[llvm-bugs] [Bug 42022] New: Failure to merge 2 * <2 x float> load + fadd and then split the result back to 2 * <2 x float>

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=42022

            Bug ID: 42022
           Summary: Failure to merge 2 * <2 x float> load + fadd and then
                    split the result back to 2 * <2 x float>
           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: a.bataev at hotmail.com, anton.a.afanasyev at gmail.com,
                    craig.topper at gmail.com, llvm-bugs at lists.llvm.org,
                    llvm-dev at redking.me.uk, spatel+llvm at rotateright.com

Another missed vectorization opportunity for

struct Vector4 {
    float x, y, z, w;
};

https://godbolt.org/z/lbAN92

// Bad: Failed to combine loads + split result
Vector4 Add(const Vector4 &a, const Vector4 &b) {
    Vector4 r;
    r.x = a.x + b.x;
    r.y = a.y + b.y;
    r.z = a.z + b.z;
    r.w = a.w + b.w;
    return r;
}

_Z3AddRK7Vector4S1_: # @_Z3AddRK7Vector4S1_
  vmovsd (%rdi), %xmm0 # xmm0 = mem[0],zero
  vmovsd (%rsi), %xmm2 # xmm2 = mem[0],zero
  vmovsd 8(%rdi), %xmm1 # xmm1 = mem[0],zero
  vaddps %xmm2, %xmm0, %xmm0
  vmovsd 8(%rsi), %xmm2 # xmm2 = mem[0],zero
  vaddps %xmm2, %xmm1, %xmm1
  retq

Not sure whether this can be handled in the SLP or DAG but we should be able to
do something like:

_Z3AddRK7Vector4S1_: # @_Z3AddRK7Vector4S1_
  vmovups (%rdi), %xmm0
  vaddps (%rsi), %xmm0, %xmm0
  vunpckhpd %xmm0, %xmm0, %xmm1 # xmm1 = xmm0[2,3,2,3]
  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/20190526/4bad5156/attachment.html>