READ-ONLY FORGE CAMPAIGN

NPBench Jacobi 2D

Historical campaign evidence projected from its recorded ledger and checked Accelerator bundle.

01 / WORKLOAD + REFERENCE

npbench.jacobi_2d.m

array_hashes
["06662ca7642f1952e510b3e97bb0c0f54344fa9432894bdefedd7406a784c8ee","51690f111f92b8933b08de46c8ff0ac8ed077b647af0e3256adb79dee9bf5280"]
array_shapes
[[350,350],[350,350]]
atol
1e-8
benchmark
jacobi_2d
dtype
float64
norm_error
0.00001
output_args
["A","B"]
parameters
{"N":350,"TSTEPS":80}
preset
M
rtol
0.00001
suite
NPBench

Reference: npbench.jacobi_2d.numpy vf2d7f27d87ba4881991e62c18a607aa6ba52b260 · Reference source

git:f2d7f27d87ba4881991e62c18a607aa6ba52b260
Reference invocation
NPBench pinned NumPy kernel with M preset; reset mutated input arrays before each call
02 / BASELINE + PROFILE

94.75 ms median

1 warm-up · 5 measured calls · perf_counter_ns

Raw reference samples (ns): 93323354 · 97585654 · 94751804 · 98080185 · 93476721

  1. campaign.py:<lambda> · 1 calls · 0.093 s cumulative
  2. campaign.py:run · 1 calls · 0.093 s cumulative
  3. jacobi_2d_numpy.py:kernel · 1 calls · 0.093 s cumulative
  4. ~:<method 'disable' of '_lsprof.Profiler' objects> · 1 calls · 0.000 s cumulative
  5. campaign.py:_np_outputs · 1 calls · 0.000 s cumulative

linux · Python 3.12.12 · NVIDIA A100-SXM4-40GB · 40960 MiB · driver 550.54.15 · raddle 0.3.0 · raddle_bench 0.1.0 · numpy 2.5.3 · torch 2.8.0+cu128 · torch_cuda_runtime 12.8

03 / EXPERIMENTS

Candidate history

  1. phase-c-001-npbench.jacobi_2d.maccepted

    Source revision: sha256:342c861c2f9d10da576d319ebe8427f8f96f9542726130d69e97129b2a8c60af

    Source SHA-256: 342c861c2f9d10da576d319ebe8427f8f96f9542726130d69e97129b2a8c60af

    Build: passed

    Validation: matched · 2 outputs checked · max abs 0 · max rel 0 · npbench.numpy.v1

    Benchmark median 24.21 ms · raw ns: 24379364 · 24719719 · 24209097 · 23956734 · 24034313

    Confirmed median 24.24 ms · raw ns: 24566588 · 24505539 · 24244062 · 23940414 · 23890016

  2. phase-c-002-npbench.jacobi_2d.mbuild_failed

    Source revision: sha256:342c861c2f9d10da576d319ebe8427f8f96f9542726130d69e97129b2a8c60af

    Source SHA-256:

    Build: failed

    Validation: not run

    candidate initialization or JIT failed

  3. phase-c-003-npbench.jacobi_2d.maccepted

    Source revision: sha256:3321418832a62e17d8ae348863a9dda63444d559f7a9a22bc2ac84d466fc6c68

    Source SHA-256: 3321418832a62e17d8ae348863a9dda63444d559f7a9a22bc2ac84d466fc6c68

    Build: passed

    Validation: matched · 2 outputs checked · max abs 4.263256414560601e-13 · max rel 5.883400614287084e-16 · npbench.numpy.v1

    Benchmark median 4.13 ms · raw ns: 4456950 · 4120456 · 4126018 · 4123707 · 4142555

    Confirmed median 4.10 ms · raw ns: 4048467 · 4100603 · 4101346 · 4097130 · 4195230

04 / INCUMBENT + ARTIFACT

Experiment phase-c-003-npbench.jacobi_2d.m

npbench.jacobi_2d.numpy → raddle_bench.npbench.cupy.elementwise.explicit_size · 94.75 ms → 4.10 ms · 23.11× end-to-end (host input → GPU → host output)

Transfers: 1.05 ms · 25.4% of end-to-end time · device-only 2.37 ms (secondary)

Validation: matched · 2 mutated outputs checked

Artifact: external.npbench.jacobi_2d.m v0.1.0 · readback verified

Artifact identity SHA-256: 52928f6bccab52db027bb580373b4296fe926726a2212522d6a22bd9ba85be7d

Manifest SHA-256: cc17c3c073646e1b8ad68d876e62c15bbde6a1eafd3cb03fc8b0b4602f2c5ac5

Wheel SHA-256: ab71dc9f340bb302aa7b3656321d842ff754f02129dca9346f321b34b0ee43b4

Benchmark SHA-256: fddb8510833cddae691bfddc7d1a21a16a920c55d3811a7aebd2e8fe626a12b3

Checked artifact bundle receipt (separate run): 93.54 ms → 4.58 ms · 20.41× · end_to_end

Provenance: Linux · Python 3.12.12 · NPBench f2d7f27d87ba4881991e62c18a607aa6ba52b260 · cupy-cuda12x 13.6.0 · numpy 2.5.3 · nvidia-cublas-cu12 12.4.5.8 · raddle 0.3.0 · raddle-bench 0.1.0