1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 11:13:28 +01:00
llvm-mirror/test/CodeGen/X86/vec3.ll
Simon Pilgrim 21418c85a1 [X86][SSE] getV4X86ShuffleImm8 - canonicalize broadcast masks
If the mask input to getV4X86ShuffleImm8 only refers to a single source element (+ undefs) then canonicalize to a full broadcast.

getV4X86ShuffleImm8 defaults to inline values for undefs, which can be useful for shuffle widening/narrowing but does leave SimplifyDemanded* calls thinking the shuffle depends on unnecessary elements.

I'm still investigating what we should do more generally to avoid these undemanded elements, but broadcast cases was a simpler win.
2020-07-29 11:32:44 +01:00

31 lines
1.1 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=sse | FileCheck %s
define <3 x float> @fadd(<3 x float> %v, float %d) {
; CHECK-LABEL: fadd:
; CHECK: # %bb.0:
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
; CHECK-NEXT: addps %xmm1, %xmm0
; CHECK-NEXT: retq
%ins = insertelement <3 x float> undef, float %d, i32 0
%splat = shufflevector <3 x float> %ins, <3 x float> undef, <3 x i32> zeroinitializer
%add = fadd <3 x float> %splat, %v
ret <3 x float> %add
}
; PR29114: https://llvm.org/bugs/show_bug.cgi?id=29114
define <3 x float> @fdiv(<3 x float> %v, float %d) {
; CHECK-LABEL: fdiv:
; CHECK: # %bb.0:
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
; CHECK-NEXT: divps %xmm0, %xmm1
; CHECK-NEXT: movaps %xmm1, %xmm0
; CHECK-NEXT: retq
%ins = insertelement <3 x float> undef, float %d, i32 0
%splat = shufflevector <3 x float> %ins, <3 x float> undef, <3 x i32> zeroinitializer
%div = fdiv <3 x float> %splat, %v
ret <3 x float> %div
}