| 1 |
On Mon, Aug 1, 2016 at 5:08 PM, David Seifert <soap@g.o> 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 |
> avx512 - introduced with Skylake and Knights Landing |
| 7 |
|
| 8 |
According to Wikipedia, "AVX-512 consists of multiple extensions not |
| 9 |
all meant to be supported by all processors implementing them." |
| 10 |
|
| 11 |
https://en.wikipedia.org/wiki/AVX-512 |
| 12 |
|
| 13 |
https://en.wikipedia.org/wiki/CPUID#EAX.3D7.2C_ECX.3D0:_Extended_Features |
| 14 |
|
| 15 |
> I also propose creating a CPU_FLAGS_PPC variable containing (at least): |
| 16 |
> |
| 17 |
> altivec - PowerPC's SSE |
| 18 |
|
| 19 |
I see this one exists in use.desc. |
| 20 |
|
| 21 |
> vsx - PowerPC's AVX |
| 22 |
|
| 23 |
I don't see any ebuids using this. |
| 24 |
|
| 25 |
> Finally, I also propose creating CPU_FLAGS_ARM variable containing: |
| 26 |
> |
| 27 |
> neon |
| 28 |
> |
| 29 |
> Any ideas? Currently cpu-feature flags for non-x86 are package local, |
| 30 |
> whereas making them globally available makes masking easier for non- |
| 31 |
> applicable archs etc. and makes enabling them easier for users. |
| 32 |
|
| 33 |
It seems strange to create a USE_EXPAND variable with exactly one value. |
Archives