KI": 7.383263760799923, "W_D": 115.75800000000001, "J_D": 1531.429685421467, "W_D_1KI": 5.61877487622561, "J_D_1KI": 0.272729583352374}

106
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_500000_1e-05.output
View File

@ -1,15 +1,15 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '500000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.5443899631500244}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '500000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 0.08110809326171875}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 7, 13, ..., 2499984,
2499993, 2500000]),
col_indices=tensor([ 29642, 73796, 205405, ..., 362365, 387524,
440531]),
values=tensor([0.6565, 0.4150, 0.8341, ..., 0.7997, 0.8212, 0.8706]),
tensor(crow_indices=tensor([ 0, 6, 8, ..., 2499988,
2499994, 2500000]),
col_indices=tensor([ 61750, 191731, 192878, ..., 292292, 347392,
413452]),
values=tensor([0.4333, 0.7749, 0.6975, ..., 0.5571, 0.2303, 0.6423]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.3188, 0.4041, 0.2486, ..., 0.5189, 0.6175, 0.2446])
tensor([0.7573, 0.7811, 0.2609, ..., 0.7028, 0.0683, 0.1077])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -17,20 +17,41 @@ Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 0.5443899631500244 seconds
Time: 0.08110809326171875 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '19287', '-ss', '500000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 9.420541286468506}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '12945', '-ss', '500000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 6.597289562225342}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 4, 12, ..., 2499994,
tensor(crow_indices=tensor([ 0, 4, 9, ..., 2499996,
2499997, 2500000]),
col_indices=tensor([304373, 374974, 396567, ..., 161828, 243938,
306700]),
values=tensor([0.0234, 0.0111, 0.7752, ..., 0.4123, 0.0911, 0.7333]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.2563, 0.2400, 0.1997, ..., 0.9331, 0.1838, 0.9541])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 6.597289562225342 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '20602', '-ss', '500000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.051029920578003}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 4, 7, ..., 2499995,
2499998, 2500000]),
col_indices=tensor([131466, 192610, 285983, ..., 398857, 7127,
216070]),
values=tensor([0.3766, 0.1095, 0.0818, ..., 0.7673, 0.9998, 0.7256]),
col_indices=tensor([ 84683, 221772, 250792, ..., 457280, 123381,
490345]),
values=tensor([0.6671, 0.6498, 0.8275, ..., 0.5282, 0.6912, 0.3058]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.6672, 0.9862, 0.6354, ..., 0.4943, 0.9100, 0.2548])
tensor([0.8099, 0.6830, 0.6662, ..., 0.4435, 0.6731, 0.4595])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -38,20 +59,17 @@ Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 9.420541286468506 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '21497', '-ss', '500000', '-sd', '1e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 1e-05, "TIME_S": 10.492619514465332}
Time: 10.051029920578003 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 10, 12, ..., 2499992,
2499996, 2500000]),
col_indices=tensor([124091, 157764, 160136, ..., 120950, 171105,
490445]),
values=tensor([0.1739, 0.8424, 0.0842, ..., 0.2028, 0.9911, 0.7243]),
tensor(crow_indices=tensor([ 0, 4, 7, ..., 2499995,
2499998, 2500000]),
col_indices=tensor([ 84683, 221772, 250792, ..., 457280, 123381,
490345]),
values=tensor([0.6671, 0.6498, 0.8275, ..., 0.5282, 0.6912, 0.3058]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.9645, 0.6044, 0.9036, ..., 0.9779, 0.7664, 0.8298])
tensor([0.8099, 0.6830, 0.6662, ..., 0.4435, 0.6731, 0.4595])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -59,31 +77,13 @@ Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 10.492619514465332 seconds
Time: 10.051029920578003 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 10, 12, ..., 2499992,
2499996, 2500000]),
col_indices=tensor([124091, 157764, 160136, ..., 120950, 171105,
490445]),
values=tensor([0.1739, 0.8424, 0.0842, ..., 0.2028, 0.9911, 0.7243]),
size=(500000, 500000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.9645, 0.6044, 0.9036, ..., 0.9779, 0.7664, 0.8298])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
Rows: 500000
Size: 250000000000
NNZ: 2500000
Density: 1e-05
Time: 10.492619514465332 seconds
[41.35, 39.34, 39.96, 39.31, 39.6, 39.93, 39.84, 39.39, 39.3, 39.31]
[155.25]
13.370337009429932
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 21497, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.492619514465332, 'TIME_S_1KI': 0.488096921173435, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2075.744820713997, 'W': 155.25}
[41.35, 39.34, 39.96, 39.31, 39.6, 39.93, 39.84, 39.39, 39.3, 39.31, 40.05, 40.08, 39.33, 39.83, 39.32, 39.88, 39.28, 39.25, 39.29, 39.13]
712.85
35.6425
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 21497, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.492619514465332, 'TIME_S_1KI': 0.488096921173435, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2075.744820713997, 'W': 155.25, 'J_1KI': 96.55974418356035, 'W_1KI': 7.221937944829511, 'W_D': 119.6075, 'J_D': 1599.1925838553905, 'W_D_1KI': 5.5639158952411965, 'J_D_1KI': 0.25882290064851826}
[40.55, 45.24, 39.91, 39.85, 39.84, 39.77, 40.43, 40.43, 42.0, 39.73]
[152.11]
13.22957968711853
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 20602, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.051029920578003, 'TIME_S_1KI': 0.4878667081146492, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2012.3513662075998, 'W': 152.11}
[40.55, 45.24, 39.91, 39.85, 39.84, 39.77, 40.43, 40.43, 42.0, 39.73, 40.67, 40.21, 39.79, 39.83, 39.69, 40.04, 39.86, 39.69, 40.08, 39.81]
727.04
36.352
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 20602, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 1e-05, 'TIME_S': 10.051029920578003, 'TIME_S_1KI': 0.4878667081146492, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2012.3513662075998, 'W': 152.11, 'J_1KI': 97.67747627451702, 'W_1KI': 7.383263760799923, 'W_D': 115.75800000000001, 'J_D': 1531.429685421467, 'W_D_1KI': 5.61877487622561, 'J_D_1KI': 0.272729583352374}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_500000_5e-05.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 2443, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 11.091211557388306, "TIME_S_1KI": 4.5399965441622205, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2093.1614422798157, "W": 125.74, "J_1KI": 856.7996079737272, "W_1KI": 51.469504707327054, "W_D": 89.3905, "J_D": 1488.0606641173363, "W_D_1KI": 36.59046254604994, "J_D_1KI": 14.977676031948402}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 2268, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.338330507278442, "TIME_S_1KI": 4.558346784514304, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1929.4526886749268, "W": 124.72, "J_1KI": 850.7286987102852, "W_1KI": 54.99118165784832, "W_D": 88.6155, "J_D": 1370.9061476368904, "W_D_1KI": 39.07208994708994, "J_D_1KI": 17.227552886723963}

87
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_500000_5e-05.output
View File

@ -1,15 +1,15 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '500000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 4.620434761047363}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '500000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 0.46280455589294434}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 28, 55, ..., 12499942,
12499972, 12500000]),
col_indices=tensor([ 19855, 24177, 33309, ..., 430292, 468270,
470726]),
values=tensor([0.1735, 0.2720, 0.9086, ..., 0.2697, 0.0473, 0.0416]),
tensor(crow_indices=tensor([ 0, 21, 39, ..., 12499960,
12499981, 12500000]),
col_indices=tensor([ 5530, 18658, 36900, ..., 388989, 426254,
497258]),
values=tensor([0.8053, 0.3880, 0.4779, ..., 0.4773, 0.4279, 0.6817]),
size=(500000, 500000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.2844, 0.4487, 0.9137, ..., 0.5004, 0.3000, 0.1233])
tensor([0.5886, 0.4606, 0.7255, ..., 0.1606, 0.2608, 0.5232])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -17,20 +17,20 @@ Rows: 500000
Size: 250000000000
NNZ: 12500000
Density: 5e-05
Time: 4.620434761047363 seconds
Time: 0.46280455589294434 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '2272', '-ss', '500000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 9.761992931365967}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '2268', '-ss', '500000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 10.338330507278442}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 25, 50, ..., 12499939,
12499968, 12500000]),
col_indices=tensor([ 35309, 102593, 109410, ..., 438712, 452154,
489935]),
values=tensor([0.4991, 0.7582, 0.4985, ..., 0.8355, 0.6986, 0.3665]),
tensor(crow_indices=tensor([ 0, 30, 63, ..., 12499957,
12499979, 12500000]),
col_indices=tensor([ 14790, 16334, 55074, ..., 466420, 486794,
499923]),
values=tensor([0.8543, 0.1686, 0.8292, ..., 0.6567, 0.2357, 0.6950]),
size=(500000, 500000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.1755, 0.5499, 0.0031, ..., 0.2944, 0.6143, 0.3232])
tensor([0.8639, 0.3423, 0.4800, ..., 0.1443, 0.7816, 0.0060])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -38,20 +38,17 @@ Rows: 500000
Size: 250000000000
NNZ: 12500000
Density: 5e-05
Time: 9.761992931365967 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '2443', '-ss', '500000', '-sd', '5e-05', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [500000, 500000], "MATRIX_ROWS": 500000, "MATRIX_SIZE": 250000000000, "MATRIX_NNZ": 12500000, "MATRIX_DENSITY": 5e-05, "TIME_S": 11.091211557388306}
Time: 10.338330507278442 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 29, 54, ..., 12499940,
12499966, 12500000]),
col_indices=tensor([ 15104, 53699, 66016, ..., 451008, 478949,
498027]),
values=tensor([0.3834, 0.0166, 0.8269, ..., 0.7083, 0.6634, 0.5753]),
tensor(crow_indices=tensor([ 0, 30, 63, ..., 12499957,
12499979, 12500000]),
col_indices=tensor([ 14790, 16334, 55074, ..., 466420, 486794,
499923]),
values=tensor([0.8543, 0.1686, 0.8292, ..., 0.6567, 0.2357, 0.6950]),
size=(500000, 500000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.3028, 0.2567, 0.4384, ..., 0.4923, 0.2329, 0.9462])
tensor([0.8639, 0.3423, 0.4800, ..., 0.1443, 0.7816, 0.0060])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
@ -59,31 +56,13 @@ Rows: 500000
Size: 250000000000
NNZ: 12500000
Density: 5e-05
Time: 11.091211557388306 seconds
Time: 10.338330507278442 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 29, 54, ..., 12499940,
12499966, 12500000]),
col_indices=tensor([ 15104, 53699, 66016, ..., 451008, 478949,
498027]),
values=tensor([0.3834, 0.0166, 0.8269, ..., 0.7083, 0.6634, 0.5753]),
size=(500000, 500000), nnz=12500000, layout=torch.sparse_csr)
tensor([0.3028, 0.2567, 0.4384, ..., 0.4923, 0.2329, 0.9462])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([500000, 500000])
Rows: 500000
Size: 250000000000
NNZ: 12500000
Density: 5e-05
Time: 11.091211557388306 seconds
[41.47, 41.03, 39.87, 40.83, 39.9, 40.09, 40.23, 40.2, 39.74, 39.65]
[125.74]
16.646742820739746
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2443, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 11.091211557388306, 'TIME_S_1KI': 4.5399965441622205, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2093.1614422798157, 'W': 125.74}
[41.47, 41.03, 39.87, 40.83, 39.9, 40.09, 40.23, 40.2, 39.74, 39.65, 40.36, 39.88, 39.75, 39.86, 39.72, 40.13, 40.1, 39.56, 45.29, 40.14]
726.99
36.3495
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2443, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 11.091211557388306, 'TIME_S_1KI': 4.5399965441622205, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2093.1614422798157, 'W': 125.74, 'J_1KI': 856.7996079737272, 'W_1KI': 51.469504707327054, 'W_D': 89.3905, 'J_D': 1488.0606641173363, 'W_D_1KI': 36.59046254604994, 'J_D_1KI': 14.977676031948402}
[40.4, 42.56, 40.55, 40.15, 39.72, 40.38, 39.75, 39.66, 39.85, 39.57]
[124.72]
15.470274925231934
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2268, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.338330507278442, 'TIME_S_1KI': 4.558346784514304, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1929.4526886749268, 'W': 124.72}
[40.4, 42.56, 40.55, 40.15, 39.72, 40.38, 39.75, 39.66, 39.85, 39.57, 40.6, 40.35, 40.27, 40.13, 39.75, 39.68, 39.73, 39.6, 39.89, 39.57]
722.09
36.1045
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 2268, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [500000, 500000], 'MATRIX_ROWS': 500000, 'MATRIX_SIZE': 250000000000, 'MATRIX_NNZ': 12500000, 'MATRIX_DENSITY': 5e-05, 'TIME_S': 10.338330507278442, 'TIME_S_1KI': 4.558346784514304, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1929.4526886749268, 'W': 124.72, 'J_1KI': 850.7286987102852, 'W_1KI': 54.99118165784832, 'W_D': 88.6155, 'J_D': 1370.9061476368904, 'W_D_1KI': 39.07208994708994, 'J_D_1KI': 17.227552886723963}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_50000_0.0001.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 91834, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.82576298713684, "TIME_S_1KI": 0.11788404062914433, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1647.9860616016388, "W": 116.97, "J_1KI": 17.945271485524305, "W_1KI": 1.273711261624235, "W_D": 80.83175, "J_D": 1138.8355760867596, "W_D_1KI": 0.8801941546703835, "J_D_1KI": 0.009584621759592129}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 89538, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.289806604385376, "TIME_S_1KI": 0.11492111287258344, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1521.4007213592529, "W": 116.96, "J_1KI": 16.991676398392332, "W_1KI": 1.3062610288369183, "W_D": 80.71074999999999, "J_D": 1049.8751134699583, "W_D_1KI": 0.9014133663919228, "J_D_1KI": 0.010067383305322017}

102
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_50000_0.0001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.14647722244262695}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 0.049301862716674805}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 8, ..., 249988, 249995,
tensor(crow_indices=tensor([ 0, 5, 13, ..., 249993, 249996,
250000]),
col_indices=tensor([ 544, 6056, 19594, ..., 16208, 31107, 37035]),
values=tensor([0.8576, 0.5005, 0.2810, ..., 0.0063, 0.7171, 0.8258]),
col_indices=tensor([ 1709, 19790, 28830, ..., 3831, 22257, 48856]),
values=tensor([0.9244, 0.7522, 0.6687, ..., 0.7540, 0.7318, 0.7260]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.4318, 0.7107, 0.2576, ..., 0.8496, 0.3705, 0.3608])
tensor([0.0785, 0.8938, 0.5541, ..., 0.5935, 0.2052, 0.2232])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -16,19 +16,19 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 0.14647722244262695 seconds
Time: 0.049301862716674805 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '71683', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 9.233005046844482}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '21297', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 2.7824935913085938}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 7, 10, ..., 249988, 249994,
tensor(crow_indices=tensor([ 0, 4, 8, ..., 249989, 249993,
250000]),
col_indices=tensor([ 4979, 12449, 23825, ..., 32585, 40358, 48594]),
values=tensor([0.7825, 0.8569, 0.5029, ..., 0.3250, 0.4106, 0.3303]),
col_indices=tensor([16415, 16632, 32449, ..., 45169, 45288, 48610]),
values=tensor([0.0101, 0.6954, 0.6241, ..., 0.3711, 0.7246, 0.3748]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.8033, 0.4755, 0.5204, ..., 0.8611, 0.9528, 0.0172])
tensor([0.6515, 0.7514, 0.0204, ..., 0.8861, 0.6124, 0.4798])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -36,19 +36,19 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 9.233005046844482 seconds
Time: 2.7824935913085938 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '81519', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 9.805182695388794}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '80366', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 9.42440915107727}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 6, 11, ..., 249983, 249992,
tensor(crow_indices=tensor([ 0, 7, 10, ..., 249987, 249993,
250000]),
col_indices=tensor([ 7422, 17911, 31055, ..., 30707, 32021, 38558]),
values=tensor([0.7718, 0.8036, 0.8293, ..., 0.2159, 0.0251, 0.0647]),
col_indices=tensor([ 2445, 24855, 26173, ..., 23560, 26333, 46130]),
values=tensor([0.2012, 0.2713, 0.8391, ..., 0.5844, 0.7972, 0.4463]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.3183, 0.3041, 0.1046, ..., 0.2603, 0.8118, 0.2097])
tensor([0.5580, 0.1767, 0.6905, ..., 0.9860, 0.6709, 0.2165])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -56,19 +56,19 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 9.805182695388794 seconds
Time: 9.42440915107727 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '87295', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 9.980920553207397}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '89538', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.289806604385376}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 3, 5, ..., 249989, 249993,
tensor(crow_indices=tensor([ 0, 2, 7, ..., 249988, 249996,
250000]),
col_indices=tensor([19530, 21432, 40127, ..., 33319, 45642, 48654]),
values=tensor([0.8438, 0.0330, 0.2387, ..., 0.6115, 0.5796, 0.5067]),
col_indices=tensor([ 2244, 34732, 7243, ..., 9132, 13610, 19520]),
values=tensor([0.6983, 0.0446, 0.9216, ..., 0.0232, 0.0374, 0.6300]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.1992, 0.5617, 0.3460, ..., 0.4818, 0.9372, 0.6597])
tensor([0.8539, 0.6321, 0.4259, ..., 0.2899, 0.6274, 0.3350])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -76,19 +76,16 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 9.980920553207397 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '91834', '-ss', '50000', '-sd', '0.0001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 250000, "MATRIX_DENSITY": 0.0001, "TIME_S": 10.82576298713684}
Time: 10.289806604385376 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 7, 12, ..., 249987, 249995,
tensor(crow_indices=tensor([ 0, 2, 7, ..., 249988, 249996,
250000]),
col_indices=tensor([ 2714, 5631, 18387, ..., 39061, 48792, 49070]),
values=tensor([0.1970, 0.9435, 0.9859, ..., 0.7944, 0.6863, 0.0587]),
col_indices=tensor([ 2244, 34732, 7243, ..., 9132, 13610, 19520]),
values=tensor([0.6983, 0.0446, 0.9216, ..., 0.0232, 0.0374, 0.6300]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.5128, 0.8861, 0.8900, ..., 0.3721, 0.4809, 0.7353])
tensor([0.8539, 0.6321, 0.4259, ..., 0.2899, 0.6274, 0.3350])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -96,30 +93,13 @@ Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 10.82576298713684 seconds
Time: 10.289806604385376 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 7, 12, ..., 249987, 249995,
250000]),
col_indices=tensor([ 2714, 5631, 18387, ..., 39061, 48792, 49070]),
values=tensor([0.1970, 0.9435, 0.9859, ..., 0.7944, 0.6863, 0.0587]),
size=(50000, 50000), nnz=250000, layout=torch.sparse_csr)
tensor([0.5128, 0.8861, 0.8900, ..., 0.3721, 0.4809, 0.7353])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
Rows: 50000
Size: 2500000000
NNZ: 250000
Density: 0.0001
Time: 10.82576298713684 seconds
[40.37, 39.3, 39.3, 39.19, 39.7, 39.15, 40.14, 44.33, 39.43, 39.81]
[116.97]
14.088963508605957
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 91834, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.82576298713684, 'TIME_S_1KI': 0.11788404062914433, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1647.9860616016388, 'W': 116.97}
[40.37, 39.3, 39.3, 39.19, 39.7, 39.15, 40.14, 44.33, 39.43, 39.81, 40.26, 39.09, 41.97, 44.11, 39.87, 39.01, 40.55, 38.99, 38.97, 38.89]
722.7650000000001
36.138250000000006
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 91834, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.82576298713684, 'TIME_S_1KI': 0.11788404062914433, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1647.9860616016388, 'W': 116.97, 'J_1KI': 17.945271485524305, 'W_1KI': 1.273711261624235, 'W_D': 80.83175, 'J_D': 1138.8355760867596, 'W_D_1KI': 0.8801941546703835, 'J_D_1KI': 0.009584621759592129}
[40.27, 39.54, 39.5, 39.65, 39.93, 39.44, 40.01, 39.51, 45.05, 39.37]
[116.96]
13.007872104644775
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 89538, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.289806604385376, 'TIME_S_1KI': 0.11492111287258344, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1521.4007213592529, 'W': 116.96}
[40.27, 39.54, 39.5, 39.65, 39.93, 39.44, 40.01, 39.51, 45.05, 39.37, 40.45, 39.41, 39.86, 39.36, 39.91, 44.54, 39.53, 39.97, 39.84, 39.78]
724.985
36.24925
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 89538, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 250000, 'MATRIX_DENSITY': 0.0001, 'TIME_S': 10.289806604385376, 'TIME_S_1KI': 0.11492111287258344, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1521.4007213592529, 'W': 116.96, 'J_1KI': 16.991676398392332, 'W_1KI': 1.3062610288369183, 'W_D': 80.71074999999999, 'J_D': 1049.8751134699583, 'W_D_1KI': 0.9014133663919228, 'J_D_1KI': 0.010067383305322017}

2
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_50000_0.001.json
View File

@ -1 +1 @@
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 46775, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.762548208236694, "TIME_S_1KI": 0.2300918911434889, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 2033.0924378275872, "W": 149.27, "J_1KI": 43.46536478519695, "W_1KI": 3.1912346338856232, "W_D": 113.87475, "J_D": 1551.0008245763183, "W_D_1KI": 2.434521646178514, "J_D_1KI": 0.052047496444222636}
{"CPU": "Epyc 7313P", "CORES": 16, "ITERATIONS": 45908, "MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.364075660705566, "TIME_S_1KI": 0.22575750763931268, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "J": 1982.6904290771486, "W": 146.58, "J_1KI": 43.18834253457238, "W_1KI": 3.192907554238913, "W_D": 110.64150000000001, "J_D": 1496.5741786651613, "W_D_1KI": 2.410070140280561, "J_D_1KI": 0.05249782478610615}

76
pytorch/output_synthetic_16core/epyc_7313p_16_csr_10_10_10_synthetic_50000_0.001.output
View File

@ -1,14 +1,14 @@
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '1000', '-ss', '50000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.29659008979797363}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '100', '-ss', '50000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 0.0605926513671875}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 46, 83, ..., 2499888,
2499946, 2500000]),
col_indices=tensor([ 2168, 2264, 3614, ..., 46868, 47216, 48811]),
values=tensor([0.2788, 0.0512, 0.3475, ..., 0.9281, 0.1898, 0.0144]),
tensor(crow_indices=tensor([ 0, 42, 99, ..., 2499902,
2499955, 2500000]),
col_indices=tensor([ 1009, 1628, 5292, ..., 43455, 47256, 47946]),
values=tensor([0.2339, 0.7843, 0.8407, ..., 0.0388, 0.2390, 0.6904]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.5080, 0.1629, 0.0847, ..., 0.6599, 0.4582, 0.2341])
tensor([0.3494, 0.3893, 0.8826, ..., 0.0693, 0.0070, 0.7582])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -16,19 +16,19 @@ Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 0.29659008979797363 seconds
Time: 0.0605926513671875 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '35402', '-ss', '50000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 7.946921348571777}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '17328', '-ss', '50000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 3.963207721710205}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 54, 117, ..., 2499905,
2499953, 2500000]),
col_indices=tensor([ 1300, 1442, 2491, ..., 47415, 49147, 49910]),
values=tensor([0.1149, 0.9707, 0.0968, ..., 0.7933, 0.6392, 0.9343]),
tensor(crow_indices=tensor([ 0, 45, 85, ..., 2499905,
2499952, 2500000]),
col_indices=tensor([ 2138, 2192, 2629, ..., 48532, 49646, 49876]),
values=tensor([0.7824, 0.0061, 0.7967, ..., 0.1635, 0.4732, 0.5157]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.2903, 0.7408, 0.0968, ..., 0.3344, 0.5691, 0.3821])
tensor([0.8165, 0.7580, 0.0903, ..., 0.6290, 0.7559, 0.6116])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -36,19 +36,19 @@ Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 7.946921348571777 seconds
Time: 3.963207721710205 seconds
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '46775', '-ss', '50000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.762548208236694}
['apptainer', 'run', '--env', 'OMP_PROC_BIND=true', '--env', 'OMP_PLACES={0:16}', 'pytorch-epyc_7313p.sif', 'python3', 'spmv.py', 'synthetic', 'csr', '45908', '-ss', '50000', '-sd', '0.001', '-c', '16']
{"MATRIX_TYPE": "synthetic", "MATRIX_FORMAT": "csr", "MATRIX_SHAPE": [50000, 50000], "MATRIX_ROWS": 50000, "MATRIX_SIZE": 2500000000, "MATRIX_NNZ": 2500000, "MATRIX_DENSITY": 0.001, "TIME_S": 10.364075660705566}
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 65, 117, ..., 2499903,
2499954, 2500000]),
col_indices=tensor([ 2232, 2981, 3015, ..., 49447, 49836, 49877]),
values=tensor([0.3281, 0.9452, 0.1004, ..., 0.4282, 0.2346, 0.1167]),
tensor(crow_indices=tensor([ 0, 48, 99, ..., 2499901,
2499955, 2500000]),
col_indices=tensor([ 2242, 2630, 4307, ..., 47333, 48170, 49131]),
values=tensor([0.3970, 0.2919, 0.1690, ..., 0.5693, 0.6652, 0.4283]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.5238, 0.6081, 0.9163, ..., 0.2866, 0.2457, 0.9117])
tensor([0.7545, 0.7866, 0.4331, ..., 0.1722, 0.5406, 0.9467])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -56,16 +56,16 @@ Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 10.762548208236694 seconds
Time: 10.364075660705566 seconds
/nfshomes/vut/ampere_research/pytorch/spmv.py:75: UserWarning: Sparse CSR tensor support is in beta state. If you miss a functionality in the sparse tensor support, please submit a feature request to https://github.com/pytorch/pytorch/issues. (Triggered internally at ../aten/src/ATen/SparseCsrTensorImpl.cpp:53.)
matrix = matrix.to_sparse_csr().type(torch.float32)
tensor(crow_indices=tensor([ 0, 65, 117, ..., 2499903,
2499954, 2500000]),
col_indices=tensor([ 2232, 2981, 3015, ..., 49447, 49836, 49877]),
values=tensor([0.3281, 0.9452, 0.1004, ..., 0.4282, 0.2346, 0.1167]),
tensor(crow_indices=tensor([ 0, 48, 99, ..., 2499901,
2499955, 2500000]),
col_indices=tensor([ 2242, 2630, 4307, ..., 47333, 48170, 49131]),
values=tensor([0.3970, 0.2919, 0.1690, ..., 0.5693, 0.6652, 0.4283]),
size=(50000, 50000), nnz=2500000, layout=torch.sparse_csr)
tensor([0.5238, 0.6081, 0.9163, ..., 0.2866, 0.2457, 0.9117])
tensor([0.7545, 0.7866, 0.4331, ..., 0.1722, 0.5406, 0.9467])
Matrix Type: synthetic
Matrix Format: csr
Shape: torch.Size([50000, 50000])
@ -73,13 +73,13 @@ Rows: 50000
Size: 2500000000
NNZ: 2500000
Density: 0.001
Time: 10.762548208236694 seconds
Time: 10.364075660705566 seconds
[41.04, 39.15, 39.26, 39.11, 39.17, 39.29, 39.32, 39.24, 39.56, 39.01]
[149.27]
13.620234727859497
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 46775, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.762548208236694, 'TIME_S_1KI': 0.2300918911434889, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2033.0924378275872, 'W': 149.27}
[41.04, 39.15, 39.26, 39.11, 39.17, 39.29, 39.32, 39.24, 39.56, 39.01, 39.97, 39.18, 39.31, 39.08, 39.46, 39.32, 39.13, 39.45, 39.15, 39.43]
707.905
35.39525
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 46775, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.762548208236694, 'TIME_S_1KI': 0.2300918911434889, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 2033.0924378275872, 'W': 149.27, 'J_1KI': 43.46536478519695, 'W_1KI': 3.1912346338856232, 'W_D': 113.87475, 'J_D': 1551.0008245763183, 'W_D_1KI': 2.434521646178514, 'J_D_1KI': 0.052047496444222636}
[40.59, 39.91, 39.52, 39.54, 39.7, 39.41, 39.47, 40.95, 40.05, 39.39]
[146.58]
13.526336669921875
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 45908, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.364075660705566, 'TIME_S_1KI': 0.22575750763931268, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1982.6904290771486, 'W': 146.58}
[40.59, 39.91, 39.52, 39.54, 39.7, 39.41, 39.47, 40.95, 40.05, 39.39, 40.78, 39.79, 39.69, 41.93, 40.24, 39.57, 39.58, 39.7, 39.52, 39.64]
718.77
35.9385
{'CPU': 'Epyc 7313P', 'CORES': 16, 'ITERATIONS': 45908, 'MATRIX_TYPE': 'synthetic', 'MATRIX_FORMAT': 'csr', 'MATRIX_SHAPE': [50000, 50000], 'MATRIX_ROWS': 50000, 'MATRIX_SIZE': 2500000000, 'MATRIX_NNZ': 2500000, 'MATRIX_DENSITY': 0.001, 'TIME_S': 10.364075660705566, 'TIME_S_1KI': 0.22575750763931268, 'BASELINE_TIME_S': 10, 'BASELINE_DELAY_S': 10, 'J': 1982.6904290771486, 'W': 146.58, 'J_1KI': 43.18834253457238, 'W_1KI': 3.192907554238913, 'W_D': 110.64150000000001, 'J_D': 1496.5741786651613, 'W_D_1KI': 2.410070140280561, 'J_D_1KI': 0.05249782478610615}

Some files were not shown because too many files have changed in this diff Show More