[llvm-bugs] [Bug 38971] New: [X86] Reductions should use smaller vector types later on

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=38971

            Bug ID: 38971
           Summary: [X86] Reductions should use smaller vector types later
                    on
           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: andrea.dibiagio at gmail.com, craig.topper at gmail.com,
                    llvm-bugs at lists.llvm.org, spatel+llvm at rotateright.com

https://godbolt.org/z/v--jOA

define float @test_v8f32(float %a0, <8 x float> %a1) {
  %1 = call fast float
@llvm.experimental.vector.reduce.fadd.f32.f32.v8f32(float %a0, <8 x float> %a1)
  ret float %1
}
declare float @llvm.experimental.vector.reduce.fadd.f32.f32.v8f32(float, <8 x
float>)

llc -mcpu=btver2

test_v8f32:
  vextractf128 $1, %ymm1, %xmm0
  vaddps %ymm0, %ymm1, %ymm0
  vpermilpd $1, %xmm0, %xmm1 # xmm1 = xmm0[1,0]
  vaddps %ymm1, %ymm0, %ymm0
  vhaddps %ymm0, %ymm0, %ymm0
  retq

All of the vaddps/vhaddps stages could occur using xmm registers, which would
have a notable performance effect on btver2 which double pumps ymm ops as
2*128-bit.

-- 
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/20180916/496bb6f8/attachment-0001.html>