https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103571
--- Comment #27 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Hongtao.liu from comment #17)
> (In reply to Hongtao.liu from comment #16)
> > There're already testcases for vec_extract/vec_set/vec_duplicate, but those
> > testcases are written under TARGET_AVX512FP16, i'll make a copy of them and
> > test them w/o avx512fp16.
>
> Also we can relax condition of extendv*hfv*sf and truncv*sfv*hf to
> avx512vl/f16c so that vect-float16-1.c could be vectorized.
>
> vect-float16-1.c
>
> void
> foo (_Float16 *__restrict__ a, _Float16 *__restrict__ b,
> _Float16 *__restrict__ c)
> {
> for (int i = 0; i < 256; i++)
> a[i] = b[i] + c[i];
> }
This was recently fixed, for -O2 -ftree-vectorize -mfp16c I get:
vpxor %xmm2, %xmm2, %xmm2
vpinsrw $0, (%rsi,%rax), %xmm2, %xmm0
vpinsrw $0, (%rdx,%rax), %xmm2, %xmm1
vcvtph2ps %xmm0, %xmm0
vcvtph2ps %xmm1, %xmm1
vaddss %xmm1, %xmm0, %xmm0
vinsertps $0xe, %xmm0, %xmm0, %xmm0
vcvtps2ph $4, %xmm0, %xmm0
vpextrw $0, %xmm0, (%rdi,%rax)
addq $2, %rax
cmpq $512, %rax
jne .L2
ret
While it would be nice to partially vectorize with vcvtph2ps/vcvtps2ph, the
compiler doesn't reach that far.
