fix test_mm_dp_pd test

[alexorlov124]

When I ran make check with -O2 passed to ARCH_CFLAGS, the test_mm_dp_pd failed. After debugging, I discovered that a buffer initialized with floats was being cast to a buffer of doubles. The root cause of the test failure was that the exact match of the dot product obtained with intrinsics and calculated with C++ doubles showed differences. This occurred because the numbers were too large, and the dot product had rounding errors.

[howjmay]

I think you use the different version of clang-format. It should be clang-format-12. But you can refer CI directly to format the code

[alexorlov124]

I think you use the different version of clang-format. It should be clang-format-12. But you can refer CI directly to format the code

Thank you, I installed clang-formatter 12 and now format is fixed

[howjmay]

Btw may I ask which compiler and platform you used to trigger the error?

[alexorlov124]

Btw may I ask which compiler and platform you used to trigger the error?

I used gcc 10.2 and 11.4, Graviton3 and Amazon Linux 2023

[howjmay]

Sorry it seems not to pass the CI. May you help me to enable -O2 in the CI and check it ?

https://github.com/howjmay/sse2neon/actions/runs/10019289097

[alexorlov124]

Sorry it seems not to pass the CI. May you help me to enable -O2 in the CI and check it ?

https://github.com/howjmay/sse2neon/actions/runs/10019289097

I updated .github\workflows\main.yml with -O2 flag not sure we should merge it in master but we can make sure tests are green

[alexorlov124]

Here my pull request with optimization enabled #644

[jserv]

Squash the git commits and refine the commit messages.
See https://cbea.ms/git-commit/

[howjmay]

Here my pull request with optimization enabled #644

Thank you for the test. However, it seems the test didn't pass. Not sure if I did something wrong.

I think this PR should be ok to merge after squash. But I just feel strange. It seems this PR can't solve the error when O2 is given

[jserv]

Thank @alexorlov124 for contributing!

[howjmay]

I will investigate what causes O2 fails in #642, since this PR doesnt solve the problem in O2

[alexorlov124]

Welcome, I could participate in investigating other test fails, Right now I need to finish a project in my work. Thank you!