[llvm-bugs] [Bug 24654] New: [x86] consecutive store merging failure

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=24654

            Bug ID: 24654
           Summary: [x86] consecutive store merging failure
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: spatel+llvm at rotateright.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

$ cat merge.ll
define void @merge_8_float_zero_stores(float* %ptr) {
  %idx0 = getelementptr float, float* %ptr, i64 0
  %idx1 = getelementptr float, float* %ptr, i64 1
  %idx2 = getelementptr float, float* %ptr, i64 2
  %idx3 = getelementptr float, float* %ptr, i64 3
  %idx4 = getelementptr float, float* %ptr, i64 4
  %idx5 = getelementptr float, float* %ptr, i64 5
  %idx6 = getelementptr float, float* %ptr, i64 6
  %idx7 = getelementptr float, float* %ptr, i64 7
  store float 0.0, float* %idx0, align 4
  store float 0.0, float* %idx1, align 4
  store float 0.0, float* %idx2, align 4
  store float 0.0, float* %idx3, align 4
  store float 0.0, float* %idx4, align 4
  store float 0.0, float* %idx5, align 4
  store float 0.0, float* %idx6, align 4
  store float 0.0, float* %idx7, align 4
  ret void
}

I think this should be an xor to generate the zero and one 32-byte AVX store,
but:

$ ./llc -o - merge.ll -mattr=avx
...
    movl    $0, (%rdi)
    movl    $0, 4(%rdi)
    movl    $0, 8(%rdi)
    movl    $0, 12(%rdi)
    movl    $0, 16(%rdi)
    movl    $0, 20(%rdi)
    movq    $0, 24(%rdi)    <--- six 4-byte stores and one 8-byte store?!
    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/20150901/2782d443/attachment.html>