[PATCH] D96231: [X86] Always assign reassoc flag for intrinsics *reduce_add/mul_ps/pd.
Sanjay Patel via Phabricator via cfe-commits
cfe-commits at lists.llvm.org
Mon Feb 8 08:51:14 PST 2021
spatel added inline comments.
================
Comment at: clang/lib/Headers/avx512fintrin.h:9352
static __inline__ double __DEFAULT_FN_ATTRS512 _mm512_reduce_add_pd(__m512d __W) {
return __builtin_ia32_reduce_fadd_pd512(0.0, __W);
}
----------------
Ah - this is where the +0.0 is specified. This should be -0.0. We could still add 'nsz' flag to be safe.
================
Comment at: clang/lib/Headers/avx512fintrin.h:9362
__W = _mm512_maskz_mov_pd(__M, __W);
return __builtin_ia32_reduce_fadd_pd512(0.0, __W);
}
----------------
This also should be changed to -0.0?
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D96231/new/
https://reviews.llvm.org/D96231
More information about the cfe-commits
mailing list