[llvm-bugs] [Bug 31327] New: [ARM] Extra vmov converting <1 x i64> to i64

[via llvm-bugs]

https://llvm.org/bugs/show_bug.cgi?id=31327

            Bug ID: 31327
           Summary: [ARM] Extra vmov converting <1 x i64> to i64
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: ARM
          Assignee: unassignedbugs at nondot.org
          Reporter: efriedma at codeaurora.org
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

Testcase:

define i64 @func1(<2 x i64> %x) {
entry:
  %shuffle.i = shufflevector <2 x i64> %x, <2 x i64> undef, <1 x i32>
zeroinitializer
  %shuffle.i6 = shufflevector <2 x i64> %x, <2 x i64> undef, <1 x i32> <i32 1>
  %add.i = add <1 x i64> %shuffle.i6, %shuffle.i
  %vget_lane = extractelement <1 x i64> %add.i, i32 0
  ret i64 %vget_lane
}

Current output from llc -mtriple=thumbv7-arm-linux-gnueabi:

        vmov    d17, r2, r3
        vmov    d16, r0, r1
        vadd.i64        d16, d17, d16
        vmov.32 r0, d16[0]
        vmov.32 r1, d16[1]
        bx      lr

The two vmov.32 instructions could be reduced to one "vmov r0, r1, d16".

-- 
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/20161209/c57e91c3/attachment.html>