llvm/llvm/test/CodeGen/AMDGPU/shared-op-cycle.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=r600 -mcpu=redwood | FileCheck %s

define amdgpu_vs void @main(<4 x float> inreg %reg0, <4 x float> inreg %reg1, <4 x float> inreg %reg2) {
; CHECK-LABEL: main:
; CHECK:       ; %bb.0:
; CHECK-NEXT:    CALL_FS
; CHECK-NEXT:    ALU 7, @4, KC0[], KC1[]
; CHECK-NEXT:    EXPORT T0.X___
; CHECK-NEXT:    CF_END
; CHECK-NEXT:    ALU clause starting at 4:
; CHECK-NEXT:     MULADD_IEEE T0.X, T0.W, T0.W, literal.x,
; CHECK-NEXT:     MULADD_IEEE T0.Y, T1.W, T1.W, literal.x, BS:VEC_120/SCL_212
; CHECK-NEXT:     MULADD_IEEE * T0.Z, T2.W, T2.W, literal.x, BS:VEC_201
; CHECK-NEXT:    1073741824(2.000000e+00), 0(0.000000e+00)
; CHECK-NEXT:     DOT4 T0.X, T0.X, T0.X,
; CHECK-NEXT:     DOT4 T0.Y (MASKED), T0.Y, T0.Y,
; CHECK-NEXT:     DOT4 T0.Z (MASKED), T0.Z, T0.Z,
; CHECK-NEXT:     DOT4 * T0.W (MASKED), T0.W, T0.W,
   %w0 = extractelement <4 x float> %reg0, i32 3
   %w1 = extractelement <4 x float> %reg1, i32 3
   %w2 = extractelement <4 x float> %reg2, i32 3
   %sq0 = fmul float %w0, %w0
   %r0 = fadd float %sq0, 2.0
   %sq1 = fmul float %w1, %w1
   %r1 = fadd float %sq1, 2.0
   %sq2 = fmul float %w2, %w2
   %r2 = fadd float %sq2, 2.0
   %v0 = insertelement <4 x float> undef, float %r0, i32 0
   %v1 = insertelement <4 x float> %v0, float %r1, i32 1
   %v2 = insertelement <4 x float> %v1, float %r2, i32 2
   %res = call float @llvm.r600.dot4(<4 x float> %v2, <4 x float> %v2)
   %vecres = insertelement <4 x float> undef, float %res, i32 0
   call void @llvm.r600.store.swizzle(<4 x float> %vecres, i32 0, i32 2)
   ret void
}

; Function Attrs: readnone
declare float @llvm.r600.dot4(<4 x float>, <4 x float>) #1

declare void @llvm.r600.store.swizzle(<4 x float>, i32, i32)

attributes #1 = { readnone }