[llvm-bugs] [Bug 47572] New: Better codegen for AVX512 scatter

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=47572

            Bug ID: 47572
           Summary: Better codegen for AVX512 scatter
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: david.bolvansky at gmail.com
                CC: craig.topper at gmail.com, llvm-bugs at lists.llvm.org,
                    llvm-dev at redking.me.uk, spatel+llvm at rotateright.com

#include <immintrin.h>
void sct(int *data) {
  _mm512_mask_i32scatter_epi32(data - 1, 0xffff, _mm512_set1_epi32(1),
_mm512_set1_epi32(1), 1);
}

-O3 -march=skylake-avx512


Clang:
.LCPI0_0:
        .long   1                               # 0x1
sct(int*):                               # @sct(int*)
        add     rdi, -4
        kxnorw  k1, k0, k0
        vpbroadcastd    zmm0, dword ptr [rip + .LCPI0_0] # zmm0 =
[1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
        vpscatterdd     zmmword ptr [rdi + zmm0] {k1}, zmm0
        vzeroupper
        ret


ICC:
sct(int*):
        vmovups   zmm0, ZMMWORD PTR .L_2il0floatpacket.0[rip]   #3.3
        vpcmpeqb  k1, xmm0, xmm0                                #3.3
        vpscatterdd DWORD PTR [-4+rdi+zmm0]{k1}, zmm0           #3.3
        vzeroupper                                              #4.1
        ret                                                     #4.1
.L_2il0floatpacket.0:
        .long  
0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001,0x00000001

https://godbolt.org/z/xKbePG

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20200918/513c22f0/attachment.html>