Intel Adds AVX2/FMA Optimized Math Functions To Glibc 2.27

[phoronix]

Phoronix: Intel Adds AVX2/FMA Optimized Math Functions To Glibc 2.27

Intel engineers have introduced AVX2/FMA-optimized math functions for glibc and will appear in the project's next stable release...

http://www.phoronix.com/scan.php?page=news_item&px=Intel-AVX2-FMA-Math-Glibc-2.27

[wdb974]

Then, Ryzen would benefit from those improvements... right?

[davidbepo]

Then, Ryzen would benefit from those improvements... right?

yes as it has both fma and avx2,
i suspect it will be a smaller improvement because zen's avx implementation is weaker

[eydee]

Then, Ryzen would benefit from those improvements... right?

Unless the code looks for the GenuineIntel string, as usual...

[AsuMagic]

Unless the code looks for the GenuineIntel string, as usual...

But... that's a glibc patch? That's not Intel's job to decide for glibc lead developers, is it?

[wdb974]

But... that's a glibc patch? That's not Intel's job to decide for glibc lead developers, is it?

I believe only ICC looks for the magic string.

[PuckPoltergeist]

I believe only ICC looks for the magic string.

And this doen't affect the optimizations mentioned here. We're speaking about assembler optimizations within glibc, not something the compiler generates.

It isn't mentioned explicitly, but I assume FMA3 here. So for AMD all CPUs since Piledriver will benefit from this, when it comes to FMA optimizations. For AVX2, all CPUs since Excavtor are affected. But, as Intel CPUs are way more optimized for these instructions, they will benefit more than CPUs from AMD. Additionally, we're speaking about special math functions. Only few programs will take benefit from this, and there are already special libraries for this, with highly optimized code.

[szymon_g]

I believe only ICC looks for the magic string.

does it still happen?
also- if my memory serves me well- AMD went to the courts about it- how did it go? TIA

[carewolf]

does it still happen?
also- if my memory serves me well- AMD went to the courts about it- how did it go? TIA

I think Intel settled and paid AMD and continued doing it.

[arjan_intel]

But... that's a glibc patch? That's not Intel's job to decide for glibc lead developers, is it?

well the full source code patch was linked in the article? did you click on it and check for that string?