1; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s --check-prefix=R600-CHECK 2; RUN: llc < %s -march=r600 -mcpu=SI -verify-machineinstrs | FileCheck %s --check-prefix=SI-CHECK 3 4; R600-CHECK: @fsub_f32 5; R600-CHECK: ADD {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].Z, -KC0[2].W 6; SI-CHECK: @fsub_f32 7; SI-CHECK: V_SUB_F32 8define void @fsub_f32(float addrspace(1)* %out, float %a, float %b) { 9entry: 10 %0 = fsub float %a, %b 11 store float %0, float addrspace(1)* %out 12 ret void 13} 14 15declare float @llvm.R600.load.input(i32) readnone 16 17declare void @llvm.AMDGPU.store.output(float, i32) 18 19; R600-CHECK: @fsub_v2f32 20; R600-CHECK-DAG: ADD {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[3].X, -KC0[3].Z 21; R600-CHECK-DAG: ADD {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].W, -KC0[3].Y 22; SI-CHECK: @fsub_v2f32 23; SI-CHECK: V_SUB_F32 24; SI-CHECK: V_SUB_F32 25define void @fsub_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) { 26entry: 27 %0 = fsub <2 x float> %a, %b 28 store <2 x float> %0, <2 x float> addrspace(1)* %out 29 ret void 30} 31 32; R600-CHECK: @fsub_v4f32 33; R600-CHECK: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}} 34; R600-CHECK: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}} 35; R600-CHECK: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}} 36; R600-CHECK: ADD {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW], -T[0-9]+\.[XYZW]}} 37; SI-CHECK: @fsub_v4f32 38; SI-CHECK: V_SUB_F32 39; SI-CHECK: V_SUB_F32 40; SI-CHECK: V_SUB_F32 41; SI-CHECK: V_SUB_F32 42define void @fsub_v4f32(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) { 43 %b_ptr = getelementptr <4 x float> addrspace(1)* %in, i32 1 44 %a = load <4 x float> addrspace(1) * %in 45 %b = load <4 x float> addrspace(1) * %b_ptr 46 %result = fsub <4 x float> %a, %b 47 store <4 x float> %result, <4 x float> addrspace(1)* %out 48 ret void 49} 50