[Bug rtl-optimization/118662] [14 regression] -ftree-slp-vectorize with -mavx causes incorrect math since r14-9316-g7890836de20912

[ubizjak at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118662

--- Comment #15 from Uroš Bizjak <ubizjak at gmail dot com> ---
The testcase now generates (-O2 -ftree-slp-vectorize -fno-vect-cost-model
-msse4):

addup:
        pmovsxbd        (%rdi), %xmm0
        movd    (%rdi), %xmm1
        movdqa  %xmm0, %xmm2
        pextrb  $3, %xmm1, %edx
        ...

One possible improvement would be to move QImode value to %xmm1 and sign-extend
to %xmm0 from %xmm1. Something like:

addup:
        movd    (%rdi), %xmm1
        pmovsxbd        %xmm1, %xmm0
        movdqa  %xmm0, %xmm2
        pextrb  $3, %xmm1, %edx
        ...

This would save memory read.