| 1 |
On Mon, Aug 01, 2016 at 11:08:47PM +0200, David Seifert wrote: |
| 2 |
> Dear friends, |
| 3 |
> while version bumping sci-libs/fftw, I've noticed our CPU_FLAGS_X86 |
| 4 |
> list could be expanded a bit: |
| 5 |
|
| 6 |
I like the idea of expanding the list of CPU_FLAGS_X86, and having flags |
| 7 |
for other arches too, but I would be in favor of dropping the _X86 |
| 8 |
suffix and using CPU_FLAGS for all arches. Your machine most likely will |
| 9 |
only have one kind anyway. Not sure what others think about it, though. |
| 10 |
|
| 11 |
> avx512 - introduced with Skylake and Knights Landing |
| 12 |
|
| 13 |
Only Skylake Xeons (yet to be launched) will support AVX512, and |
| 14 |
as mentioned in another email, the subsets supported by Skylake and |
| 15 |
KNL are a bit different. So, adding each subset should be better |
| 16 |
(e.g. avx512f, avx512cd, avx512er, etc). |
| 17 |
|
| 18 |
> I also propose creating a CPU_FLAGS_PPC variable containing (at least): |
| 19 |
> |
| 20 |
> altivec - PowerPC's SSE |
| 21 |
> vsx - PowerPC's AVX |
| 22 |
> |
| 23 |
> Finally, I also propose creating CPU_FLAGS_ARM variable containing: |
| 24 |
> |
| 25 |
> neon |
| 26 |
> |
| 27 |
> Any ideas? Currently cpu-feature flags for non-x86 are package local, |
| 28 |
> whereas making them globally available makes masking easier for non- |
| 29 |
> applicable archs etc. and makes enabling them easier for users. |
| 30 |
|
| 31 |
Masking flags is not needed if they are determined from /proc/cpuinfo. |
| 32 |
|
| 33 |
Cheers, |
| 34 |
—Guilherme |
Archives