# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # RUN: llc -march=amdgcn -mcpu=tahiti -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX6 %s # RUN: llc -march=amdgcn -mcpu=fiji -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX8 %s # RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX8 %s # RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX8 %s # RUN: llc -march=amdgcn -mcpu=gfx1100 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX11 %s --- name: trunc_sgpr_v2s32_to_v2s16 legalized: true regBankSelected: true body: | bb.0: liveins: $sgpr0_sgpr1 ; GFX6-LABEL: name: trunc_sgpr_v2s32_to_v2s16 ; GFX6: liveins: $sgpr0_sgpr1 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub0 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub1 ; GFX6-NEXT: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY2]], 16, implicit-def $scc ; GFX6-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 65535 ; GFX6-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[COPY1]], [[S_MOV_B32_]], implicit-def $scc ; GFX6-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32 = S_OR_B32 [[S_LSHL_B32_]], [[S_AND_B32_]], implicit-def $scc ; GFX6-NEXT: S_ENDPGM 0, implicit [[S_OR_B32_]] ; GFX8-LABEL: name: trunc_sgpr_v2s32_to_v2s16 ; GFX8: liveins: $sgpr0_sgpr1 ; GFX8-NEXT: {{ $}} ; GFX8-NEXT: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1 ; GFX8-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub0 ; GFX8-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub1 ; GFX8-NEXT: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY2]], 16, implicit-def $scc ; GFX8-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 65535 ; GFX8-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[COPY1]], [[S_MOV_B32_]], implicit-def $scc ; GFX8-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32 = S_OR_B32 [[S_LSHL_B32_]], [[S_AND_B32_]], implicit-def $scc ; GFX8-NEXT: S_ENDPGM 0, implicit [[S_OR_B32_]] ; GFX11-LABEL: name: trunc_sgpr_v2s32_to_v2s16 ; GFX11: liveins: $sgpr0_sgpr1 ; GFX11-NEXT: {{ $}} ; GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1 ; GFX11-NEXT: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub0 ; GFX11-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub1 ; GFX11-NEXT: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY2]], 16, implicit-def $scc ; GFX11-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 65535 ; GFX11-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[COPY1]], [[S_MOV_B32_]], implicit-def $scc ; GFX11-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32 = S_OR_B32 [[S_LSHL_B32_]], [[S_AND_B32_]], implicit-def $scc ; GFX11-NEXT: S_ENDPGM 0, implicit [[S_OR_B32_]] %0:sgpr(<2 x s32>) = COPY $sgpr0_sgpr1 %1:sgpr(<2 x s16>) = G_TRUNC %0 S_ENDPGM 0, implicit %1 ... --- name: trunc_vgpr_v2s32_to_v2s16 legalized: true regBankSelected: true body: | bb.0: liveins: $vgpr0_vgpr1 ; GFX6-LABEL: name: trunc_vgpr_v2s32_to_v2s16 ; GFX6: liveins: $vgpr0_vgpr1 ; GFX6-NEXT: {{ $}} ; GFX6-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 ; GFX6-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 ; GFX6-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 ; GFX6-NEXT: [[V_LSHLREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 16, [[COPY2]], implicit $exec ; GFX6-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 65535, implicit $exec ; GFX6-NEXT: [[V_AND_B32_e64_:%[0-9]+]]:vgpr_32 = V_AND_B32_e64 [[COPY1]], [[V_MOV_B32_e32_]], implicit $exec ; GFX6-NEXT: [[V_OR_B32_e64_:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_LSHLREV_B32_e64_]], [[V_AND_B32_e64_]], implicit $exec ; GFX6-NEXT: S_ENDPGM 0, implicit [[V_OR_B32_e64_]] ; GFX8-LABEL: name: trunc_vgpr_v2s32_to_v2s16 ; GFX8: liveins: $vgpr0_vgpr1 ; GFX8-NEXT: {{ $}} ; GFX8-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 ; GFX8-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 ; GFX8-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 ; GFX8-NEXT: [[V_MOV_B32_sdwa:%[0-9]+]]:vgpr_32 = V_MOV_B32_sdwa 0, [[COPY2]], 0, 5, 2, 4, implicit $exec, implicit [[COPY1]](tied-def 0) ; GFX8-NEXT: S_ENDPGM 0, implicit [[V_MOV_B32_sdwa]] ; GFX11-LABEL: name: trunc_vgpr_v2s32_to_v2s16 ; GFX11: liveins: $vgpr0_vgpr1 ; GFX11-NEXT: {{ $}} ; GFX11-NEXT: [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1 ; GFX11-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub0 ; GFX11-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[COPY]].sub1 ; GFX11-NEXT: [[V_LSHLREV_B32_e64_:%[0-9]+]]:vgpr_32 = V_LSHLREV_B32_e64 16, [[COPY2]], implicit $exec ; GFX11-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 65535, implicit $exec ; GFX11-NEXT: [[V_AND_B32_e64_:%[0-9]+]]:vgpr_32 = V_AND_B32_e64 [[COPY1]], [[V_MOV_B32_e32_]], implicit $exec ; GFX11-NEXT: [[V_OR_B32_e64_:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_LSHLREV_B32_e64_]], [[V_AND_B32_e64_]], implicit $exec ; GFX11-NEXT: S_ENDPGM 0, implicit [[V_OR_B32_e64_]] %0:vgpr(<2 x s32>) = COPY $vgpr0_vgpr1 %1:vgpr(<2 x s16>) = G_TRUNC %0 S_ENDPGM 0, implicit %1 ...