[PATCH v2 15/28] x86: Use -mavx instead of -msse2avx

Adhemerval Zanella Netto adhemerval.zanella@linaro.org
Tue Oct 28 20:01:59 GMT 2025



On 28/10/25 16:57, H.J. Lu wrote:
> On Wed, Oct 29, 2025 at 3:54 AM Adhemerval Zanella Netto
> <adhemerval.zanella@linaro.org> wrote:
>>
>>
>>
>> On 28/10/25 16:02, H.J. Lu wrote:
>>> On Wed, Oct 29, 2025 at 1:13 AM Adhemerval Zanella
>>> <adhemerval.zanella@linaro.org> wrote:
>>>>
>>>> clang supports it from version 19 and onwards, but it should be gated
>>>> as an option to assembler (either with -Wa or -Xassembler).  Add a
>>>> test for -Wa as well if -msse2avx fully supported by the compiler.
>>>> ---
>>>>  sysdeps/x86_64/fpu/multiarch/Makefile | 14 +++++++-------
>>>>  1 file changed, 7 insertions(+), 7 deletions(-)
>>>>
>>>> diff --git a/sysdeps/x86_64/fpu/multiarch/Makefile b/sysdeps/x86_64/fpu/multiarch/Makefile
>>>> index 708b14297f..bd887cd59c 100644
>>>> --- a/sysdeps/x86_64/fpu/multiarch/Makefile
>>>> +++ b/sysdeps/x86_64/fpu/multiarch/Makefile
>>>> @@ -150,13 +150,13 @@ CFLAGS-s_sin-fma4.c = -mfma4
>>>>  CFLAGS-s_tan-fma4.c = -mfma4
>>>>  CFLAGS-s_sincos-fma4.c = -mfma4
>>>>
>>>> -CFLAGS-e_atan2-avx.c = -msse2avx -DSSE2AVX
>>>> -CFLAGS-e_exp-avx.c = -msse2avx -DSSE2AVX
>>>> -CFLAGS-e_log-avx.c = -msse2avx -DSSE2AVX
>>>> -CFLAGS-s_atan-avx.c = -msse2avx -DSSE2AVX
>>>> -CFLAGS-s_sin-avx.c = -msse2avx -DSSE2AVX
>>>> -CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
>>>> -CFLAGS-s_sincos-avx.c = -msse2avx -DSSE2AVX
>>>> +CFLAGS-e_atan2-avx.c = -mavx -DSSE2AVX
>>>> +CFLAGS-e_exp-avx.c = -mavx -DSSE2AVX
>>>> +CFLAGS-e_log-avx.c = -mavx -DSSE2AVX
>>>> +CFLAGS-s_atan-avx.c = -mavx -DSSE2AVX
>>>> +CFLAGS-s_sin-avx.c = -mavx -DSSE2AVX
>>>> +CFLAGS-s_tan-avx.c = -mavx -DSSE2AVX
>>>> +CFLAGS-s_sincos-avx.c = -mavx -DSSE2AVX
>>>
>>> -DSSE2AVX shouldn't be necessary.
>>
>> I ended up re-using this on clang optimization for AVX [1].
>>
>> [1] https://patchwork.sourceware.org/project/glibc/patch/20251028171141.3208834-19-adhemerval.zanella@linaro.org/
>>
> 
> "|| defined SSE2AVX" can be removed.
> 

Indeed, I will remove it.


More information about the Libc-alpha mailing list