martin-frbg:issue5658 - Branch - OpenMathLib/OpenBLAS - CodSpeed

Blog Docs Changelog

Rewrite the Haswell SROT/DROT kernel tail loop with AVX2 to get consistent FMA rounding

Comparing

martin-frbg:issue5658

(

df29cc0

) with

develop

(

18638c7

)

0%

Untouched: 62

Benchmarks

62 total

test_syrk[100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

341 µs339.5 µs

test_nrm2[1000-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

30.3 µs30.2 µs

test_dgbmv[1-100-c]

benchmark/pybench/benchmarks/bench_blas.py

0%

40.4 µs40.2 µs

test_nrm2[100-dz]

benchmark/pybench/benchmarks/bench_blas.py

0%

28.7 µs28.6 µs

test_syrk[100-c]

benchmark/pybench/benchmarks/bench_blas.py

0%

473.7 µs472.3 µs

test_daxpy[100-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

24.1 µs24 µs

test_daxpy[100-z]

benchmark/pybench/benchmarks/bench_blas.py

0%

25.8 µs25.8 µs

test_daxpy[1000-z]

benchmark/pybench/benchmarks/bench_blas.py

0%

40.5 µs40.5 µs

test_daxpy[1000-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

27.5 µs27.5 µs

test_nrm2[1000-dz]

benchmark/pybench/benchmarks/bench_blas.py

0%

35.3 µs35.2 µs

test_gesv[100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

394.5 µs393.8 µs

test_dgbmv[1-100-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

37.3 µs37.2 µs

test_dot[100]

benchmark/pybench/benchmarks/bench_blas.py

0%

22.5 µs22.4 µs

test_dgbmv[1-1000-c]

benchmark/pybench/benchmarks/bench_blas.py

0%

99.6 µs99.4 µs

test_dgbmv[1-1000-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

83.7 µs83.6 µs

test_dgbmv[1-100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

37.9 µs37.9 µs

test_nrm2[100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

38.4 µs38.4 µs

test_dgemv[100-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

104.1 µs104 µs

test_dgbmv[1-100-z]

benchmark/pybench/benchmarks/bench_blas.py

0%

42.2 µs42.1 µs

test_gesv[100-c]

benchmark/pybench/benchmarks/bench_blas.py

0%

696.2 µs695.8 µs

test_gesv[100-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

256.9 µs256.7 µs

test_dgbmv[1-1000-z]

benchmark/pybench/benchmarks/bench_blas.py

0%

118.9 µs118.8 µs

test_dgemv[1000-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

7 ms7 ms

test_syev[50-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

1.3 ms1.3 ms

test_gesdd[mn0-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

120.1 µs120.1 µs

Commits

Click on a commit to change the comparison range

Base

develop

18638c7

+0.05%

Use AVX2 in the tail loop too for consistent FMA rounding

df29cc0

1 day ago

by martin-frbg

© 2026 CodSpeed Technology

Home Terms Privacy Docs