Re: [PATCH] x86: Compile some BF16 tests with -mno-avx512bf16 -mno-avxneconvert
Hongtao Liu <[email protected]>
| Newsgroups | gmane.comp.gcc.patches |
|---|---|
| Message-ID | <CAMZc-bxD2ARo7TuR2wpHG7v6ykxtXT18Uwpy-7mcdOoTJ2oQWQ@mail.gmail.com> |
On Thu, Aug 20, 2026 at 10:50 AM H.J. Lu <[email protected]> wrote: > > Compile some BF16 tests with -mno-avx512bf16 -mno-avxneconvert to avoid > > FAIL: gcc.target/i386/sse2-bfloat16-scalar-typecheck.c (test for excess errors) > FAIL: gcc.target/i386/vect-bfloat16-typecheck_1.c (test for excess errors) > FAIL: gcc.target/i386/vect-bfloat16-typecheck_2.c (test for excess errors) > > with errors like > > sse2-bfloat16-scalar-typecheck.c:16:1: warning: ‘__bfloat16’ is > redefined from typedef ‘short’ to real ‘__bf16’ since GCC 13.1, be > careful of implicit conversion between ‘__bf16’ and ‘short’; an > explicit bitcast may be needed here Ok. > > when AVXNECONVERT or AVX512BF16 are enabled by default. > > PR testsuite/126923 > * gcc.target/i386/sse2-bfloat16-scalar-typecheck.c (dg-options): > Add -mno-avx512bf16 -mno-avxneconvert. > * gcc.target/i386/vect-bfloat16-typecheck_1.c (dg-options): > Likewise. > * gcc.target/i386/vect-bfloat16-typecheck_2.c (dg-options): > Likewise. > > -- > H.J. -- BR, Hongtao