[llvm-bugs] [Bug 34553] New: [X86][AVX512VL] Infinite loop in bool vector sign extension

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=34553

            Bug ID: 34553
           Summary: [X86][AVX512VL] Infinite loop in bool vector sign
                    extension
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: llvm-dev at redking.me.uk
                CC: craig.topper at gmail.com, elena.demikhovsky at intel.com,
                    llvm-bugs at lists.llvm.org, zvi.rackover at intel.com

Found while working on D37680

-mtriple=x86_64-unknown -mattr=+avx512f,+avx512vl

define <16 x i8> @vselect_packss_v16i32(<16 x i32> %a0, <16 x i32> %a1, <16 x
i8> %a2, <16 x i8> %a3) {
  %1 = icmp eq <16 x i32> %a0, %a1
  %2 = sext <16 x i1> %1 to <16 x i8>
  %3 = and <16 x i8> %2, %a2
  %4 = xor <16 x i8> %2, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8
-1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
  %5 = and <16 x i8> %4, %a3
  %6 = or <16 x i8> %3, %5
  ret <16 x i8> %6
}

-mtriple=x86_64-unknown -mattr=+avx512f works OK:

vselect_packss_v16i32:
        vpcmpeqd        %zmm1, %zmm0, %k1
        vpternlogd      $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
        vpmovdb %zmm0, %xmm0
        vpand   %xmm2, %xmm0, %xmm1
        vpandn  %xmm3, %xmm0, %xmm0
        vpor    %xmm0, %xmm1, %xmm0
        vzeroupper
        retq

As does -mtriple=x86_64-unknown -mcpu=skx

vselect_packss_v16i32:
        vpcmpeqd        %zmm1, %zmm0, %k0
        vpmovm2b        %k0, %xmm0
        vpblendvb       %xmm0, %xmm2, %xmm3, %xmm0
        vzeroupper
        retq

-- 
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/20170911/30048c72/attachment.html>