1 ;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s 2 3 ; Exactly one constant vector can be folded into dot4, which means exactly 4 ; 4 MOV instructions 5 ; CHECK: @main 6 ; CHECK: MOV 7 ; CHECK: MOV 8 ; CHECK: MOV 9 ; CHECK: MOV 10 ; CHECK-NOT: MOV 11 ; CHECK-NOT: MOV 12 ; CHECK-NOT: MOV 13 ; CHECK-NOT: MOV 14 15 define void @main(float addrspace(1)* %out) { 16 main_body: 17 %0 = load <4 x float> addrspace(8)* null 18 %1 = load <4 x float> addrspace(8)* getelementptr ([1024 x <4 x float>] addrspace(8)* null, i64 0, i32 1) 19 %2 = call float @llvm.AMDGPU.dp4(<4 x float> %0,<4 x float> %1) 20 %3 = insertelement <4 x float> undef, float %2, i32 0 21 call void @llvm.R600.store.swizzle(<4 x float> %3, i32 0, i32 0) 22 ret void 23 } 24 25 declare float @llvm.AMDGPU.dp4(<4 x float>, <4 x float>) #1 26 declare void @llvm.R600.store.swizzle(<4 x float>, i32, i32) 27 attributes #1 = { readnone } 28
