martin-frbg:issue5372 - Branch - OpenMathLib/OpenBLAS - CodSpeed

Blog Docs Changelog

Work around gcc15.1 on POWER misoptimizing DGEMV at -O3

#5409Merged

Comparing

martin-frbg:issue5372

(

a3b9c93

) with

develop

(

9a64b32

)

0%

Untouched: 62

Benchmarks

62 total

test_dot[1000]

benchmark/pybench/benchmarks/bench_blas.py

+1%

28.4 µs28.2 µs

test_dgbmv[1-100-d]

benchmark/pybench/benchmarks/bench_blas.py

+1%

37.9 µs37.7 µs

test_dgemv[100-s]

benchmark/pybench/benchmarks/bench_blas.py

+1%

103.7 µs103.2 µs

test_dgemv[100-c]

benchmark/pybench/benchmarks/bench_blas.py

+1%

149.7 µs148.9 µs

test_daxpy[100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

24 µs23.9 µs

test_daxpy[100-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

23.9 µs23.8 µs

test_dgemv[100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

141 µs140.4 µs

test_dgbmv[1-100-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

37.4 µs37.2 µs

test_daxpy[100-z]

benchmark/pybench/benchmarks/bench_blas.py

0%

25.8 µs25.6 µs

test_daxpy[1000-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

27.4 µs27.3 µs

test_gesv[100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

395.5 µs394 µs

test_nrm2[100-dz]

benchmark/pybench/benchmarks/bench_blas.py

0%

28.8 µs28.7 µs

test_nrm2[1000-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

30.5 µs30.4 µs

test_dgbmv[1-1000-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

83.7 µs83.4 µs

test_dgbmv[1-1000-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

75.1 µs75 µs

test_nrm2[100-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

36.7 µs36.7 µs

test_daxpy[1000-z]

benchmark/pybench/benchmarks/bench_blas.py

0%

40.5 µs40.4 µs

test_dot[100]

benchmark/pybench/benchmarks/bench_blas.py

0%

22.3 µs22.2 µs

test_gesdd[mn0-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

120 µs119.9 µs

test_dgbmv[1-100-c]

benchmark/pybench/benchmarks/bench_blas.py

0%

40.3 µs40.2 µs

test_daxpy[100-c]

benchmark/pybench/benchmarks/bench_blas.py

0%

25 µs24.9 µs

test_dgbmv[1-1000-c]

benchmark/pybench/benchmarks/bench_blas.py

0%

99.5 µs99.4 µs

test_daxpy[1000-d]

benchmark/pybench/benchmarks/bench_blas.py

0%

32.3 µs32.3 µs

test_dgbmv[1-100-z]

benchmark/pybench/benchmarks/bench_blas.py

0%

42.1 µs42.1 µs

test_gesv[100-s]

benchmark/pybench/benchmarks/bench_blas.py

0%

256.8 µs256.7 µs

Commits

Click on a commit to change the comparison range

Base

develop

9a64b32

+0.12%

mark xbuffer as volatile to work around gcc15.1 optimizer bug

a3b9c93

10 months ago

by martin-frbg

© 2026 CodSpeed Technology

Home Terms Privacy Docs