| # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py |
| # RUN: llc -march=amdgcn -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX6 %s |
| # RUN: llc -march=amdgcn -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX8 %s |
| # RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s |
| # RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s |
| # RUN: llc -march=amdgcn -mcpu=gfx1100 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s |
| |
| --- |
| name: usubsat_s7 |
| body: | |
| bb.0: |
| liveins: $vgpr0, $vgpr1 |
| |
| ; GFX6-LABEL: name: usubsat_s7 |
| ; GFX6: liveins: $vgpr0, $vgpr1 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 25 |
| ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY1]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6-NEXT: $vgpr0 = COPY [[LSHR]](s32) |
| ; |
| ; GFX8-LABEL: name: usubsat_s7 |
| ; GFX8: liveins: $vgpr0, $vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 9 |
| ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX8-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX8-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; |
| ; GFX9-LABEL: name: usubsat_s7 |
| ; GFX9: liveins: $vgpr0, $vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX9-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 9 |
| ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX9-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| %0:_(s32) = COPY $vgpr0 |
| %1:_(s32) = COPY $vgpr1 |
| %2:_(s7) = G_TRUNC %0 |
| %3:_(s7) = G_TRUNC %1 |
| %4:_(s7) = G_USUBSAT %2, %3 |
| %5:_(s32) = G_ANYEXT %4 |
| $vgpr0 = COPY %5 |
| ... |
| |
| --- |
| name: usubsat_s8 |
| body: | |
| bb.0: |
| liveins: $vgpr0, $vgpr1 |
| |
| ; GFX6-LABEL: name: usubsat_s8 |
| ; GFX6: liveins: $vgpr0, $vgpr1 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 |
| ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY1]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6-NEXT: $vgpr0 = COPY [[LSHR]](s32) |
| ; |
| ; GFX8-LABEL: name: usubsat_s8 |
| ; GFX8: liveins: $vgpr0, $vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX8-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX8-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; |
| ; GFX9-LABEL: name: usubsat_s8 |
| ; GFX9: liveins: $vgpr0, $vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX9-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX9-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| %0:_(s32) = COPY $vgpr0 |
| %1:_(s32) = COPY $vgpr1 |
| %2:_(s8) = G_TRUNC %0 |
| %3:_(s8) = G_TRUNC %1 |
| %4:_(s8) = G_USUBSAT %2, %3 |
| %5:_(s32) = G_ANYEXT %4 |
| $vgpr0 = COPY %5 |
| ... |
| |
| --- |
| name: usubsat_v2s8 |
| body: | |
| bb.0: |
| liveins: $vgpr0, $vgpr1 |
| |
| ; GFX6-LABEL: name: usubsat_v2s8 |
| ; GFX6: liveins: $vgpr0, $vgpr1 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 |
| ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32) |
| ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 |
| ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY]], [[C1]](s32) |
| ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY1]], [[C1]](s32) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C1]](s32) |
| ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C1]](s32) |
| ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C1]](s32) |
| ; GFX6-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C1]](s32) |
| ; GFX6-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) |
| ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C]](s32) |
| ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LSHR3]], [[COPY2]](s32) |
| ; GFX6-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL4]](s32) |
| ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[TRUNC]], [[TRUNC1]] |
| ; GFX6-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) |
| ; GFX6-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; |
| ; GFX8-LABEL: name: usubsat_v2s8 |
| ; GFX8: liveins: $vgpr0, $vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 |
| ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32) |
| ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C1]](s16) |
| ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C1]](s16) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C1]](s16) |
| ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[C1]](s16) |
| ; GFX8-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[TRUNC3]], [[C1]](s16) |
| ; GFX8-NEXT: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL2]], [[SHL3]] |
| ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT1]], [[C1]](s16) |
| ; GFX8-NEXT: [[SHL4:%[0-9]+]]:_(s16) = G_SHL [[LSHR3]], [[C1]](s16) |
| ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[LSHR2]], [[SHL4]] |
| ; GFX8-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) |
| ; GFX8-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; |
| ; GFX9-LABEL: name: usubsat_v2s8 |
| ; GFX9: liveins: $vgpr0, $vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 |
| ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32) |
| ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32) |
| ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[TRUNC1]](s16) |
| ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC2]](s16), [[TRUNC3]](s16) |
| ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[C2]](s16), [[C2]](s16) |
| ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR]], [[BUILD_VECTOR2]](<2 x s16>) |
| ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR1]], [[BUILD_VECTOR2]](<2 x s16>) |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[USUBSAT]], [[BUILD_VECTOR2]](<2 x s16>) |
| ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR2]](<2 x s16>) |
| ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32) |
| ; GFX9-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) |
| ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 |
| ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC4]], [[C3]] |
| ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC5]], [[C3]] |
| ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16) |
| ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]] |
| ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) |
| ; GFX9-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| %0:_(s32) = COPY $vgpr0 |
| %1:_(s32) = COPY $vgpr1 |
| %2:_(s16) = G_TRUNC %0 |
| %3:_(s16) = G_TRUNC %1 |
| %4:_(<2 x s8>) = G_BITCAST %2 |
| %5:_(<2 x s8>) = G_BITCAST %3 |
| %6:_(<2 x s8>) = G_USUBSAT %4, %5 |
| %7:_(s16) = G_BITCAST %6 |
| %8:_(s32) = G_ANYEXT %7 |
| $vgpr0 = COPY %8 |
| ... |
| |
| --- |
| name: usubsat_s16 |
| body: | |
| bb.0: |
| liveins: $vgpr0, $vgpr1 |
| |
| ; GFX6-LABEL: name: usubsat_s16 |
| ; GFX6: liveins: $vgpr0, $vgpr1 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY1]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6-NEXT: $vgpr0 = COPY [[LSHR]](s32) |
| ; |
| ; GFX8-LABEL: name: usubsat_s16 |
| ; GFX8: liveins: $vgpr0, $vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC1]] |
| ; GFX8-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[USUBSAT]](s16) |
| ; GFX8-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; |
| ; GFX9-LABEL: name: usubsat_s16 |
| ; GFX9: liveins: $vgpr0, $vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC1]] |
| ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[USUBSAT]](s16) |
| ; GFX9-NEXT: $vgpr0 = COPY [[ANYEXT]](s32) |
| %0:_(s32) = COPY $vgpr0 |
| %1:_(s32) = COPY $vgpr1 |
| %2:_(s16) = G_TRUNC %0 |
| %3:_(s16) = G_TRUNC %1 |
| %4:_(s16) = G_USUBSAT %2, %3 |
| %5:_(s32) = G_ANYEXT %4 |
| $vgpr0 = COPY %5 |
| ... |
| |
| --- |
| name: usubsat_v2s16 |
| body: | |
| bb.0: |
| liveins: $vgpr0, $vgpr1 |
| |
| ; GFX6-LABEL: name: usubsat_v2s16 |
| ; GFX6: liveins: $vgpr0, $vgpr1 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 |
| ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) |
| ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LSHR3]], [[C]](s32) |
| ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[LSHR2]], [[SHL4]] |
| ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX6-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>) |
| ; |
| ; GFX8-LABEL: name: usubsat_v2s16 |
| ; GFX8: liveins: $vgpr0, $vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 |
| ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) |
| ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC2]] |
| ; GFX8-NEXT: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC1]], [[TRUNC3]] |
| ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT]](s16) |
| ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT1]](s16) |
| ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) |
| ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] |
| ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX8-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>) |
| ; |
| ; GFX9-LABEL: name: usubsat_v2s16 |
| ; GFX9: liveins: $vgpr0, $vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[COPY]], [[COPY1]] |
| ; GFX9-NEXT: $vgpr0 = COPY [[USUBSAT]](<2 x s16>) |
| %0:_(<2 x s16>) = COPY $vgpr0 |
| %1:_(<2 x s16>) = COPY $vgpr1 |
| %2:_(<2 x s16>) = G_USUBSAT %0, %1 |
| $vgpr0 = COPY %2 |
| ... |
| |
| --- |
| name: usubsat_v3s16 |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1_vgpr2 |
| |
| ; GFX6-LABEL: name: usubsat_v3s16 |
| ; GFX6: liveins: $vgpr0_vgpr1_vgpr2 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 |
| ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) |
| ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) |
| ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LSHR2]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN2:%[0-9]+]]:_(s32) = G_UMIN [[SHL4]], [[SHL5]] |
| ; GFX6-NEXT: [[SUB2:%[0-9]+]]:_(s32) = G_SUB [[SHL4]], [[UMIN2]] |
| ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[SUB2]], [[C]](s32) |
| ; GFX6-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>) |
| ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) |
| ; GFX6-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) |
| ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LSHR4]], [[C]](s32) |
| ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[LSHR3]], [[SHL6]] |
| ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 |
| ; GFX6-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]] |
| ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32) |
| ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[LSHR5]], [[SHL7]] |
| ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]] |
| ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) |
| ; GFX6-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[LSHR6]], [[SHL8]] |
| ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) |
| ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) |
| ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) |
| ; |
| ; GFX8-LABEL: name: usubsat_v3s16 |
| ; GFX8: liveins: $vgpr0_vgpr1_vgpr2 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 |
| ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) |
| ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) |
| ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) |
| ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC3]] |
| ; GFX8-NEXT: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC1]], [[TRUNC4]] |
| ; GFX8-NEXT: [[USUBSAT2:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC2]], [[TRUNC5]] |
| ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>) |
| ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) |
| ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) |
| ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT]](s16) |
| ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT1]](s16) |
| ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) |
| ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] |
| ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT2]](s16) |
| ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 |
| ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]] |
| ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32) |
| ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] |
| ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]] |
| ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) |
| ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[LSHR3]], [[SHL2]] |
| ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) |
| ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) |
| ; |
| ; GFX9-LABEL: name: usubsat_v3s16 |
| ; GFX9: liveins: $vgpr0_vgpr1_vgpr2 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 |
| ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) |
| ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) |
| ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) |
| ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF |
| ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC]](s16), [[DEF]](s16) |
| ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC1]](s16), [[TRUNC2]](s16) |
| ; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC3]](s16), [[DEF]](s16) |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[UV]], [[BUILD_VECTOR1]] |
| ; GFX9-NEXT: [[USUBSAT1:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[BUILD_VECTOR]], [[BUILD_VECTOR2]] |
| ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT]](<2 x s16>) |
| ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) |
| ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX9-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) |
| ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT1]](<2 x s16>) |
| ; GFX9-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) |
| ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>) |
| ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) |
| ; GFX9-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32) |
| ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) |
| ; GFX9-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) |
| ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) |
| ; GFX9-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32) |
| ; GFX9-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC4]](s16), [[TRUNC5]](s16) |
| ; GFX9-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC6]](s16), [[TRUNC7]](s16) |
| ; GFX9-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[TRUNC8]](s16), [[TRUNC9]](s16) |
| ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR3]](<2 x s16>), [[BUILD_VECTOR4]](<2 x s16>), [[BUILD_VECTOR5]](<2 x s16>) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) |
| %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 |
| %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0 |
| %3:_(<3 x s16>) = G_USUBSAT %1, %2 |
| %4:_(<3 x s16>) = G_IMPLICIT_DEF |
| %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4 |
| $vgpr0_vgpr1_vgpr2 = COPY %5 |
| ... |
| |
| --- |
| name: usubsat_v4s16 |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| |
| ; GFX6-LABEL: name: usubsat_v4s16 |
| ; GFX6: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 |
| ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) |
| ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX6-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) |
| ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) |
| ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) |
| ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LSHR2]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[BITCAST3]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN2:%[0-9]+]]:_(s32) = G_UMIN [[SHL4]], [[SHL5]] |
| ; GFX6-NEXT: [[SUB2:%[0-9]+]]:_(s32) = G_SUB [[SHL4]], [[UMIN2]] |
| ; GFX6-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[SUB2]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[LSHR3]], [[C]](s32) |
| ; GFX6-NEXT: [[UMIN3:%[0-9]+]]:_(s32) = G_UMIN [[SHL6]], [[SHL7]] |
| ; GFX6-NEXT: [[SUB3:%[0-9]+]]:_(s32) = G_SUB [[SHL6]], [[UMIN3]] |
| ; GFX6-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[SUB3]], [[C]](s32) |
| ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[LSHR5]], [[C]](s32) |
| ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[LSHR4]], [[SHL8]] |
| ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX6-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[LSHR7]], [[C]](s32) |
| ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[LSHR6]], [[SHL9]] |
| ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) |
| ; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) |
| ; |
| ; GFX8-LABEL: name: usubsat_v4s16 |
| ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 |
| ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) |
| ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) |
| ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX8-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) |
| ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) |
| ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) |
| ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) |
| ; GFX8-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) |
| ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX8-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC4]] |
| ; GFX8-NEXT: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC1]], [[TRUNC5]] |
| ; GFX8-NEXT: [[USUBSAT2:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC2]], [[TRUNC6]] |
| ; GFX8-NEXT: [[USUBSAT3:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC3]], [[TRUNC7]] |
| ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT]](s16) |
| ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT1]](s16) |
| ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) |
| ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] |
| ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT2]](s16) |
| ; GFX8-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT3]](s16) |
| ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) |
| ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] |
| ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) |
| ; |
| ; GFX9-LABEL: name: usubsat_v4s16 |
| ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 |
| ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) |
| ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[UV]], [[UV2]] |
| ; GFX9-NEXT: [[USUBSAT1:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[UV1]], [[UV3]] |
| ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[USUBSAT]](<2 x s16>), [[USUBSAT1]](<2 x s16>) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) |
| %0:_(<4 x s16>) = COPY $vgpr0_vgpr1 |
| %1:_(<4 x s16>) = COPY $vgpr2_vgpr3 |
| %2:_(<4 x s16>) = G_USUBSAT %0, %1 |
| $vgpr0_vgpr1 = COPY %2 |
| ... |
| |
| --- |
| name: usubsat_s32 |
| body: | |
| bb.0: |
| liveins: $vgpr0, $vgpr1 |
| |
| ; GFX6-LABEL: name: usubsat_s32 |
| ; GFX6: liveins: $vgpr0, $vgpr1 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[COPY]], [[COPY1]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[COPY]], [[UMIN]] |
| ; GFX6-NEXT: $vgpr0 = COPY [[SUB]](s32) |
| ; |
| ; GFX8-LABEL: name: usubsat_s32 |
| ; GFX8: liveins: $vgpr0, $vgpr1 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[COPY]], [[COPY1]] |
| ; GFX8-NEXT: $vgpr0 = COPY [[USUBSAT]](s32) |
| ; |
| ; GFX9-LABEL: name: usubsat_s32 |
| ; GFX9: liveins: $vgpr0, $vgpr1 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[COPY]], [[COPY1]] |
| ; GFX9-NEXT: $vgpr0 = COPY [[USUBSAT]](s32) |
| %0:_(s32) = COPY $vgpr0 |
| %1:_(s32) = COPY $vgpr1 |
| %2:_(s32) = G_USUBSAT %0, %1 |
| $vgpr0 = COPY %2 |
| ... |
| |
| --- |
| name: usubsat_v2s32 |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| |
| ; GFX6-LABEL: name: usubsat_v2s32 |
| ; GFX6: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 |
| ; GFX6-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) |
| ; GFX6-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) |
| ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[UV]], [[UV2]] |
| ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[UV]], [[UMIN]] |
| ; GFX6-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[UV1]], [[UV3]] |
| ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[UV1]], [[UMIN1]] |
| ; GFX6-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[SUB]](s32), [[SUB1]](s32) |
| ; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) |
| ; |
| ; GFX8-LABEL: name: usubsat_v2s32 |
| ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 |
| ; GFX8-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) |
| ; GFX8-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) |
| ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[UV]], [[UV2]] |
| ; GFX8-NEXT: [[USUBSAT1:%[0-9]+]]:_(s32) = G_USUBSAT [[UV1]], [[UV3]] |
| ; GFX8-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[USUBSAT]](s32), [[USUBSAT1]](s32) |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) |
| ; |
| ; GFX9-LABEL: name: usubsat_v2s32 |
| ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 |
| ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) |
| ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) |
| ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[UV]], [[UV2]] |
| ; GFX9-NEXT: [[USUBSAT1:%[0-9]+]]:_(s32) = G_USUBSAT [[UV1]], [[UV3]] |
| ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[USUBSAT]](s32), [[USUBSAT1]](s32) |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) |
| %0:_(<2 x s32>) = COPY $vgpr0_vgpr1 |
| %1:_(<2 x s32>) = COPY $vgpr2_vgpr3 |
| %2:_(<2 x s32>) = G_USUBSAT %0, %1 |
| $vgpr0_vgpr1 = COPY %2 |
| ... |
| |
| --- |
| name: usubsat_s64 |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| |
| ; GFX6-LABEL: name: usubsat_s64 |
| ; GFX6: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 |
| ; GFX6-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) |
| ; GFX6-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64) |
| ; GFX6-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]] |
| ; GFX6-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]] |
| ; GFX6-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[USUBE1]](s1), [[C]], [[MV]] |
| ; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64) |
| ; |
| ; GFX8-LABEL: name: usubsat_s64 |
| ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 |
| ; GFX8-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) |
| ; GFX8-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64) |
| ; GFX8-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]] |
| ; GFX8-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]] |
| ; GFX8-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[USUBE1]](s1), [[C]], [[MV]] |
| ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64) |
| ; |
| ; GFX9-LABEL: name: usubsat_s64 |
| ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 |
| ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) |
| ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64) |
| ; GFX9-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]] |
| ; GFX9-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]] |
| ; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[USUBE1]](s1), [[C]], [[MV]] |
| ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64) |
| %0:_(s64) = COPY $vgpr0_vgpr1 |
| %1:_(s64) = COPY $vgpr2_vgpr3 |
| %2:_(s64) = G_USUBSAT %0, %1 |
| $vgpr0_vgpr1 = COPY %2 |
| ... |
| |
| --- |
| name: usubsat_v2s64 |
| body: | |
| bb.0: |
| liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6_vgpr7 |
| |
| ; GFX6-LABEL: name: usubsat_v2s64 |
| ; GFX6: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX6-NEXT: {{ $}} |
| ; GFX6-NEXT: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 |
| ; GFX6-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX6-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) |
| ; GFX6-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) |
| ; GFX6-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64) |
| ; GFX6-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64) |
| ; GFX6-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]] |
| ; GFX6-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]] |
| ; GFX6-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX6-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[USUBE1]](s1), [[C]], [[MV]] |
| ; GFX6-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64) |
| ; GFX6-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64) |
| ; GFX6-NEXT: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV8]], [[UV10]] |
| ; GFX6-NEXT: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV9]], [[UV11]], [[USUBO3]] |
| ; GFX6-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32) |
| ; GFX6-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[USUBE3]](s1), [[C]], [[MV1]] |
| ; GFX6-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64) |
| ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) |
| ; |
| ; GFX8-LABEL: name: usubsat_v2s64 |
| ; GFX8: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX8-NEXT: {{ $}} |
| ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 |
| ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX8-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) |
| ; GFX8-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) |
| ; GFX8-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64) |
| ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64) |
| ; GFX8-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]] |
| ; GFX8-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]] |
| ; GFX8-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX8-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[USUBE1]](s1), [[C]], [[MV]] |
| ; GFX8-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64) |
| ; GFX8-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64) |
| ; GFX8-NEXT: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV8]], [[UV10]] |
| ; GFX8-NEXT: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV9]], [[UV11]], [[USUBO3]] |
| ; GFX8-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32) |
| ; GFX8-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[USUBE3]](s1), [[C]], [[MV1]] |
| ; GFX8-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64) |
| ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) |
| ; |
| ; GFX9-LABEL: name: usubsat_v2s64 |
| ; GFX9: liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX9-NEXT: {{ $}} |
| ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 |
| ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) |
| ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) |
| ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64) |
| ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64) |
| ; GFX9-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]] |
| ; GFX9-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]] |
| ; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[USUBE1]](s1), [[C]], [[MV]] |
| ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64) |
| ; GFX9-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64) |
| ; GFX9-NEXT: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV8]], [[UV10]] |
| ; GFX9-NEXT: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV9]], [[UV11]], [[USUBO3]] |
| ; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32) |
| ; GFX9-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[USUBE3]](s1), [[C]], [[MV1]] |
| ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64) |
| ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) |
| %0:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 |
| %1:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 |
| %2:_(<2 x s64>) = G_USUBSAT %0, %1 |
| $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2 |
| ... |