From d4ef9f434657161c9dadf75233a56aebf27b2154 Mon Sep 17 00:00:00 2001 From: cephi Date: Mon, 9 Dec 2024 15:06:46 -0500 Subject: [PATCH] new output --- analysis/data_stat.py | 19 -------------- .../altra_10_10_ASIC_680k_100000.json | 1 + .../altra_10_10_ASIC_680k_100000.output | 26 +++++++++++++++++++ .../altra_10_10_Oregon-2_100000.json | 1 + .../altra_10_10_Oregon-2_100000.output | 23 ++++++++++++++++ .../altra_10_10_as-caida_100000.json | 1 + .../altra_10_10_as-caida_100000.output | 24 +++++++++++++++++ .../output_test2/altra_10_10_dc2_100000.json | 1 + .../altra_10_10_dc2_100000.output | 26 +++++++++++++++++++ .../altra_10_10_de2010_100000.json | 1 + .../altra_10_10_de2010_100000.output | 25 ++++++++++++++++++ .../altra_10_10_email-Enron_100000.json | 1 + .../altra_10_10_email-Enron_100000.output | 24 +++++++++++++++++ .../altra_10_10_fl2010_100000.json | 1 + .../altra_10_10_fl2010_100000.output | 25 ++++++++++++++++++ .../altra_10_10_ga2010_100000.json | 1 + .../altra_10_10_ga2010_100000.output | 25 ++++++++++++++++++ .../altra_10_10_mac_econ_fwd500_100000.json | 1 + .../altra_10_10_mac_econ_fwd500_100000.output | 26 +++++++++++++++++++ .../altra_10_10_mc2depi_100000.json | 1 + .../altra_10_10_mc2depi_100000.output | 25 ++++++++++++++++++ .../altra_10_10_p2p-Gnutella04_100000.json | 1 + .../altra_10_10_p2p-Gnutella04_100000.output | 23 ++++++++++++++++ .../altra_10_10_p2p-Gnutella24_100000.json | 1 + .../altra_10_10_p2p-Gnutella24_100000.output | 23 ++++++++++++++++ .../altra_10_10_p2p-Gnutella25_100000.json | 1 + .../altra_10_10_p2p-Gnutella25_100000.output | 23 ++++++++++++++++ .../altra_10_10_p2p-Gnutella30_100000.json | 1 + .../altra_10_10_p2p-Gnutella30_100000.output | 23 ++++++++++++++++ .../altra_10_10_ri2010_100000.json | 1 + .../altra_10_10_ri2010_100000.output | 24 +++++++++++++++++ .../altra_10_10_rma10_100000.json | 1 + .../altra_10_10_rma10_100000.output | 25 ++++++++++++++++++ ..._10_10_soc-sign-Slashdot090216_100000.json | 1 + ...0_10_soc-sign-Slashdot090216_100000.output | 24 +++++++++++++++++ ..._10_10_soc-sign-Slashdot090221_100000.json | 2 ++ ...0_10_soc-sign-Slashdot090221_100000.output | 24 +++++++++++++++++ .../altra_10_10_soc-sign-epinions_100000.json | 1 + ...ltra_10_10_soc-sign-epinions_100000.output | 25 ++++++++++++++++++ .../altra_10_10_sx-mathoverflow_100000.json | 1 + .../altra_10_10_sx-mathoverflow_100000.output | 24 +++++++++++++++++ .../altra_10_10_tn2010_100000.json | 1 + .../altra_10_10_tn2010_100000.output | 26 +++++++++++++++++++ .../altra_10_10_ut2010_100000.json | 1 + .../altra_10_10_ut2010_100000.output | 26 +++++++++++++++++++ .../altra_10_10_va2010_100000.json | 1 + .../altra_10_10_va2010_100000.output | 26 +++++++++++++++++++ .../altra_10_10_vt2010_100000.json | 1 + .../altra_10_10_vt2010_100000.output | 24 +++++++++++++++++ .../epyc_7313p_10_10_ASIC_680k_100000.json | 1 + .../epyc_7313p_10_10_ASIC_680k_100000.output | 20 ++++++++++++++ .../epyc_7313p_10_10_Oregon-2_100000.json | 1 + .../epyc_7313p_10_10_Oregon-2_100000.output | 17 ++++++++++++ .../epyc_7313p_10_10_as-caida_100000.json | 1 + .../epyc_7313p_10_10_as-caida_100000.output | 18 +++++++++++++ .../epyc_7313p_10_10_dc2_100000.json | 1 + .../epyc_7313p_10_10_dc2_100000.output | 20 ++++++++++++++ .../epyc_7313p_10_10_de2010_100000.json | 1 + .../epyc_7313p_10_10_de2010_100000.output | 19 ++++++++++++++ .../epyc_7313p_10_10_email-Enron_100000.json | 1 + ...epyc_7313p_10_10_email-Enron_100000.output | 18 +++++++++++++ .../epyc_7313p_10_10_fl2010_100000.json | 1 + .../epyc_7313p_10_10_fl2010_100000.output | 19 ++++++++++++++ .../epyc_7313p_10_10_ga2010_100000.json | 1 + .../epyc_7313p_10_10_ga2010_100000.output | 19 ++++++++++++++ ...yc_7313p_10_10_mac_econ_fwd500_100000.json | 1 + ..._7313p_10_10_mac_econ_fwd500_100000.output | 20 ++++++++++++++ .../epyc_7313p_10_10_mc2depi_100000.json | 1 + .../epyc_7313p_10_10_mc2depi_100000.output | 19 ++++++++++++++ ...pyc_7313p_10_10_p2p-Gnutella04_100000.json | 1 + ...c_7313p_10_10_p2p-Gnutella04_100000.output | 17 ++++++++++++ ...pyc_7313p_10_10_p2p-Gnutella24_100000.json | 1 + ...c_7313p_10_10_p2p-Gnutella24_100000.output | 17 ++++++++++++ ...pyc_7313p_10_10_p2p-Gnutella25_100000.json | 1 + ...c_7313p_10_10_p2p-Gnutella25_100000.output | 17 ++++++++++++ ...pyc_7313p_10_10_p2p-Gnutella30_100000.json | 1 + ...c_7313p_10_10_p2p-Gnutella30_100000.output | 17 ++++++++++++ .../epyc_7313p_10_10_ri2010_100000.json | 1 + .../epyc_7313p_10_10_ri2010_100000.output | 18 +++++++++++++ .../epyc_7313p_10_10_rma10_100000.json | 1 + .../epyc_7313p_10_10_rma10_100000.output | 19 ++++++++++++++ ..._10_10_soc-sign-Slashdot090216_100000.json | 1 + ...0_10_soc-sign-Slashdot090216_100000.output | 18 +++++++++++++ ..._10_10_soc-sign-Slashdot090221_100000.json | 1 + ...0_10_soc-sign-Slashdot090221_100000.output | 18 +++++++++++++ ..._7313p_10_10_soc-sign-epinions_100000.json | 1 + ...313p_10_10_soc-sign-epinions_100000.output | 19 ++++++++++++++ ...yc_7313p_10_10_sx-mathoverflow_100000.json | 1 + ..._7313p_10_10_sx-mathoverflow_100000.output | 18 +++++++++++++ .../epyc_7313p_10_10_tn2010_100000.json | 1 + .../epyc_7313p_10_10_tn2010_100000.output | 21 +++++++++++++++ .../epyc_7313p_10_10_ut2010_100000.json | 1 + .../epyc_7313p_10_10_ut2010_100000.output | 20 ++++++++++++++ .../epyc_7313p_10_10_va2010_100000.json | 1 + .../epyc_7313p_10_10_va2010_100000.output | 20 ++++++++++++++ .../epyc_7313p_10_10_vt2010_100000.json | 1 + .../epyc_7313p_10_10_vt2010_100000.output | 18 +++++++++++++ pytorch/spmv.py | 2 ++ 98 files changed, 1086 insertions(+), 19 deletions(-) create mode 100644 pytorch/output_test2/altra_10_10_ASIC_680k_100000.json create mode 100644 pytorch/output_test2/altra_10_10_ASIC_680k_100000.output create mode 100644 pytorch/output_test2/altra_10_10_Oregon-2_100000.json create mode 100644 pytorch/output_test2/altra_10_10_Oregon-2_100000.output create mode 100644 pytorch/output_test2/altra_10_10_as-caida_100000.json create mode 100644 pytorch/output_test2/altra_10_10_as-caida_100000.output create mode 100644 pytorch/output_test2/altra_10_10_dc2_100000.json create mode 100644 pytorch/output_test2/altra_10_10_dc2_100000.output create mode 100644 pytorch/output_test2/altra_10_10_de2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_de2010_100000.output create mode 100644 pytorch/output_test2/altra_10_10_email-Enron_100000.json create mode 100644 pytorch/output_test2/altra_10_10_email-Enron_100000.output create mode 100644 pytorch/output_test2/altra_10_10_fl2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_fl2010_100000.output create mode 100644 pytorch/output_test2/altra_10_10_ga2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_ga2010_100000.output create mode 100644 pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.json create mode 100644 pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.output create mode 100644 pytorch/output_test2/altra_10_10_mc2depi_100000.json create mode 100644 pytorch/output_test2/altra_10_10_mc2depi_100000.output create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.json create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.output create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.json create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.output create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.json create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.output create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.json create mode 100644 pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.output create mode 100644 pytorch/output_test2/altra_10_10_ri2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_ri2010_100000.output create mode 100644 pytorch/output_test2/altra_10_10_rma10_100000.json create mode 100644 pytorch/output_test2/altra_10_10_rma10_100000.output create mode 100644 pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.json create mode 100644 pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.output create mode 100644 pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.json create mode 100644 pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.output create mode 100644 pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.json create mode 100644 pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.output create mode 100644 pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.json create mode 100644 pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.output create mode 100644 pytorch/output_test2/altra_10_10_tn2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_tn2010_100000.output create mode 100644 pytorch/output_test2/altra_10_10_ut2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_ut2010_100000.output create mode 100644 pytorch/output_test2/altra_10_10_va2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_va2010_100000.output create mode 100644 pytorch/output_test2/altra_10_10_vt2010_100000.json create mode 100644 pytorch/output_test2/altra_10_10_vt2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_dc2_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_dc2_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_de2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_de2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_rma10_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_rma10_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_va2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_va2010_100000.output create mode 100644 pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.json create mode 100644 pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.output diff --git a/analysis/data_stat.py b/analysis/data_stat.py index 391bf87..9783392 100644 --- a/analysis/data_stat.py +++ b/analysis/data_stat.py @@ -108,25 +108,6 @@ names = { } } -def parse_output_old(filename: str, data: dict[str, str]) -> dict: - result: dict[str, int | float] = dict() - cpu: CPU = CPU[data['cpu'].upper()] - - with open(filename, 'r') as file: - for line in file: - for stat in [x for x in Stat if x in cpu.value]: - regex = r'^\W*([\d+(,|\.)?]+)\W*.*' + cpu.value[stat] - value = re.search(regex, line) - - if value is None: - continue - elif stat == Stat.TASK_CLK: - result[stat.value] = float(value.group(1).replace(',', '')) - else: - result[stat.value] = int(value.group(1).replace(',', '')) - - return result | parse_power(filename, cpu) - def parse_output(output: str, cpu: Cpu) -> dict: result = dict() diff --git a/pytorch/output_test2/altra_10_10_ASIC_680k_100000.json b/pytorch/output_test2/altra_10_10_ASIC_680k_100000.json new file mode 100644 index 0000000..94de9d0 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ASIC_680k_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "ASIC_680k", "MATRIX_SHAPE": [682862, 682862], "MATRIX_SIZE": 466300511044, "MATRIX_NNZ": 3871773, "MATRIX_DENSITY": 8.303171256088674e-06, "TIME_S": 143.85276532173157, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [22.6, 22.52, 22.76, 22.68, 22.88, 22.84, 22.72, 22.48, 22.36, 22.64], "POWER": [101.6, 97.76, 86.32, 69.24, 56.32, 52.24, 57.16, 57.16, 75.88, 89.6, 100.72, 101.6, 103.6, 103.0, 104.08, 109.28, 107.8, 106.44, 104.68, 104.68, 100.04, 98.72, 98.28, 101.4, 98.96, 97.04, 94.12, 92.84, 88.4, 88.8, 93.36, 93.36, 94.12, 94.92, 95.6, 92.16, 91.6, 94.88, 95.88, 97.28, 98.36, 98.64, 96.52, 96.52, 97.24, 98.36, 95.12, 94.92, 98.72, 97.52, 94.56, 96.2, 98.04, 98.52, 102.44, 106.08, 106.08, 108.4, 107.52, 104.56, 103.16, 101.8, 103.24, 107.64, 105.52, 103.64, 104.84, 101.64, 101.64, 98.92, 95.64, 96.16, 100.24, 104.36, 105.52, 105.64, 102.0, 97.16, 95.4, 98.28, 98.28, 100.16, 102.76, 101.96, 103.16, 101.8, 105.32, 100.96, 98.44, 97.68, 97.6, 97.16, 97.16, 100.4, 101.48, 100.6, 98.96, 95.88, 93.68, 93.72, 94.44, 98.64, 100.44, 99.52, 101.52, 101.52, 98.16, 97.68, 98.28, 101.56, 100.2, 102.72, 103.8, 100.68, 103.12, 102.24, 101.28, 101.28, 100.04, 97.48, 95.08, 95.8, 94.92, 96.12, 95.16, 100.08, 104.08, 104.48, 107.4, 107.4, 109.68, 102.6, 100.44, 102.16, 99.48, 97.88, 95.96, 98.92, 102.84, 101.36, 102.48, 102.48, 100.92, 100.68, 96.48, 100.0, 102.04], "JOULES": 14125.656173429492, "POWER_AFTER": [23.04, 23.28, 23.0, 22.88, 22.84, 22.84, 22.76, 22.8, 22.96, 22.8]} diff --git a/pytorch/output_test2/altra_10_10_ASIC_680k_100000.output b/pytorch/output_test2/altra_10_10_ASIC_680k_100000.output new file mode 100644 index 0000000..145bbe8 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ASIC_680k_100000.output @@ -0,0 +1,26 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471856 queued and waiting for resources +srun: job 3471856 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 4, ..., 3871767, + 3871770, 3871773]), + col_indices=tensor([ 0, 11698, 11699, ..., 169456, 645874, + 682861]), + values=tensor([ 3.8333e-04, -3.3333e-04, -5.0000e-05, ..., + 0.0000e+00, 0.0000e+00, 7.9289e-02]), + size=(682862, 682862), nnz=3871773, layout=torch.sparse_csr) +tensor([0.6052, 0.7917, 0.7066, ..., 0.3876, 0.8366, 0.5267]) +Matrix: ASIC_680k +Shape: torch.Size([682862, 682862]) +Size: 466300511044 +NNZ: 3871773 +Density: 8.303171256088674e-06 +Time: 143.85276532173157 seconds + diff --git a/pytorch/output_test2/altra_10_10_Oregon-2_100000.json b/pytorch/output_test2/altra_10_10_Oregon-2_100000.json new file mode 100644 index 0000000..ebc7194 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_Oregon-2_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "Oregon-2", "MATRIX_SHAPE": [11806, 11806], "MATRIX_SIZE": 139381636, "MATRIX_NNZ": 65460, "MATRIX_DENSITY": 0.0004696458003979807, "TIME_S": 8.373449563980103, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [22.8, 22.64, 22.52, 22.56, 22.68, 22.44, 22.44, 22.36, 22.24, 22.36], "POWER": [97.2, 97.84, 89.08, 72.32, 58.36, 58.4, 58.84, 74.52, 74.52, 88.96, 99.68], "JOULES": 612.2254525375366, "POWER_AFTER": [22.16, 21.92, 21.92, 21.96, 21.96, 22.36, 22.24, 22.24, 22.04, 21.88]} diff --git a/pytorch/output_test2/altra_10_10_Oregon-2_100000.output b/pytorch/output_test2/altra_10_10_Oregon-2_100000.output new file mode 100644 index 0000000..d2559a2 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_Oregon-2_100000.output @@ -0,0 +1,23 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471857 queued and waiting for resources +srun: job 3471857 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 583, 584, ..., 65459, 65460, 65460]), + col_indices=tensor([ 2, 23, 27, ..., 3324, 958, 841]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(11806, 11806), + nnz=65460, layout=torch.sparse_csr) +tensor([0.8374, 0.0143, 0.3251, ..., 0.2693, 0.4062, 0.8940]) +Matrix: Oregon-2 +Shape: torch.Size([11806, 11806]) +Size: 139381636 +NNZ: 65460 +Density: 0.0004696458003979807 +Time: 8.373449563980103 seconds + diff --git a/pytorch/output_test2/altra_10_10_as-caida_100000.json b/pytorch/output_test2/altra_10_10_as-caida_100000.json new file mode 100644 index 0000000..76f9282 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_as-caida_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "as-caida", "MATRIX_SHAPE": [31379, 31379], "MATRIX_SIZE": 984641641, "MATRIX_NNZ": 106762, "MATRIX_DENSITY": 0.00010842726485909405, "TIME_S": 7.69922399520874, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.24, 21.28, 21.28, 21.12, 21.28, 21.0, 21.0, 21.16, 21.16, 21.08], "POWER": [101.04, 100.48, 90.04, 74.8, 62.16, 59.8, 62.6, 62.6, 77.84, 92.96, 103.24, 101.24, 101.84], "JOULES": 631.4889716720581, "POWER_AFTER": [21.24, 21.44, 21.36, 21.4, 21.4, 21.36, 21.44, 21.36, 21.28, 21.36]} diff --git a/pytorch/output_test2/altra_10_10_as-caida_100000.output b/pytorch/output_test2/altra_10_10_as-caida_100000.output new file mode 100644 index 0000000..b3a7860 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_as-caida_100000.output @@ -0,0 +1,24 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471796 queued and waiting for resources +srun: job 3471796 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 0, 0, ..., 106761, 106761, + 106762]), + col_indices=tensor([ 106, 329, 1040, ..., 155, 160, 12170]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(31379, 31379), + nnz=106762, layout=torch.sparse_csr) +tensor([0.3242, 0.9198, 0.8266, ..., 0.4648, 0.8946, 0.6351]) +Matrix: as-caida +Shape: torch.Size([31379, 31379]) +Size: 984641641 +NNZ: 106762 +Density: 0.00010842726485909405 +Time: 7.69922399520874 seconds + diff --git a/pytorch/output_test2/altra_10_10_dc2_100000.json b/pytorch/output_test2/altra_10_10_dc2_100000.json new file mode 100644 index 0000000..f164ec6 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_dc2_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "dc2", "MATRIX_SHAPE": [116835, 116835], "MATRIX_SIZE": 13650417225, "MATRIX_NNZ": 766396, "MATRIX_DENSITY": 5.614451099680581e-05, "TIME_S": 37.14217662811279, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.24, 20.92, 21.08, 21.16, 21.36, 21.6, 21.6, 21.28, 21.04, 21.0], "POWER": [102.16, 103.44, 92.84, 76.08, 59.2, 59.48, 59.48, 66.88, 79.48, 96.96, 106.32, 105.92, 102.72, 102.68, 100.88, 99.92, 99.32, 98.04, 97.4, 97.4, 97.08, 94.72, 94.08, 96.16, 94.52, 95.32, 94.76, 92.16, 92.76, 95.88, 96.48, 96.48, 97.4, 98.08, 97.92, 97.56, 98.44, 97.36, 97.88, 99.72, 99.52, 99.0, 97.76, 96.36, 96.36], "JOULES": 3585.0201398849495, "POWER_AFTER": [23.36, 23.08, 22.72, 22.52, 22.52, 22.12, 22.0, 21.96, 21.72, 21.72]} diff --git a/pytorch/output_test2/altra_10_10_dc2_100000.output b/pytorch/output_test2/altra_10_10_dc2_100000.output new file mode 100644 index 0000000..79c34b5 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_dc2_100000.output @@ -0,0 +1,26 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471782 queued and waiting for resources +srun: job 3471782 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 1, 2, ..., 766390, 766394, + 766396]), + col_indices=tensor([ 0, 1, 2, ..., 116833, 89, + 116834]), + values=tensor([-1.0000e+00, -1.0000e+00, -1.0000e+00, ..., + 1.0331e+01, -1.0000e-03, 1.0000e-03]), + size=(116835, 116835), nnz=766396, layout=torch.sparse_csr) +tensor([0.1528, 0.7657, 0.8355, ..., 0.4682, 0.1999, 0.0103]) +Matrix: dc2 +Shape: torch.Size([116835, 116835]) +Size: 13650417225 +NNZ: 766396 +Density: 5.614451099680581e-05 +Time: 37.14217662811279 seconds + diff --git a/pytorch/output_test2/altra_10_10_de2010_100000.json b/pytorch/output_test2/altra_10_10_de2010_100000.json new file mode 100644 index 0000000..3d331a8 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_de2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "de2010", "MATRIX_SHAPE": [24115, 24115], "MATRIX_SIZE": 581533225, "MATRIX_NNZ": 116056, "MATRIX_DENSITY": 0.0001995689928120616, "TIME_S": 8.169610738754272, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.48, 21.44, 21.32, 21.44, 21.6, 21.64, 21.44, 21.44, 21.4, 21.24], "POWER": [104.88, 103.2, 87.28, 72.12, 56.64, 56.64, 57.16, 62.64, 82.24, 97.56, 105.76, 102.64, 99.36], "JOULES": 638.1325230026245, "POWER_AFTER": [21.24, 21.32, 21.32, 21.24, 20.84, 20.84, 20.84, 20.84, 21.16, 21.56]} diff --git a/pytorch/output_test2/altra_10_10_de2010_100000.output b/pytorch/output_test2/altra_10_10_de2010_100000.output new file mode 100644 index 0000000..bbe38c5 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_de2010_100000.output @@ -0,0 +1,25 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471783 queued and waiting for resources +srun: job 3471783 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 13, 21, ..., 116047, 116051, + 116056]), + col_indices=tensor([ 250, 251, 757, ..., 23334, 23553, 24050]), + values=tensor([ 14900., 33341., 20255., ..., 164227., 52413., + 16949.]), size=(24115, 24115), nnz=116056, + layout=torch.sparse_csr) +tensor([0.3069, 0.2208, 0.9592, ..., 0.2726, 0.0490, 0.9363]) +Matrix: de2010 +Shape: torch.Size([24115, 24115]) +Size: 581533225 +NNZ: 116056 +Density: 0.0001995689928120616 +Time: 8.169610738754272 seconds + diff --git a/pytorch/output_test2/altra_10_10_email-Enron_100000.json b/pytorch/output_test2/altra_10_10_email-Enron_100000.json new file mode 100644 index 0000000..7897789 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_email-Enron_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "email-Enron", "MATRIX_SHAPE": [36692, 36692], "MATRIX_SIZE": 1346302864, "MATRIX_NNZ": 367662, "MATRIX_DENSITY": 0.0002730901120626302, "TIME_S": 12.88691234588623, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [20.76, 20.88, 20.84, 21.0, 21.16, 21.24, 21.44, 21.44, 21.48, 21.36], "POWER": [99.08, 99.68, 91.12, 77.0, 63.04, 57.0, 58.84, 75.68, 90.8, 90.8, 104.72, 103.32, 100.28, 97.32, 95.44, 93.56, 93.36], "JOULES": 1113.6021366119385, "POWER_AFTER": [21.72, 21.72, 21.8, 21.72, 21.92, 21.88, 21.88, 21.76, 21.44, 21.28]} diff --git a/pytorch/output_test2/altra_10_10_email-Enron_100000.output b/pytorch/output_test2/altra_10_10_email-Enron_100000.output new file mode 100644 index 0000000..49f40f7 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_email-Enron_100000.output @@ -0,0 +1,24 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471786 queued and waiting for resources +srun: job 3471786 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 1, 71, ..., 367660, 367661, + 367662]), + col_indices=tensor([ 1, 0, 2, ..., 36690, 36689, 8203]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(36692, 36692), + nnz=367662, layout=torch.sparse_csr) +tensor([0.6039, 0.3557, 0.6656, ..., 0.1586, 0.2866, 0.7610]) +Matrix: email-Enron +Shape: torch.Size([36692, 36692]) +Size: 1346302864 +NNZ: 367662 +Density: 0.0002730901120626302 +Time: 12.88691234588623 seconds + diff --git a/pytorch/output_test2/altra_10_10_fl2010_100000.json b/pytorch/output_test2/altra_10_10_fl2010_100000.json new file mode 100644 index 0000000..92ea71d --- /dev/null +++ b/pytorch/output_test2/altra_10_10_fl2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "fl2010", "MATRIX_SHAPE": [484481, 484481], "MATRIX_SIZE": 234721839361, "MATRIX_NNZ": 2346294, "MATRIX_DENSITY": 9.99606174861054e-06, "TIME_S": 31.069382905960083, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.64, 21.72, 21.64, 21.68, 21.4, 21.44, 21.24, 21.12, 21.12, 21.04], "POWER": [120.04, 120.92, 98.92, 79.32, 58.8, 57.84, 62.68, 62.68, 77.64, 100.44, 114.44, 116.72, 118.8, 121.36, 118.48, 116.4, 114.24, 113.88, 109.72, 109.72, 117.92, 119.64, 115.56, 112.28, 107.04, 104.52, 105.32, 109.56, 109.76, 110.6, 113.36, 113.36, 116.64], "JOULES": 3220.0128221511845, "POWER_AFTER": [22.36, 22.24, 22.52, 22.4, 22.44, 22.4, 22.28, 23.56, 23.56, 25.52]} diff --git a/pytorch/output_test2/altra_10_10_fl2010_100000.output b/pytorch/output_test2/altra_10_10_fl2010_100000.output new file mode 100644 index 0000000..163aa12 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_fl2010_100000.output @@ -0,0 +1,25 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471829 queued and waiting for resources +srun: job 3471829 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 2, 5, ..., 2346288, + 2346292, 2346294]), + col_indices=tensor([ 1513, 5311, 947, ..., 484460, 482463, + 484022]), + values=tensor([28364., 12497., 11567., ..., 8532., 22622., 35914.]), + size=(484481, 484481), nnz=2346294, layout=torch.sparse_csr) +tensor([0.5157, 0.5811, 0.2529, ..., 0.9249, 0.1469, 0.4136]) +Matrix: fl2010 +Shape: torch.Size([484481, 484481]) +Size: 234721839361 +NNZ: 2346294 +Density: 9.99606174861054e-06 +Time: 31.069382905960083 seconds + diff --git a/pytorch/output_test2/altra_10_10_ga2010_100000.json b/pytorch/output_test2/altra_10_10_ga2010_100000.json new file mode 100644 index 0000000..d45c8be --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ga2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "ga2010", "MATRIX_SHAPE": [291086, 291086], "MATRIX_SIZE": 84731059396, "MATRIX_NNZ": 1418056, "MATRIX_DENSITY": 1.6735964475229304e-05, "TIME_S": 17.813313722610474, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.56, 21.72, 21.68, 21.56, 21.56, 21.76, 21.2, 21.16, 20.96, 20.96], "POWER": [115.36, 114.88, 106.84, 89.28, 71.36, 58.8, 57.92, 74.0, 91.48, 110.52, 113.56, 113.56, 116.72, 113.88, 117.16, 119.4, 113.4, 113.76, 111.48, 110.64, 115.04, 121.64], "JOULES": 1821.6114812183382, "POWER_AFTER": [21.56, 21.56, 21.4, 21.28, 21.36, 21.48, 21.64, 21.84, 21.68, 21.88]} diff --git a/pytorch/output_test2/altra_10_10_ga2010_100000.output b/pytorch/output_test2/altra_10_10_ga2010_100000.output new file mode 100644 index 0000000..8557254 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ga2010_100000.output @@ -0,0 +1,25 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471788 queued and waiting for resources +srun: job 3471788 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 10, ..., 1418047, + 1418054, 1418056]), + col_indices=tensor([ 1566, 1871, 1997, ..., 291064, 289820, + 290176]), + values=tensor([18760., 17851., 18847., ..., 65219., 56729., 77629.]), + size=(291086, 291086), nnz=1418056, layout=torch.sparse_csr) +tensor([0.6229, 0.6308, 0.8573, ..., 0.9191, 0.9418, 0.6011]) +Matrix: ga2010 +Shape: torch.Size([291086, 291086]) +Size: 84731059396 +NNZ: 1418056 +Density: 1.6735964475229304e-05 +Time: 17.813313722610474 seconds + diff --git a/pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.json b/pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.json new file mode 100644 index 0000000..2fa64f9 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "mac_econ_fwd500", "MATRIX_SHAPE": [206500, 206500], "MATRIX_SIZE": 42642250000, "MATRIX_NNZ": 1273389, "MATRIX_DENSITY": 2.9862143765866013e-05, "TIME_S": 13.249896049499512, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.04, 21.08, 21.12, 21.24, 21.32, 21.32, 21.2, 21.04, 21.08, 21.08], "POWER": [109.92, 114.24, 106.48, 89.76, 70.72, 60.84, 56.4, 74.08, 74.08, 94.04, 110.84, 111.28, 109.44, 109.48, 110.12, 107.64, 114.36, 119.48, 121.56, 121.4], "JOULES": 1343.1373804092407, "POWER_AFTER": [21.44, 21.4, 21.24, 21.28, 21.92, 22.56, 23.52, 24.28, 24.28, 24.4]} diff --git a/pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.output b/pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.output new file mode 100644 index 0000000..c038e87 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_mac_econ_fwd500_100000.output @@ -0,0 +1,26 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471828 queued and waiting for resources +srun: job 3471828 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 8, ..., 1273376, + 1273379, 1273389]), + col_indices=tensor([ 3, 30, 44, ..., 206363, 206408, + 206459]), + values=tensor([-3.7877e-03, -1.5420e-01, 9.5305e-04, ..., + 1.2290e-01, 2.2235e-01, -1.0000e+00]), + size=(206500, 206500), nnz=1273389, layout=torch.sparse_csr) +tensor([0.7751, 0.0281, 0.9910, ..., 0.3020, 0.8213, 0.1857]) +Matrix: mac_econ_fwd500 +Shape: torch.Size([206500, 206500]) +Size: 42642250000 +NNZ: 1273389 +Density: 2.9862143765866013e-05 +Time: 13.249896049499512 seconds + diff --git a/pytorch/output_test2/altra_10_10_mc2depi_100000.json b/pytorch/output_test2/altra_10_10_mc2depi_100000.json new file mode 100644 index 0000000..25bae52 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_mc2depi_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "mc2depi", "MATRIX_SHAPE": [525825, 525825], "MATRIX_SIZE": 276491930625, "MATRIX_NNZ": 2100225, "MATRIX_DENSITY": 7.595972132902821e-06, "TIME_S": 19.404656887054443, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.72, 21.48, 21.32, 21.52, 21.52, 21.28, 21.4, 21.56, 21.56, 21.4], "POWER": [116.0, 110.56, 110.56, 91.76, 72.12, 61.36, 61.52, 74.2, 90.92, 117.8, 125.96, 128.68, 130.72, 124.72, 124.72, 115.08, 119.88, 117.36, 115.44, 110.44, 111.24, 110.4], "JOULES": 2049.1541203308107, "POWER_AFTER": [21.6, 21.64, 21.64, 21.84, 21.84, 21.76, 21.76, 21.72, 21.72, 21.4]} diff --git a/pytorch/output_test2/altra_10_10_mc2depi_100000.output b/pytorch/output_test2/altra_10_10_mc2depi_100000.output new file mode 100644 index 0000000..486689e --- /dev/null +++ b/pytorch/output_test2/altra_10_10_mc2depi_100000.output @@ -0,0 +1,25 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471790 queued and waiting for resources +srun: job 3471790 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 2, 5, ..., 2100220, + 2100223, 2100225]), + col_indices=tensor([ 0, 1, 1, ..., 525824, 525821, + 525824]), + values=tensor([-2025., 2025., -2026., ..., 2025., 1024., -1024.]), + size=(525825, 525825), nnz=2100225, layout=torch.sparse_csr) +tensor([0.4809, 0.5361, 0.4713, ..., 0.3506, 0.4153, 0.4817]) +Matrix: mc2depi +Shape: torch.Size([525825, 525825]) +Size: 276491930625 +NNZ: 2100225 +Density: 7.595972132902821e-06 +Time: 19.404656887054443 seconds + diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.json b/pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.json new file mode 100644 index 0000000..e13712a --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella04", "MATRIX_SHAPE": [10879, 10879], "MATRIX_SIZE": 118352641, "MATRIX_NNZ": 39994, "MATRIX_DENSITY": 0.0003379223282393842, "TIME_S": 7.197759389877319, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.12, 21.2, 21.04, 20.92, 21.0, 21.04, 21.04, 20.72, 20.76, 21.12], "POWER": [100.84, 100.04, 85.4, 71.68, 71.68, 56.24, 57.76, 66.84, 79.88, 94.24, 101.2, 100.24, 98.4, 96.36, 95.08], "JOULES": 655.9629627895355, "POWER_AFTER": [21.68, 21.04, 20.88, 21.4, 21.28, 21.28, 21.32, 21.04, 21.04, 21.04]} diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.output b/pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.output new file mode 100644 index 0000000..d5225b0 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella04_100000.output @@ -0,0 +1,23 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471826 queued and waiting for resources +srun: job 3471826 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 10, 20, ..., 39994, 39994, 39994]), + col_indices=tensor([ 1, 2, 3, ..., 9711, 10875, 10876]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(10879, 10879), + nnz=39994, layout=torch.sparse_csr) +tensor([0.2810, 0.9768, 0.5232, ..., 0.2583, 0.8876, 0.2861]) +Matrix: p2p-Gnutella04 +Shape: torch.Size([10879, 10879]) +Size: 118352641 +NNZ: 39994 +Density: 0.0003379223282393842 +Time: 7.197759389877319 seconds + diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.json b/pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.json new file mode 100644 index 0000000..ef0ec71 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella24", "MATRIX_SHAPE": [26518, 26518], "MATRIX_SIZE": 703204324, "MATRIX_NNZ": 65369, "MATRIX_DENSITY": 9.295875717624285e-05, "TIME_S": 8.68448281288147, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.84, 22.0, 21.92, 22.0, 21.92, 21.56, 21.52, 21.56, 21.48, 21.48], "POWER": [94.8, 95.36, 83.32, 69.76, 57.92, 59.48, 65.64, 65.64, 82.24, 99.92, 105.68, 103.0, 101.92, 99.32], "JOULES": 751.5028329753875, "POWER_AFTER": [24.12, 24.56, 24.56, 24.56, 24.28, 24.16, 24.24, 24.28, 24.28, 24.08]} diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.output b/pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.output new file mode 100644 index 0000000..6d6442d --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella24_100000.output @@ -0,0 +1,23 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471854 queued and waiting for resources +srun: job 3471854 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 9, 9, ..., 65369, 65369, 65369]), + col_indices=tensor([ 1, 2, 3, ..., 15065, 9401, 26517]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(26518, 26518), + nnz=65369, layout=torch.sparse_csr) +tensor([0.4761, 0.4887, 0.0195, ..., 0.5651, 0.2234, 0.2511]) +Matrix: p2p-Gnutella24 +Shape: torch.Size([26518, 26518]) +Size: 703204324 +NNZ: 65369 +Density: 9.295875717624285e-05 +Time: 8.68448281288147 seconds + diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.json b/pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.json new file mode 100644 index 0000000..c6e61c4 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella25", "MATRIX_SHAPE": [22687, 22687], "MATRIX_SIZE": 514699969, "MATRIX_NNZ": 54705, "MATRIX_DENSITY": 0.00010628522108964806, "TIME_S": 8.185347080230713, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.32, 21.32, 21.2, 21.04, 20.96, 21.04, 21.16, 21.28, 21.24, 21.2], "POWER": [97.72, 94.2, 94.2, 82.44, 65.96, 58.0, 62.48, 71.32, 83.8, 97.0, 99.48, 98.56, 99.76], "JOULES": 655.0902247238158, "POWER_AFTER": [21.28, 21.28, 21.32, 21.48, 21.28, 21.48, 21.44, 21.12, 20.8, 20.76]} diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.output b/pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.output new file mode 100644 index 0000000..860e3b3 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella25_100000.output @@ -0,0 +1,23 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471827 queued and waiting for resources +srun: job 3471827 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 9, 9, ..., 54704, 54704, 54705]), + col_indices=tensor([ 1, 2, 3, ..., 17949, 22685, 144]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(22687, 22687), + nnz=54705, layout=torch.sparse_csr) +tensor([0.4250, 0.5008, 0.7599, ..., 0.4696, 0.2842, 0.9247]) +Matrix: p2p-Gnutella25 +Shape: torch.Size([22687, 22687]) +Size: 514699969 +NNZ: 54705 +Density: 0.00010628522108964806 +Time: 8.185347080230713 seconds + diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.json b/pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.json new file mode 100644 index 0000000..4f06737 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella30", "MATRIX_SHAPE": [36682, 36682], "MATRIX_SIZE": 1345569124, "MATRIX_NNZ": 88328, "MATRIX_DENSITY": 6.564359899804003e-05, "TIME_S": 9.74808645248413, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.08, 21.04, 21.16, 21.16, 21.24, 21.04, 21.08, 21.04, 21.0, 21.08], "POWER": [97.28, 94.08, 85.16, 71.12, 58.08, 59.6, 62.64, 79.28, 97.68, 97.68, 108.48, 106.6, 104.24, 101.56], "JOULES": 850.2556601142884, "POWER_AFTER": [20.92, 20.84, 20.96, 21.32, 21.28, 21.4, 21.44, 21.16, 21.16, 21.24]} diff --git a/pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.output b/pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.output new file mode 100644 index 0000000..f8a4d57 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_p2p-Gnutella30_100000.output @@ -0,0 +1,23 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471825 queued and waiting for resources +srun: job 3471825 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 10, 10, ..., 88328, 88328, 88328]), + col_indices=tensor([ 1, 2, 3, ..., 36675, 36676, 36677]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(36682, 36682), + nnz=88328, layout=torch.sparse_csr) +tensor([0.5280, 0.0933, 0.8124, ..., 0.0433, 0.2447, 0.2625]) +Matrix: p2p-Gnutella30 +Shape: torch.Size([36682, 36682]) +Size: 1345569124 +NNZ: 88328 +Density: 6.564359899804003e-05 +Time: 9.74808645248413 seconds + diff --git a/pytorch/output_test2/altra_10_10_ri2010_100000.json b/pytorch/output_test2/altra_10_10_ri2010_100000.json new file mode 100644 index 0000000..9ae93fc --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ri2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "ri2010", "MATRIX_SHAPE": [25181, 25181], "MATRIX_SIZE": 634082761, "MATRIX_NNZ": 125750, "MATRIX_DENSITY": 0.00019831796057928155, "TIME_S": 7.650730133056641, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [17.16, 16.92, 16.6, 16.44, 16.52, 16.56, 16.84, 16.76, 16.84, 17.12], "POWER": [100.16, 98.52, 90.92, 76.44, 58.2, 51.52, 52.8, 65.68, 82.84, 97.64, 96.8, 94.84, 94.84], "JOULES": 603.8352458190918, "POWER_AFTER": [16.72, 16.96, 17.0, 16.92, 17.28, 17.12, 17.0, 17.04, 17.28, 17.2]} diff --git a/pytorch/output_test2/altra_10_10_ri2010_100000.output b/pytorch/output_test2/altra_10_10_ri2010_100000.output new file mode 100644 index 0000000..320f268 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ri2010_100000.output @@ -0,0 +1,24 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471779 queued and waiting for resources +srun: job 3471779 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 8, ..., 125742, 125747, + 125750]), + col_indices=tensor([ 25, 56, 662, ..., 21738, 22279, 23882]), + values=tensor([17171., 37318., 5284., ..., 25993., 24918., 803.]), + size=(25181, 25181), nnz=125750, layout=torch.sparse_csr) +tensor([0.2875, 0.2982, 0.0876, ..., 0.4058, 0.8442, 0.7364]) +Matrix: ri2010 +Shape: torch.Size([25181, 25181]) +Size: 634082761 +NNZ: 125750 +Density: 0.00019831796057928155 +Time: 7.650730133056641 seconds + diff --git a/pytorch/output_test2/altra_10_10_rma10_100000.json b/pytorch/output_test2/altra_10_10_rma10_100000.json new file mode 100644 index 0000000..0de6415 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_rma10_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "rma10", "MATRIX_SHAPE": [46835, 46835], "MATRIX_SIZE": 2193517225, "MATRIX_NNZ": 2374001, "MATRIX_DENSITY": 0.0010822805369125833, "TIME_S": 18.274461030960083, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [22.08, 21.84, 22.0, 22.0, 21.84, 21.96, 21.92, 21.6, 21.68, 21.84], "POWER": [117.12, 115.04, 110.24, 95.24, 76.96, 64.8, 58.44, 69.48, 90.84, 109.8, 115.72, 121.2, 121.2, 120.8, 116.48, 112.48, 110.0, 110.48, 109.6, 107.6, 108.0, 110.32], "JOULES": 1849.3985409355162, "POWER_AFTER": [22.12, 22.0, 21.84, 21.76, 21.72, 21.64, 21.6, 21.56, 21.64, 21.64]} diff --git a/pytorch/output_test2/altra_10_10_rma10_100000.output b/pytorch/output_test2/altra_10_10_rma10_100000.output new file mode 100644 index 0000000..dda985b --- /dev/null +++ b/pytorch/output_test2/altra_10_10_rma10_100000.output @@ -0,0 +1,25 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471832 queued and waiting for resources +srun: job 3471832 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 17, 34, ..., 2373939, + 2373970, 2374001]), + col_indices=tensor([ 0, 1, 2, ..., 46831, 46833, 46834]), + values=tensor([ 1.2636e+05, -1.6615e+07, -8.2015e+04, ..., + 8.3378e+01, 2.5138e+00, 1.2184e+03]), + size=(46835, 46835), nnz=2374001, layout=torch.sparse_csr) +tensor([0.9389, 0.2472, 0.7378, ..., 0.8609, 0.3319, 0.1508]) +Matrix: rma10 +Shape: torch.Size([46835, 46835]) +Size: 2193517225 +NNZ: 2374001 +Density: 0.0010822805369125833 +Time: 18.274461030960083 seconds + diff --git a/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.json b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.json new file mode 100644 index 0000000..2e64644 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "soc-sign-Slashdot090216", "MATRIX_SHAPE": [81871, 81871], "MATRIX_SIZE": 6702860641, "MATRIX_NNZ": 545671, "MATRIX_DENSITY": 8.140867447881048e-05, "TIME_S": 21.485024452209473, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.56, 21.36, 21.36, 21.72, 21.72, 21.88, 21.88, 22.0, 21.76, 21.36], "POWER": [101.32, 102.64, 100.4, 82.28, 66.8, 62.8, 62.88, 75.04, 91.52, 107.12, 104.28, 103.72, 102.2, 102.2, 103.2, 103.24, 107.08, 108.24, 106.2], "JOULES": 1738.469596824646, "POWER_AFTER": [21.72, 21.68, 21.52, 21.4, 21.56, 21.72, 21.84, 22.08, 22.08, 22.08]} diff --git a/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.output b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.output new file mode 100644 index 0000000..ba7cba7 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090216_100000.output @@ -0,0 +1,24 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471855 queued and waiting for resources +srun: job 3471855 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 29, 124, ..., 545669, 545669, + 545671]), + col_indices=tensor([ 1, 2, 3, ..., 81869, 81699, 81863]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(81871, 81871), + nnz=545671, layout=torch.sparse_csr) +tensor([0.7599, 0.3131, 0.1356, ..., 0.5599, 0.7303, 0.7084]) +Matrix: soc-sign-Slashdot090216 +Shape: torch.Size([81871, 81871]) +Size: 6702860641 +NNZ: 545671 +Density: 8.140867447881048e-05 +Time: 21.485024452209473 seconds + diff --git a/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.json b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.json new file mode 100644 index 0000000..7134d2c --- /dev/null +++ b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.json @@ -0,0 +1,2 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "soc-sign-Slashdot090221", "MATRIX_SHAPE": [82144, 82144], "MATRIX_SIZE": 6747636736, "MATRIX_NNZ": 549202, "MATRIX_DENSITY": 8.13917555860553e-05, "TIME_S": 9.906620264053345, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.48, 21.28, 21.6, 21.6, 21.88, 21.96, 21.96, 21.84, 21.4, 21.44], "POWER": [102.12, 100.8, 88.12, 72.96, 55.76, 56.68, 60.92, 60.92, 77.88, 98.52, 109.4, 109.76, 111.0, 109.04, 106.48, 104.8, 105.32, 102.52], "JOULES": 1025.1467094707486, "POWER_AFTER": [21.88, 21.84, 21.76, 21.24, 21.36, 21.4, 21.32, 21.48, 21.6, 21.48]} +48, 21.4, 21.12, 20.92, 20.6]} diff --git a/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.output b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.output new file mode 100644 index 0000000..bdcacf0 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_soc-sign-Slashdot090221_100000.output @@ -0,0 +1,24 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471791 queued and waiting for resources +srun: job 3471791 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 29, 124, ..., 549200, 549200, + 549202]), + col_indices=tensor([ 1, 2, 3, ..., 82142, 81974, 82136]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(82144, 82144), + nnz=549202, layout=torch.sparse_csr) +tensor([0.7291, 0.8277, 0.0975, ..., 0.0057, 0.6109, 0.6944]) +Matrix: soc-sign-Slashdot090221 +Shape: torch.Size([82144, 82144]) +Size: 6747636736 +NNZ: 549202 +Density: 8.13917555860553e-05 +Time: 9.906620264053345 seconds + diff --git a/pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.json b/pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.json new file mode 100644 index 0000000..7a4388d --- /dev/null +++ b/pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "soc-sign-epinions", "MATRIX_SHAPE": [131828, 131828], "MATRIX_SIZE": 17378621584, "MATRIX_NNZ": 841372, "MATRIX_DENSITY": 4.841419648464106e-05, "TIME_S": 31.47378420829773, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [17.16, 17.16, 17.12, 17.2, 17.16, 17.36, 17.52, 17.6, 17.48, 17.44], "POWER": [93.96, 93.96, 93.72, 75.92, 62.56, 53.68, 55.84, 68.12, 84.12, 103.88, 107.88, 106.84, 107.16, 107.16, 104.4, 100.96, 96.64, 96.08, 98.24, 100.16, 97.92, 97.44, 96.24, 94.64, 90.36, 92.96, 92.96, 92.96, 91.56, 90.88, 91.24, 93.72, 95.72], "JOULES": 2885.550624418259, "POWER_AFTER": [18.24, 18.04, 18.0, 18.0, 17.88, 17.8, 18.2, 18.28, 18.48, 18.52]} diff --git a/pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.output b/pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.output new file mode 100644 index 0000000..22be23c --- /dev/null +++ b/pytorch/output_test2/altra_10_10_soc-sign-epinions_100000.output @@ -0,0 +1,25 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471823 queued and waiting for resources +srun: job 3471823 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 1, 2, ..., 841371, 841371, + 841372]), + col_indices=tensor([ 1, 128552, 3, ..., 131824, 131826, + 7714]), + values=tensor([-1., -1., 1., ..., 1., 1., 1.]), + size=(131828, 131828), nnz=841372, layout=torch.sparse_csr) +tensor([0.4186, 0.4768, 0.7650, ..., 0.7266, 0.5735, 0.6056]) +Matrix: soc-sign-epinions +Shape: torch.Size([131828, 131828]) +Size: 17378621584 +NNZ: 841372 +Density: 4.841419648464106e-05 +Time: 31.47378420829773 seconds + diff --git a/pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.json b/pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.json new file mode 100644 index 0000000..2f91ad3 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "sx-mathoverflow", "MATRIX_SHAPE": [24818, 24818], "MATRIX_SIZE": 615933124, "MATRIX_NNZ": 239978, "MATRIX_DENSITY": 0.00038961697406616504, "TIME_S": 9.512531042098999, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.16, 20.8, 20.84, 20.76, 21.0, 21.28, 21.36, 21.36, 21.56, 21.6], "POWER": [99.92, 100.08, 87.32, 72.88, 59.16, 50.88, 50.88, 54.48, 71.84, 89.72, 105.24, 106.76, 106.32, 104.48], "JOULES": 748.8292432785034, "POWER_AFTER": [21.2, 20.92, 20.92, 20.92, 21.2, 21.04, 21.08, 21.4, 21.08, 21.16]} diff --git a/pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.output b/pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.output new file mode 100644 index 0000000..deb95fa --- /dev/null +++ b/pytorch/output_test2/altra_10_10_sx-mathoverflow_100000.output @@ -0,0 +1,24 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471784 queued and waiting for resources +srun: job 3471784 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 317, 416, ..., 239976, 239977, + 239978]), + col_indices=tensor([ 0, 1, 2, ..., 1483, 2179, 24817]), + values=tensor([151., 17., 6., ..., 1., 1., 1.]), + size=(24818, 24818), nnz=239978, layout=torch.sparse_csr) +tensor([0.0721, 0.7772, 0.5440, ..., 0.2599, 0.9247, 0.3684]) +Matrix: sx-mathoverflow +Shape: torch.Size([24818, 24818]) +Size: 615933124 +NNZ: 239978 +Density: 0.00038961697406616504 +Time: 9.512531042098999 seconds + diff --git a/pytorch/output_test2/altra_10_10_tn2010_100000.json b/pytorch/output_test2/altra_10_10_tn2010_100000.json new file mode 100644 index 0000000..740f027 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_tn2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "tn2010", "MATRIX_SHAPE": [240116, 240116], "MATRIX_SIZE": 57655693456, "MATRIX_NNZ": 1193966, "MATRIX_DENSITY": 2.070855328296721e-05, "TIME_S": 16.210495948791504, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.44, 21.16, 21.24, 21.36, 21.16, 21.48, 21.28, 21.16, 21.0, 20.84], "POWER": [107.88, 109.4, 94.08, 72.28, 58.64, 54.36, 59.12, 78.8, 78.8, 95.44, 110.56, 109.2, 109.28, 105.92, 108.24, 107.16, 106.4, 109.0, 111.52], "JOULES": 1480.7545082092288, "POWER_AFTER": [21.68, 21.64, 21.64, 21.6, 21.6, 21.52, 21.56, 21.76, 22.08, 22.48]} diff --git a/pytorch/output_test2/altra_10_10_tn2010_100000.output b/pytorch/output_test2/altra_10_10_tn2010_100000.output new file mode 100644 index 0000000..a98c8ff --- /dev/null +++ b/pytorch/output_test2/altra_10_10_tn2010_100000.output @@ -0,0 +1,26 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471794 queued and waiting for resources +srun: job 3471794 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 4, 20, ..., 1193961, + 1193963, 1193966]), + col_indices=tensor([ 1152, 1272, 1961, ..., 238254, 239142, + 240113]), + values=tensor([ 5728., 2871., 418449., ..., 10058., 33324., + 34928.]), size=(240116, 240116), nnz=1193966, + layout=torch.sparse_csr) +tensor([0.7187, 0.4492, 0.0121, ..., 0.1002, 0.2839, 0.4108]) +Matrix: tn2010 +Shape: torch.Size([240116, 240116]) +Size: 57655693456 +NNZ: 1193966 +Density: 2.070855328296721e-05 +Time: 16.210495948791504 seconds + diff --git a/pytorch/output_test2/altra_10_10_ut2010_100000.json b/pytorch/output_test2/altra_10_10_ut2010_100000.json new file mode 100644 index 0000000..1ee8891 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ut2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "ut2010", "MATRIX_SHAPE": [115406, 115406], "MATRIX_SIZE": 13318544836, "MATRIX_NNZ": 572066, "MATRIX_DENSITY": 4.295259032005559e-05, "TIME_S": 14.674797296524048, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [20.72, 21.0, 20.96, 20.92, 20.76, 20.68, 20.84, 20.64, 20.72, 20.96], "POWER": [105.44, 105.04, 90.04, 74.36, 58.48, 58.48, 56.4, 64.68, 79.0, 96.68, 106.04, 107.04, 105.64, 109.16, 108.88, 108.28, 106.32, 106.16, 106.16, 103.52], "JOULES": 1388.7750161361696, "POWER_AFTER": [21.24, 21.04, 21.08, 21.08, 21.12, 20.88, 20.88, 20.96, 20.88, 21.0]} diff --git a/pytorch/output_test2/altra_10_10_ut2010_100000.output b/pytorch/output_test2/altra_10_10_ut2010_100000.output new file mode 100644 index 0000000..cb8b45e --- /dev/null +++ b/pytorch/output_test2/altra_10_10_ut2010_100000.output @@ -0,0 +1,26 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471824 queued and waiting for resources +srun: job 3471824 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 9, ..., 572056, 572061, + 572066]), + col_indices=tensor([ 453, 1291, 1979, ..., 113521, 114509, + 114602]), + values=tensor([160642., 31335., 282373., ..., 88393., 99485., + 18651.]), size=(115406, 115406), nnz=572066, + layout=torch.sparse_csr) +tensor([0.7125, 0.8600, 0.2723, ..., 0.9659, 0.9794, 0.8036]) +Matrix: ut2010 +Shape: torch.Size([115406, 115406]) +Size: 13318544836 +NNZ: 572066 +Density: 4.295259032005559e-05 +Time: 14.674797296524048 seconds + diff --git a/pytorch/output_test2/altra_10_10_va2010_100000.json b/pytorch/output_test2/altra_10_10_va2010_100000.json new file mode 100644 index 0000000..90e8f6c --- /dev/null +++ b/pytorch/output_test2/altra_10_10_va2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "va2010", "MATRIX_SHAPE": [285762, 285762], "MATRIX_SIZE": 81659920644, "MATRIX_NNZ": 1402128, "MATRIX_DENSITY": 1.717033263003816e-05, "TIME_S": 21.11183762550354, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [21.8, 21.72, 21.76, 21.88, 21.84, 21.92, 21.96, 22.08, 22.08, 22.12], "POWER": [110.68, 110.72, 94.88, 76.76, 76.76, 61.96, 63.16, 70.68, 91.52, 111.84, 121.16, 120.44, 118.12, 116.68, 114.48, 116.84, 116.84, 113.72, 113.04, 110.4], "JOULES": 1932.6268738555905, "POWER_AFTER": [22.2, 22.44, 22.88, 22.52, 22.32, 22.32, 22.04, 21.76, 21.68, 21.92]} diff --git a/pytorch/output_test2/altra_10_10_va2010_100000.output b/pytorch/output_test2/altra_10_10_va2010_100000.output new file mode 100644 index 0000000..9c2d785 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_va2010_100000.output @@ -0,0 +1,26 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471831 queued and waiting for resources +srun: job 3471831 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 2, 8, ..., 1402119, + 1402123, 1402128]), + col_indices=tensor([ 2006, 2464, 1166, ..., 285581, 285634, + 285760]), + values=tensor([125334., 3558., 1192., ..., 10148., 1763., + 9832.]), size=(285762, 285762), nnz=1402128, + layout=torch.sparse_csr) +tensor([0.4623, 0.7205, 0.5451, ..., 0.0101, 0.1478, 0.8275]) +Matrix: va2010 +Shape: torch.Size([285762, 285762]) +Size: 81659920644 +NNZ: 1402128 +Density: 1.717033263003816e-05 +Time: 21.11183762550354 seconds + diff --git a/pytorch/output_test2/altra_10_10_vt2010_100000.json b/pytorch/output_test2/altra_10_10_vt2010_100000.json new file mode 100644 index 0000000..eeec04a --- /dev/null +++ b/pytorch/output_test2/altra_10_10_vt2010_100000.json @@ -0,0 +1 @@ +{"CPU": "ALTRA", "ITERATIONS": 100000, "MATRIX_FILE": "vt2010", "MATRIX_SHAPE": [32580, 32580], "MATRIX_SIZE": 1061456400, "MATRIX_NNZ": 155598, "MATRIX_DENSITY": 0.00014658915806621921, "TIME_S": 10.058021783828735, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [22.12, 21.92, 21.84, 21.48, 21.44, 21.6, 21.72, 21.48, 21.88, 22.0], "POWER": [107.8, 104.12, 91.0, 91.0, 73.2, 62.64, 62.2, 72.84, 88.0, 105.08, 109.68, 107.92, 107.12], "JOULES": 869.7752934837342, "POWER_AFTER": [21.96, 21.84, 22.04, 21.92, 21.96, 21.88, 21.96, 21.92, 21.84, 21.76]} diff --git a/pytorch/output_test2/altra_10_10_vt2010_100000.output b/pytorch/output_test2/altra_10_10_vt2010_100000.output new file mode 100644 index 0000000..708f767 --- /dev/null +++ b/pytorch/output_test2/altra_10_10_vt2010_100000.output @@ -0,0 +1,24 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: ################################################################################ +srun: # Please note that the oasis compute nodes have aarch64 architecture CPUs. # +srun: # All submission nodes and all other compute nodes have x86_64 architecture # +srun: # CPUs. Programs, environments, or other software that was built on x86_64 # +srun: # nodes may need to be rebuilt to properly execute on these nodes. # +srun: ################################################################################ +srun: job 3471830 queued and waiting for resources +srun: job 3471830 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 4, 7, ..., 155588, 155592, + 155598]), + col_indices=tensor([ 131, 561, 996, ..., 32237, 32238, 32570]), + values=tensor([79040., 7820., 15136., ..., 2828., 17986., 2482.]), + size=(32580, 32580), nnz=155598, layout=torch.sparse_csr) +tensor([0.7636, 0.2831, 0.7866, ..., 0.4359, 0.2796, 0.7453]) +Matrix: vt2010 +Shape: torch.Size([32580, 32580]) +Size: 1061456400 +NNZ: 155598 +Density: 0.00014658915806621921 +Time: 10.058021783828735 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.json b/pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.json new file mode 100644 index 0000000..4e4efee --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "ASIC_680k", "MATRIX_SHAPE": [682862, 682862], "MATRIX_SIZE": 466300511044, "MATRIX_NNZ": 3871773, "MATRIX_DENSITY": 8.303171256088674e-06, "TIME_S": 77.6055359840393, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [41.35, 40.33, 40.46, 40.22, 39.91, 39.8, 40.53, 39.93, 40.0, 39.65], "POWER": [139.19], "JOULES": 10801.914553618431, "POWER_AFTER": [42.62, 39.91, 41.87, 45.85, 40.24, 40.39, 40.2, 39.74, 40.32, 39.74]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.output b/pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.output new file mode 100644 index 0000000..99df8d4 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ASIC_680k_100000.output @@ -0,0 +1,20 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471836 queued and waiting for resources +srun: job 3471836 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 4, ..., 3871767, + 3871770, 3871773]), + col_indices=tensor([ 0, 11698, 11699, ..., 169456, 645874, + 682861]), + values=tensor([ 3.8333e-04, -3.3333e-04, -5.0000e-05, ..., + 0.0000e+00, 0.0000e+00, 7.9289e-02]), + size=(682862, 682862), nnz=3871773, layout=torch.sparse_csr) +tensor([0.9586, 0.4554, 0.3276, ..., 0.2106, 0.5062, 0.3303]) +Matrix: ASIC_680k +Shape: torch.Size([682862, 682862]) +Size: 466300511044 +NNZ: 3871773 +Density: 8.303171256088674e-06 +Time: 77.6055359840393 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.json b/pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.json new file mode 100644 index 0000000..9b1a164 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "Oregon-2", "MATRIX_SHAPE": [11806, 11806], "MATRIX_SIZE": 139381636, "MATRIX_NNZ": 65460, "MATRIX_DENSITY": 0.0004696458003979807, "TIME_S": 4.933578252792358, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [41.99, 39.07, 45.1, 40.09, 40.19, 39.0, 39.86, 40.01, 39.88, 38.97], "POWER": [95.59], "JOULES": 471.60074518442156, "POWER_AFTER": [42.25, 39.42, 39.09, 38.85, 39.96, 39.01, 40.48, 38.81, 39.77, 39.06]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.output b/pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.output new file mode 100644 index 0000000..d802d44 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_Oregon-2_100000.output @@ -0,0 +1,17 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471837 queued and waiting for resources +srun: job 3471837 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 583, 584, ..., 65459, 65460, 65460]), + col_indices=tensor([ 2, 23, 27, ..., 3324, 958, 841]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(11806, 11806), + nnz=65460, layout=torch.sparse_csr) +tensor([0.4829, 0.4475, 0.1256, ..., 0.6137, 0.5875, 0.8973]) +Matrix: Oregon-2 +Shape: torch.Size([11806, 11806]) +Size: 139381636 +NNZ: 65460 +Density: 0.0004696458003979807 +Time: 4.933578252792358 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.json b/pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.json new file mode 100644 index 0000000..5a08e97 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "as-caida", "MATRIX_SHAPE": [31379, 31379], "MATRIX_SIZE": 984641641, "MATRIX_NNZ": 106762, "MATRIX_DENSITY": 0.00010842726485909405, "TIME_S": 7.285882234573364, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.52, 39.15, 38.66, 39.53, 38.32, 39.24, 38.32, 39.16, 38.39, 39.11], "POWER": [99.77], "JOULES": 726.9124705433845, "POWER_AFTER": [40.76, 39.27, 38.94, 41.07, 38.55, 38.52, 38.78, 39.4, 38.59, 39.27]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.output b/pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.output new file mode 100644 index 0000000..81d5172 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_as-caida_100000.output @@ -0,0 +1,18 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471778 queued and waiting for resources +srun: job 3471778 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 0, 0, ..., 106761, 106761, + 106762]), + col_indices=tensor([ 106, 329, 1040, ..., 155, 160, 12170]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(31379, 31379), + nnz=106762, layout=torch.sparse_csr) +tensor([0.1183, 0.9529, 0.6144, ..., 0.4979, 0.4476, 0.7005]) +Matrix: as-caida +Shape: torch.Size([31379, 31379]) +Size: 984641641 +NNZ: 106762 +Density: 0.00010842726485909405 +Time: 7.285882234573364 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_dc2_100000.json b/pytorch/output_test2/epyc_7313p_10_10_dc2_100000.json new file mode 100644 index 0000000..56bc384 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_dc2_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "dc2", "MATRIX_SHAPE": [116835, 116835], "MATRIX_SIZE": 13650417225, "MATRIX_NNZ": 766396, "MATRIX_DENSITY": 5.614451099680581e-05, "TIME_S": 20.959667444229126, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [41.86, 41.11, 41.33, 45.34, 39.89, 39.64, 39.81, 39.46, 39.95, 39.44], "POWER": [130.0], "JOULES": 2724.7567677497864, "POWER_AFTER": [41.62, 39.66, 40.09, 40.08, 39.79, 40.13, 39.9, 39.48, 39.76, 39.46]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_dc2_100000.output b/pytorch/output_test2/epyc_7313p_10_10_dc2_100000.output new file mode 100644 index 0000000..a8105ee --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_dc2_100000.output @@ -0,0 +1,20 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471792 queued and waiting for resources +srun: job 3471792 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 1, 2, ..., 766390, 766394, + 766396]), + col_indices=tensor([ 0, 1, 2, ..., 116833, 89, + 116834]), + values=tensor([-1.0000e+00, -1.0000e+00, -1.0000e+00, ..., + 1.0331e+01, -1.0000e-03, 1.0000e-03]), + size=(116835, 116835), nnz=766396, layout=torch.sparse_csr) +tensor([0.7371, 0.0385, 0.5513, ..., 0.8646, 0.4043, 0.6262]) +Matrix: dc2 +Shape: torch.Size([116835, 116835]) +Size: 13650417225 +NNZ: 766396 +Density: 5.614451099680581e-05 +Time: 20.959667444229126 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_de2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_de2010_100000.json new file mode 100644 index 0000000..b6adaa9 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_de2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "de2010", "MATRIX_SHAPE": [24115, 24115], "MATRIX_SIZE": 581533225, "MATRIX_NNZ": 116056, "MATRIX_DENSITY": 0.0001995689928120616, "TIME_S": 5.716832399368286, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [42.3, 38.99, 39.9, 38.72, 39.6, 38.7, 39.72, 39.07, 40.88, 38.68], "POWER": [102.08], "JOULES": 583.5742513275146, "POWER_AFTER": [41.13, 39.37, 39.29, 39.48, 39.23, 39.31, 38.94, 48.65, 45.1, 39.44]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_de2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_de2010_100000.output new file mode 100644 index 0000000..e843cc2 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_de2010_100000.output @@ -0,0 +1,19 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471785 queued and waiting for resources +srun: job 3471785 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 13, 21, ..., 116047, 116051, + 116056]), + col_indices=tensor([ 250, 251, 757, ..., 23334, 23553, 24050]), + values=tensor([ 14900., 33341., 20255., ..., 164227., 52413., + 16949.]), size=(24115, 24115), nnz=116056, + layout=torch.sparse_csr) +tensor([0.8359, 0.4165, 0.5742, ..., 0.6583, 0.2127, 0.8459]) +Matrix: de2010 +Shape: torch.Size([24115, 24115]) +Size: 581533225 +NNZ: 116056 +Density: 0.0001995689928120616 +Time: 5.716832399368286 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.json b/pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.json new file mode 100644 index 0000000..d492d5b --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "email-Enron", "MATRIX_SHAPE": [36692, 36692], "MATRIX_SIZE": 1346302864, "MATRIX_NNZ": 367662, "MATRIX_DENSITY": 0.0002730901120626302, "TIME_S": 12.811992168426514, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.69, 39.52, 38.51, 39.32, 39.3, 39.55, 38.71, 39.44, 38.55, 39.0], "POWER": [111.47], "JOULES": 1428.1527670145035, "POWER_AFTER": [40.08, 39.68, 38.83, 39.62, 38.73, 39.72, 38.93, 39.63, 38.72, 39.59]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.output b/pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.output new file mode 100644 index 0000000..75a8121 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_email-Enron_100000.output @@ -0,0 +1,18 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471780 queued and waiting for resources +srun: job 3471780 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 1, 71, ..., 367660, 367661, + 367662]), + col_indices=tensor([ 1, 0, 2, ..., 36690, 36689, 8203]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(36692, 36692), + nnz=367662, layout=torch.sparse_csr) +tensor([0.2683, 0.9357, 0.3150, ..., 0.9101, 0.5382, 0.3808]) +Matrix: email-Enron +Shape: torch.Size([36692, 36692]) +Size: 1346302864 +NNZ: 367662 +Density: 0.0002730901120626302 +Time: 12.811992168426514 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.json new file mode 100644 index 0000000..9ce2194 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "fl2010", "MATRIX_SHAPE": [484481, 484481], "MATRIX_SIZE": 234721839361, "MATRIX_NNZ": 2346294, "MATRIX_DENSITY": 9.99606174861054e-06, "TIME_S": 38.57364296913147, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [43.41, 39.71, 40.64, 39.57, 40.01, 39.59, 40.67, 39.73, 41.95, 40.44], "POWER": [146.49], "JOULES": 5650.652958548069, "POWER_AFTER": [43.12, 40.88, 40.74, 46.15, 40.03, 39.97, 39.96, 40.9, 40.37, 40.98]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.output new file mode 100644 index 0000000..fbb1249 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_fl2010_100000.output @@ -0,0 +1,19 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471812 queued and waiting for resources +srun: job 3471812 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 2, 5, ..., 2346288, + 2346292, 2346294]), + col_indices=tensor([ 1513, 5311, 947, ..., 484460, 482463, + 484022]), + values=tensor([28364., 12497., 11567., ..., 8532., 22622., 35914.]), + size=(484481, 484481), nnz=2346294, layout=torch.sparse_csr) +tensor([0.8274, 0.0613, 0.8619, ..., 0.9125, 0.2679, 0.3813]) +Matrix: fl2010 +Shape: torch.Size([484481, 484481]) +Size: 234721839361 +NNZ: 2346294 +Density: 9.99606174861054e-06 +Time: 38.57364296913147 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.json new file mode 100644 index 0000000..40e95da --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "ga2010", "MATRIX_SHAPE": [291086, 291086], "MATRIX_SIZE": 84731059396, "MATRIX_NNZ": 1418056, "MATRIX_DENSITY": 1.6735964475229304e-05, "TIME_S": 33.085010051727295, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [68.04, 68.74, 71.06, 68.14, 65.41, 69.01, 74.03, 67.13, 66.27, 68.83], "POWER": [151.87], "JOULES": 5024.620476555824, "POWER_AFTER": [42.56, 40.18, 39.86, 39.64, 39.71, 40.05, 40.15, 40.33, 39.91, 39.93]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.output new file mode 100644 index 0000000..940efc8 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ga2010_100000.output @@ -0,0 +1,19 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471787 queued and waiting for resources +srun: job 3471787 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 10, ..., 1418047, + 1418054, 1418056]), + col_indices=tensor([ 1566, 1871, 1997, ..., 291064, 289820, + 290176]), + values=tensor([18760., 17851., 18847., ..., 65219., 56729., 77629.]), + size=(291086, 291086), nnz=1418056, layout=torch.sparse_csr) +tensor([0.9989, 0.8982, 0.5822, ..., 0.5453, 0.7727, 0.8878]) +Matrix: ga2010 +Shape: torch.Size([291086, 291086]) +Size: 84731059396 +NNZ: 1418056 +Density: 1.6735964475229304e-05 +Time: 33.085010051727295 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.json b/pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.json new file mode 100644 index 0000000..a0e52b3 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "mac_econ_fwd500", "MATRIX_SHAPE": [206500, 206500], "MATRIX_SIZE": 42642250000, "MATRIX_NNZ": 1273389, "MATRIX_DENSITY": 2.9862143765866013e-05, "TIME_S": 10.857311248779297, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [41.53, 40.0, 40.81, 39.62, 40.1, 39.65, 39.53, 39.92, 39.46, 39.53], "POWER": [155.03], "JOULES": 1683.2089628982544, "POWER_AFTER": [41.01, 40.13, 40.25, 39.99, 39.62, 40.85, 40.86, 45.62, 40.19, 40.22]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.output b/pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.output new file mode 100644 index 0000000..3aa52d7 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_mac_econ_fwd500_100000.output @@ -0,0 +1,20 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471816 queued and waiting for resources +srun: job 3471816 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 8, ..., 1273376, + 1273379, 1273389]), + col_indices=tensor([ 3, 30, 44, ..., 206363, 206408, + 206459]), + values=tensor([-3.7877e-03, -1.5420e-01, 9.5305e-04, ..., + 1.2290e-01, 2.2235e-01, -1.0000e+00]), + size=(206500, 206500), nnz=1273389, layout=torch.sparse_csr) +tensor([0.1058, 0.5873, 0.8242, ..., 0.1574, 0.8351, 0.1537]) +Matrix: mac_econ_fwd500 +Shape: torch.Size([206500, 206500]) +Size: 42642250000 +NNZ: 1273389 +Density: 2.9862143765866013e-05 +Time: 10.857311248779297 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.json b/pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.json new file mode 100644 index 0000000..c57d4b5 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "mc2depi", "MATRIX_SHAPE": [525825, 525825], "MATRIX_SIZE": 276491930625, "MATRIX_NNZ": 2100225, "MATRIX_DENSITY": 7.595972132902821e-06, "TIME_S": 14.108525037765503, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [50.67, 65.78, 62.81, 66.06, 69.62, 62.45, 50.62, 61.89, 58.86, 60.32], "POWER": [159.06], "JOULES": 2244.101992506981, "POWER_AFTER": [41.84, 39.6, 40.57, 39.58, 40.13, 39.78, 40.32, 39.37, 41.46, 39.79]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.output b/pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.output new file mode 100644 index 0000000..7c54e85 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_mc2depi_100000.output @@ -0,0 +1,19 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471789 queued and waiting for resources +srun: job 3471789 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 2, 5, ..., 2100220, + 2100223, 2100225]), + col_indices=tensor([ 0, 1, 1, ..., 525824, 525821, + 525824]), + values=tensor([-2025., 2025., -2026., ..., 2025., 1024., -1024.]), + size=(525825, 525825), nnz=2100225, layout=torch.sparse_csr) +tensor([0.0548, 0.4624, 0.2352, ..., 0.4021, 0.8916, 0.8349]) +Matrix: mc2depi +Shape: torch.Size([525825, 525825]) +Size: 276491930625 +NNZ: 2100225 +Density: 7.595972132902821e-06 +Time: 14.108525037765503 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.json b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.json new file mode 100644 index 0000000..67f66d8 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella04", "MATRIX_SHAPE": [10879, 10879], "MATRIX_SIZE": 118352641, "MATRIX_NNZ": 39994, "MATRIX_DENSITY": 0.0003379223282393842, "TIME_S": 3.682297468185425, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.77, 39.53, 40.23, 39.78, 39.57, 39.96, 39.93, 40.04, 40.05, 39.92], "POWER": [96.71], "JOULES": 356.11498814821243, "POWER_AFTER": [40.97, 39.73, 39.97, 39.94, 39.55, 39.74, 39.76, 40.24, 39.22, 39.1]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.output b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.output new file mode 100644 index 0000000..c6aa4a4 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella04_100000.output @@ -0,0 +1,17 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471809 queued and waiting for resources +srun: job 3471809 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 10, 20, ..., 39994, 39994, 39994]), + col_indices=tensor([ 1, 2, 3, ..., 9711, 10875, 10876]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(10879, 10879), + nnz=39994, layout=torch.sparse_csr) +tensor([0.0521, 0.7363, 0.1682, ..., 0.5599, 0.1291, 0.8935]) +Matrix: p2p-Gnutella04 +Shape: torch.Size([10879, 10879]) +Size: 118352641 +NNZ: 39994 +Density: 0.0003379223282393842 +Time: 3.682297468185425 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.json b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.json new file mode 100644 index 0000000..121d587 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella24", "MATRIX_SHAPE": [26518, 26518], "MATRIX_SIZE": 703204324, "MATRIX_NNZ": 65369, "MATRIX_DENSITY": 9.295875717624285e-05, "TIME_S": 5.905890703201294, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.57, 39.41, 39.78, 39.28, 45.08, 41.07, 39.52, 39.33, 39.72, 39.39], "POWER": [100.57], "JOULES": 593.9554280209541, "POWER_AFTER": [41.15, 39.21, 39.39, 39.22, 39.59, 39.76, 39.22, 39.3, 39.72, 39.16]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.output b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.output new file mode 100644 index 0000000..c23a0b2 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella24_100000.output @@ -0,0 +1,17 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471834 queued and waiting for resources +srun: job 3471834 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 9, 9, ..., 65369, 65369, 65369]), + col_indices=tensor([ 1, 2, 3, ..., 15065, 9401, 26517]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(26518, 26518), + nnz=65369, layout=torch.sparse_csr) +tensor([0.1894, 0.0975, 0.5835, ..., 0.5367, 0.6746, 0.5669]) +Matrix: p2p-Gnutella24 +Shape: torch.Size([26518, 26518]) +Size: 703204324 +NNZ: 65369 +Density: 9.295875717624285e-05 +Time: 5.905890703201294 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.json b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.json new file mode 100644 index 0000000..895385c --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella25", "MATRIX_SHAPE": [22687, 22687], "MATRIX_SIZE": 514699969, "MATRIX_NNZ": 54705, "MATRIX_DENSITY": 0.00010628522108964806, "TIME_S": 5.152007102966309, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [42.83, 40.24, 39.56, 40.23, 39.39, 40.63, 39.47, 40.26, 40.06, 41.04], "POWER": [97.57], "JOULES": 502.6813330364227, "POWER_AFTER": [42.03, 40.41, 39.25, 39.24, 39.63, 39.78, 40.31, 39.31, 39.76, 39.21]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.output b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.output new file mode 100644 index 0000000..ce9222e --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella25_100000.output @@ -0,0 +1,17 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471810 queued and waiting for resources +srun: job 3471810 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 9, 9, ..., 54704, 54704, 54705]), + col_indices=tensor([ 1, 2, 3, ..., 17949, 22685, 144]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(22687, 22687), + nnz=54705, layout=torch.sparse_csr) +tensor([0.2183, 0.6576, 0.4780, ..., 0.0534, 0.0208, 0.5648]) +Matrix: p2p-Gnutella25 +Shape: torch.Size([22687, 22687]) +Size: 514699969 +NNZ: 54705 +Density: 0.00010628522108964806 +Time: 5.152007102966309 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.json b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.json new file mode 100644 index 0000000..2c738d3 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "p2p-Gnutella30", "MATRIX_SHAPE": [36682, 36682], "MATRIX_SIZE": 1345569124, "MATRIX_NNZ": 88328, "MATRIX_DENSITY": 6.564359899804003e-05, "TIME_S": 5.514855861663818, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [41.78, 39.68, 40.37, 39.72, 40.3, 45.14, 41.88, 39.66, 39.77, 39.66], "POWER": [102.52], "JOULES": 565.3830229377746, "POWER_AFTER": [41.48, 40.47, 41.27, 39.44, 39.67, 40.54, 39.51, 40.24, 39.31, 41.6]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.output b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.output new file mode 100644 index 0000000..1505e8b --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_p2p-Gnutella30_100000.output @@ -0,0 +1,17 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471815 queued and waiting for resources +srun: job 3471815 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 10, 10, ..., 88328, 88328, 88328]), + col_indices=tensor([ 1, 2, 3, ..., 36675, 36676, 36677]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(36682, 36682), + nnz=88328, layout=torch.sparse_csr) +tensor([0.5178, 0.0281, 0.3608, ..., 0.6911, 0.2357, 0.6596]) +Matrix: p2p-Gnutella30 +Shape: torch.Size([36682, 36682]) +Size: 1345569124 +NNZ: 88328 +Density: 6.564359899804003e-05 +Time: 5.514855861663818 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.json new file mode 100644 index 0000000..2f4b486 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "ri2010", "MATRIX_SHAPE": [25181, 25181], "MATRIX_SIZE": 634082761, "MATRIX_NNZ": 125750, "MATRIX_DENSITY": 0.00019831796057928155, "TIME_S": 6.2883217334747314, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.5, 38.99, 38.82, 39.05, 38.88, 38.78, 39.03, 39.21, 38.98, 39.37], "POWER": [104.04], "JOULES": 654.2369931507111, "POWER_AFTER": [40.71, 38.8, 39.41, 38.92, 39.31, 38.76, 38.78, 40.12, 39.21, 46.09]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.output new file mode 100644 index 0000000..9a4dc4d --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ri2010_100000.output @@ -0,0 +1,18 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471781 queued and waiting for resources +srun: job 3471781 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 8, ..., 125742, 125747, + 125750]), + col_indices=tensor([ 25, 56, 662, ..., 21738, 22279, 23882]), + values=tensor([17171., 37318., 5284., ..., 25993., 24918., 803.]), + size=(25181, 25181), nnz=125750, layout=torch.sparse_csr) +tensor([0.1942, 0.1978, 0.3462, ..., 0.1743, 0.2436, 0.9955]) +Matrix: ri2010 +Shape: torch.Size([25181, 25181]) +Size: 634082761 +NNZ: 125750 +Density: 0.00019831796057928155 +Time: 6.2883217334747314 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_rma10_100000.json b/pytorch/output_test2/epyc_7313p_10_10_rma10_100000.json new file mode 100644 index 0000000..4014dcb --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_rma10_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "rma10", "MATRIX_SHAPE": [46835, 46835], "MATRIX_SIZE": 2193517225, "MATRIX_NNZ": 2374001, "MATRIX_DENSITY": 0.0010822805369125833, "TIME_S": 38.7850341796875, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.96, 39.06, 40.4, 39.15, 40.82, 44.54, 39.97, 39.79, 39.76, 39.05], "POWER": [136.56], "JOULES": 5296.484267578125, "POWER_AFTER": [42.19, 40.68, 40.19, 40.6, 39.77, 40.74, 39.92, 40.55, 39.61, 40.63]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_rma10_100000.output b/pytorch/output_test2/epyc_7313p_10_10_rma10_100000.output new file mode 100644 index 0000000..7f77177 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_rma10_100000.output @@ -0,0 +1,19 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471808 queued and waiting for resources +srun: job 3471808 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 17, 34, ..., 2373939, + 2373970, 2374001]), + col_indices=tensor([ 0, 1, 2, ..., 46831, 46833, 46834]), + values=tensor([ 1.2636e+05, -1.6615e+07, -8.2015e+04, ..., + 8.3378e+01, 2.5138e+00, 1.2184e+03]), + size=(46835, 46835), nnz=2374001, layout=torch.sparse_csr) +tensor([0.0908, 0.0974, 0.3859, ..., 0.1088, 0.9810, 0.2978]) +Matrix: rma10 +Shape: torch.Size([46835, 46835]) +Size: 2193517225 +NNZ: 2374001 +Density: 0.0010822805369125833 +Time: 38.7850341796875 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.json b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.json new file mode 100644 index 0000000..ae5d036 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "soc-sign-Slashdot090216", "MATRIX_SHAPE": [81871, 81871], "MATRIX_SIZE": 6702860641, "MATRIX_NNZ": 545671, "MATRIX_DENSITY": 8.140867447881048e-05, "TIME_S": 16.565748691558838, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [42.54, 39.13, 39.85, 38.79, 40.21, 38.98, 39.52, 38.92, 39.48, 39.32], "POWER": [122.93], "JOULES": 2036.427486653328, "POWER_AFTER": [45.86, 39.56, 39.69, 40.31, 39.53, 40.55, 39.15, 39.99, 39.13, 39.67]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.output b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.output new file mode 100644 index 0000000..808d59b --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090216_100000.output @@ -0,0 +1,18 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471835 queued and waiting for resources +srun: job 3471835 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 29, 124, ..., 545669, 545669, + 545671]), + col_indices=tensor([ 1, 2, 3, ..., 81869, 81699, 81863]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(81871, 81871), + nnz=545671, layout=torch.sparse_csr) +tensor([0.7693, 0.4110, 0.2737, ..., 0.8913, 0.4051, 0.1845]) +Matrix: soc-sign-Slashdot090216 +Shape: torch.Size([81871, 81871]) +Size: 6702860641 +NNZ: 545671 +Density: 8.140867447881048e-05 +Time: 16.565748691558838 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.json b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.json new file mode 100644 index 0000000..e12f006 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "soc-sign-Slashdot090221", "MATRIX_SHAPE": [82144, 82144], "MATRIX_SIZE": 6747636736, "MATRIX_NNZ": 549202, "MATRIX_DENSITY": 8.13917555860553e-05, "TIME_S": 16.887201070785522, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [39.82, 40.65, 38.66, 38.38, 38.81, 38.83, 38.56, 38.43, 38.56, 38.4], "POWER": [121.66], "JOULES": 2054.4968822717665, "POWER_AFTER": [40.16, 38.88, 38.98, 39.3, 38.76, 38.72, 39.42, 38.78, 39.25, 38.73]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.output b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.output new file mode 100644 index 0000000..19faa39 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-Slashdot090221_100000.output @@ -0,0 +1,18 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471777 queued and waiting for resources +srun: job 3471777 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 29, 124, ..., 549200, 549200, + 549202]), + col_indices=tensor([ 1, 2, 3, ..., 82142, 81974, 82136]), + values=tensor([1., 1., 1., ..., 1., 1., 1.]), size=(82144, 82144), + nnz=549202, layout=torch.sparse_csr) +tensor([0.7544, 0.2107, 0.1548, ..., 0.8853, 0.5512, 0.8288]) +Matrix: soc-sign-Slashdot090221 +Shape: torch.Size([82144, 82144]) +Size: 6747636736 +NNZ: 549202 +Density: 8.13917555860553e-05 +Time: 16.887201070785522 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.json b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.json new file mode 100644 index 0000000..efc26c5 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "soc-sign-epinions", "MATRIX_SHAPE": [131828, 131828], "MATRIX_SIZE": 17378621584, "MATRIX_NNZ": 841372, "MATRIX_DENSITY": 4.841419648464106e-05, "TIME_S": 28.74003553390503, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.92, 39.62, 39.81, 39.63, 39.69, 39.93, 40.13, 39.49, 40.22, 39.73], "POWER": [128.06], "JOULES": 3680.448950471878, "POWER_AFTER": [42.14, 40.44, 40.38, 40.33, 40.04, 39.73, 40.03, 39.89, 39.97, 39.65]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.output b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.output new file mode 100644 index 0000000..7cd92e2 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_soc-sign-epinions_100000.output @@ -0,0 +1,19 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471807 queued and waiting for resources +srun: job 3471807 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 1, 2, ..., 841371, 841371, + 841372]), + col_indices=tensor([ 1, 128552, 3, ..., 131824, 131826, + 7714]), + values=tensor([-1., -1., 1., ..., 1., 1., 1.]), + size=(131828, 131828), nnz=841372, layout=torch.sparse_csr) +tensor([0.8235, 0.7193, 0.6409, ..., 0.8649, 0.4430, 0.4079]) +Matrix: soc-sign-epinions +Shape: torch.Size([131828, 131828]) +Size: 17378621584 +NNZ: 841372 +Density: 4.841419648464106e-05 +Time: 28.74003553390503 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.json b/pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.json new file mode 100644 index 0000000..bcd157a --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "sx-mathoverflow", "MATRIX_SHAPE": [24818, 24818], "MATRIX_SIZE": 615933124, "MATRIX_NNZ": 239978, "MATRIX_DENSITY": 0.00038961697406616504, "TIME_S": 9.806709051132202, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [41.51, 39.84, 40.47, 39.69, 39.5, 39.43, 40.48, 39.36, 40.49, 39.18], "POWER": [110.18], "JOULES": 1080.503203253746, "POWER_AFTER": [46.74, 40.47, 40.85, 39.36, 40.38, 40.43, 40.13, 39.54, 39.53, 39.11]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.output b/pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.output new file mode 100644 index 0000000..b179d74 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_sx-mathoverflow_100000.output @@ -0,0 +1,18 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471793 queued and waiting for resources +srun: job 3471793 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 317, 416, ..., 239976, 239977, + 239978]), + col_indices=tensor([ 0, 1, 2, ..., 1483, 2179, 24817]), + values=tensor([151., 17., 6., ..., 1., 1., 1.]), + size=(24818, 24818), nnz=239978, layout=torch.sparse_csr) +tensor([0.4836, 0.4937, 0.4802, ..., 0.5967, 0.6196, 0.8699]) +Matrix: sx-mathoverflow +Shape: torch.Size([24818, 24818]) +Size: 615933124 +NNZ: 239978 +Density: 0.00038961697406616504 +Time: 9.806709051132202 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.json new file mode 100644 index 0000000..3881548 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "tn2010", "MATRIX_SHAPE": [240116, 240116], "MATRIX_SIZE": 57655693456, "MATRIX_NNZ": 1193966, "MATRIX_DENSITY": 2.070855328296721e-05, "TIME_S": 17.589671850204468, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [42.04, 40.09, 40.08, 39.99, 39.91, 39.54, 39.73, 39.62, 39.97, 40.07], "POWER": [151.86], "JOULES": 2671.167567172051, "POWER_AFTER": [40.91, 40.25, 40.2, 40.04, 40.2, 39.75, 40.04, 39.72, 39.49, 39.74]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.output new file mode 100644 index 0000000..0147812 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_tn2010_100000.output @@ -0,0 +1,21 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471795 queued and waiting for resources +srun: job 3471795 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 4, 20, ..., 1193961, + 1193963, 1193966]), + col_indices=tensor([ 1152, 1272, 1961, ..., 238254, 239142, + 240113]), + values=tensor([ 5728., 2871., 418449., ..., 10058., 33324., + 34928.]), size=(240116, 240116), nnz=1193966, + layout=torch.sparse_csr) +tensor([3.0726e-01, 6.8295e-01, 5.4596e-01, ..., 3.6135e-01, 9.3459e-01, + 3.1018e-04]) +Matrix: tn2010 +Shape: torch.Size([240116, 240116]) +Size: 57655693456 +NNZ: 1193966 +Density: 2.070855328296721e-05 +Time: 17.589671850204468 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.json new file mode 100644 index 0000000..39245b5 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "ut2010", "MATRIX_SHAPE": [115406, 115406], "MATRIX_SIZE": 13318544836, "MATRIX_NNZ": 572066, "MATRIX_DENSITY": 4.295259032005559e-05, "TIME_S": 7.522066831588745, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [41.26, 39.12, 39.52, 39.06, 39.73, 40.19, 40.7, 41.76, 41.47, 45.37], "POWER": [133.71], "JOULES": 1005.7755560517312, "POWER_AFTER": [40.48, 39.63, 40.0, 39.19, 39.77, 39.15, 39.44, 39.69, 39.23, 39.42]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.output new file mode 100644 index 0000000..eb6b60c --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_ut2010_100000.output @@ -0,0 +1,20 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471813 queued and waiting for resources +srun: job 3471813 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 3, 9, ..., 572056, 572061, + 572066]), + col_indices=tensor([ 453, 1291, 1979, ..., 113521, 114509, + 114602]), + values=tensor([160642., 31335., 282373., ..., 88393., 99485., + 18651.]), size=(115406, 115406), nnz=572066, + layout=torch.sparse_csr) +tensor([0.9535, 0.8300, 0.1451, ..., 0.8613, 0.5153, 0.2159]) +Matrix: ut2010 +Shape: torch.Size([115406, 115406]) +Size: 13318544836 +NNZ: 572066 +Density: 4.295259032005559e-05 +Time: 7.522066831588745 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_va2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_va2010_100000.json new file mode 100644 index 0000000..1a8be31 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_va2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "va2010", "MATRIX_SHAPE": [285762, 285762], "MATRIX_SIZE": 81659920644, "MATRIX_NNZ": 1402128, "MATRIX_DENSITY": 1.717033263003816e-05, "TIME_S": 24.332262754440308, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.82, 39.63, 39.96, 39.41, 39.7, 39.59, 39.95, 39.95, 39.67, 39.46], "POWER": [151.38], "JOULES": 3683.4179357671737, "POWER_AFTER": [42.41, 39.81, 39.8, 40.35, 39.68, 39.68, 40.54, 39.6, 39.96, 40.45]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_va2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_va2010_100000.output new file mode 100644 index 0000000..9a57254 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_va2010_100000.output @@ -0,0 +1,20 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471814 queued and waiting for resources +srun: job 3471814 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 2, 8, ..., 1402119, + 1402123, 1402128]), + col_indices=tensor([ 2006, 2464, 1166, ..., 285581, 285634, + 285760]), + values=tensor([125334., 3558., 1192., ..., 10148., 1763., + 9832.]), size=(285762, 285762), nnz=1402128, + layout=torch.sparse_csr) +tensor([0.7314, 0.5884, 0.7739, ..., 0.0933, 0.1510, 0.6060]) +Matrix: va2010 +Shape: torch.Size([285762, 285762]) +Size: 81659920644 +NNZ: 1402128 +Density: 1.717033263003816e-05 +Time: 24.332262754440308 seconds + diff --git a/pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.json b/pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.json new file mode 100644 index 0000000..f761b92 --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.json @@ -0,0 +1 @@ +{"CPU": "EPYC_7313P", "ITERATIONS": 100000, "MATRIX_FILE": "vt2010", "MATRIX_SHAPE": [32580, 32580], "MATRIX_SIZE": 1061456400, "MATRIX_NNZ": 155598, "MATRIX_DENSITY": 0.00014658915806621921, "TIME_S": 7.804270267486572, "BASELINE_TIME_S": 10, "BASELINE_DELAY_S": 10, "POWER_BEFORE": [40.49, 39.46, 39.18, 39.38, 39.2, 39.17, 39.84, 39.64, 39.67, 39.25], "POWER": [106.01], "JOULES": 827.3306910562516, "POWER_AFTER": [40.57, 39.15, 39.36, 39.09, 39.14, 39.45, 39.76, 39.02, 39.12, 44.6]} diff --git a/pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.output b/pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.output new file mode 100644 index 0000000..c31abbe --- /dev/null +++ b/pytorch/output_test2/epyc_7313p_10_10_vt2010_100000.output @@ -0,0 +1,18 @@ +srun: Job time limit was unset; set to partition default of 60 minutes +srun: job 3471811 queued and waiting for resources +srun: job 3471811 has been allocated resources +/nfshomes/vut/ampere_research/pytorch/spmv.py:22: 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.) + ).to_sparse_csr().type(torch.float) +tensor(crow_indices=tensor([ 0, 4, 7, ..., 155588, 155592, + 155598]), + col_indices=tensor([ 131, 561, 996, ..., 32237, 32238, 32570]), + values=tensor([79040., 7820., 15136., ..., 2828., 17986., 2482.]), + size=(32580, 32580), nnz=155598, layout=torch.sparse_csr) +tensor([0.9036, 0.7985, 0.5047, ..., 0.6402, 0.1482, 0.0115]) +Matrix: vt2010 +Shape: torch.Size([32580, 32580]) +Size: 1061456400 +NNZ: 155598 +Density: 0.00014658915806621921 +Time: 7.804270267486572 seconds + diff --git a/pytorch/spmv.py b/pytorch/spmv.py index a367ba7..d063b35 100644 --- a/pytorch/spmv.py +++ b/pytorch/spmv.py @@ -29,6 +29,8 @@ print(vector, file=sys.stderr) start = time.time() for i in range(0, args.iterations): torch.mv(matrix, vector) + #torch.sparse.mm(matrix, vector.unsqueeze(-1)).squeeze(-1) + #print(i) end = time.time() result = dict()