27 lines · plain
1; RUN: llc -global-isel=0 -march=amdgcn -mcpu=tahiti -stop-after=amdgpu-isel -o - %s | FileCheck %s --check-prefixes=GCN,PREGFX12-SDAG2; RUN: llc -global-isel=1 -new-reg-bank-select -march=amdgcn -mcpu=tahiti -stop-after=instruction-select -o - %s | FileCheck %s --check-prefixes=GCN,PREGFX12-GISEL3; RUN: llc -global-isel=0 -march=amdgcn -mcpu=gfx1200 -stop-after=amdgpu-isel -o - %s | FileCheck %s --check-prefixes=GCN,GFX12PLUS-SDAG4; RUN: llc -global-isel=1 -new-reg-bank-select -march=amdgcn -mcpu=gfx1200 -stop-after=instruction-select -o - %s | FileCheck %s --check-prefixes=GCN,GFX12PLUS-GISEL5 6; GCN-LABEL: name: buffer_swizzle_bit_pregfx127; PREGFX12-SDAG: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN {{%[0-9]+}}, killed {{%[0-9]+}}, {{%[0-9]+}}, 0, 0, 1, implicit $exec8; PREGFX12-GISEL: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN {{%[0-9]+}}, {{%[0-9]+}}, {{%[0-9]+}}, 0, 0, 1, implicit $exec9; GFX12PLUS-SDAG: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN killed {{%[0-9]+}}, killed {{%[0-9]+}}, $sgpr_null, 0, 8, 0, implicit $exec10; GFX12PLUS-GISEL: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN {{%[0-9]+}}, {{%[0-9]+}}, $sgpr_null, 0, 8, 0, implicit $exec11define amdgpu_ps <4 x float> @buffer_swizzle_bit_pregfx12(<4 x i32> inreg %0) {12 %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 8)13 ret <4 x float> %data14}15 16; GCN-LABEL: name: buffer_swizzle_bit_gfx12plus17; PREGFX12-SDAG: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN {{%[0-9]+}}, killed {{%[0-9]+}}, {{%[0-9]+}}, 0, 0, 0, implicit $exec18; PREGFX12-GISEL: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_IDXEN {{%[0-9]+}}, {{%[0-9]+}}, {{%[0-9]+}}, 0, 0, 0, implicit $exec19; GFX12PLUS-SDAG: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN killed {{%[0-9]+}}, killed {{%[0-9]+}}, $sgpr_null, 0, 0, 1, implicit $exec20; GFX12PLUS-GISEL: {{%[0-9]+}}:vreg_128 = BUFFER_LOAD_DWORDX4_VBUFFER_IDXEN {{%[0-9]+}}, {{%[0-9]+}}, $sgpr_null, 0, 0, 1, implicit $exec21define amdgpu_ps <4 x float> @buffer_swizzle_bit_gfx12plus(<4 x i32> inreg %0) {22 %data = call <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32> %0, i32 0, i32 0, i32 0, i32 64)23 ret <4 x float> %data24}25 26declare <4 x float> @llvm.amdgcn.struct.buffer.load.v4f32(<4 x i32>, i32, i32, i32, i32)27