[llvm-bugs] [Bug 33674] New: [AVX-512] select + add can produce better sequence

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=33674

            Bug ID: 33674
           Summary: [AVX-512] select + add can produce better sequence
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: elena.demikhovsky at intel.com
                CC: llvm-bugs at lists.llvm.org

This is the simplified C-code:

    if (B[i] > 1)
      Sum += A[i];

  %8 = load <16 x i32>, <16 x i32>* %7, align 4, !dbg !27, !tbaa !30
  %9 = icmp sgt <16 x i32> %8, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32
1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
  %10 = getelementptr inbounds i32, i32* %0, i64 %4
  %11 = bitcast i32* %10 to <16 x i32>*
  %12 = call <16 x i32> @llvm.masked.load.v16i32.p0v16i32(<16 x i32>* %11, i32
4, <16 x i1> %9, <16 x i32> undef)
  %13 = select <16 x i1> %9, <16 x i32> %12, <16 x i32> zeroinitializer
  %14 = add nsw <16 x i32> %5, %13

This code generates the following sequence:

        vmovdqu32       zmm2, zmmword ptr [rsi + rax]
        vpcmpgtd        k1, zmm2, zmm0
        vmovdqu32       zmm2 {k1} {z}, zmmword ptr [rdi + rax]
        vmovdqa32       zmm2 {k1} {z}, zmm2
        vpaddd  zmm1, zmm1, zmm2

The better sequence:
        vpcmpd    k1, zmm3, ZMMWORD PTR [rsi+rcx*4], 1        
        vpaddd    zmm4{k1}, zmm4, ZMMWORD PTR [rdi+rcx*4]

-- 
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/20170702/f055b99f/attachment.html>