AVX512F/fmaddsub: Implement Instruction

[robinchrist]

First implementation of AVX512 fmaddsub (_mm512_fmaddsub_pd and _mm512_fmaddsub_ps), see #1243

Current status:
No tests yet. This is just to see whether the header itself could be merged style-wise.

[mr-c]

This looks like the correct direction, yes!

Please review the linter / formating results at https://github.com/simd-everywhere/simde/actions/runs/12282344329/job/34273355555?pr=1246

You'll also need an entry for avx512/fmaddsub.h in

simde/simde/x86/avx512.h

Lines 70 to 74 in fa6a869

	#include "avx512/flushsubnormal.h"
	#include "avx512/fmadd.h"
	#include "avx512/fmsub.h"
	#include "avx512/fnmadd.h"
	#include "avx512/fnmsub.h"

See also https://github.com/simd-everywhere/simde/wiki/Implementing-a-New-Function for hints about generating the tests. I can also help you with that.

[robinchrist]

I'll try to add the tests later today

[robinchrist]

This pipeline is massive... But looking good so far. Anything else I need to adjust?

[mr-c]

@robinchrist the CI is looking good! Did you generate the test arrays on a machine that supports AVX512F? If not, let me know and I'll run them on real hardware. The CI uses an emulator from Intel, but I like to double check with real silicon.

[robinchrist]

@mr-c All my AVX512 machines are currently not available, so if possible it would be good if you tested on a CPU with native AVX512. If not possible, I‘ll try to get one of my Cascade Lake machines up and running this weekend

[mr-c]

@robinchrist I ran the tests on a AMD EPYC 9454 using gcc 12.2.0-14 (from Debian) and your tests pass!

[robinchrist]

@mr-c Amazing, thanks a lot! So I guess we're ready to merge now?

[mr-c]

@robinchrist Nearly there! I'll merge after a few more CIs pass