| # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py |
| # RUN: llc -global-isel-abort=0 -march=amdgcn -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX6 %s |
| # RUN: llc -global-isel-abort=0 -march=amdgcn -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX8 %s |
| # RUN: llc -global-isel-abort=0 -march=amdgcn -mcpu=gfx900 -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: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY]](s32) |
| ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32) |
| ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 25 |
| ; GFX6: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32) |
| ; GFX6: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C]](s32) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX6: $vgpr0 = COPY [[COPY4]](s32) |
| ; GFX8-LABEL: name: usubsat_s7 |
| ; GFX8: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 9 |
| ; GFX8: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX8: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX8: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX8: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX8: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; GFX9-LABEL: name: usubsat_s7 |
| ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX9: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 9 |
| ; GFX9: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX9: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX9: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY]](s32) |
| ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32) |
| ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 |
| ; GFX6: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32) |
| ; GFX6: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C]](s32) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX6: $vgpr0 = COPY [[COPY4]](s32) |
| ; GFX8-LABEL: name: usubsat_s8 |
| ; GFX8: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX8: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX8: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX8: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX8: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX8: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; GFX9-LABEL: name: usubsat_s8 |
| ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX9: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX9: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C]](s16) |
| ; GFX9: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C]](s16) |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX9: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C]](s16) |
| ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16) |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 |
| ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32) |
| ; GFX6: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C1]](s32) |
| ; GFX6: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 |
| ; GFX6: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C2]](s32) |
| ; GFX6: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32) |
| ; GFX6: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C1]](s32) |
| ; GFX6: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C2]](s32) |
| ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY]](s32) |
| ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32) |
| ; GFX6: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C2]](s32) |
| ; GFX6: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C2]](s32) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C2]](s32) |
| ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX6: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) |
| ; GFX6: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C2]](s32) |
| ; GFX6: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[COPY5]], [[C2]](s32) |
| ; GFX6: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C2]](s32) |
| ; GFX6: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 |
| ; GFX6: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32) |
| ; GFX6: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C3]] |
| ; GFX6: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C]](s32) |
| ; GFX6: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 255 |
| ; GFX6: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR7]](s32) |
| ; GFX6: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C4]] |
| ; GFX6: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[COPY6]](s32) |
| ; GFX6: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[SHL4]](s32) |
| ; GFX6: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[TRUNC1]] |
| ; GFX6: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) |
| ; GFX6: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; GFX8-LABEL: name: usubsat_v2s8 |
| ; GFX8: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 |
| ; GFX8: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32) |
| ; GFX8: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX8: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C1]](s32) |
| ; GFX8: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 |
| ; GFX8: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C2]](s32) |
| ; GFX8: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32) |
| ; GFX8: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C1]](s32) |
| ; GFX8: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C2]](s32) |
| ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX8: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[C3]](s16) |
| ; GFX8: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[C3]](s16) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX8: [[LSHR6:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT]], [[C3]](s16) |
| ; GFX8: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) |
| ; GFX8: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[C3]](s16) |
| ; GFX8: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[TRUNC3]], [[C3]](s16) |
| ; GFX8: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[SHL2]], [[SHL3]] |
| ; GFX8: [[LSHR7:%[0-9]+]]:_(s16) = G_LSHR [[USUBSAT1]], [[C3]](s16) |
| ; GFX8: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 |
| ; GFX8: [[COPY2:%[0-9]+]]:_(s16) = COPY [[LSHR6]](s16) |
| ; GFX8: [[AND:%[0-9]+]]:_(s16) = G_AND [[COPY2]], [[C4]] |
| ; GFX8: [[COPY3:%[0-9]+]]:_(s16) = COPY [[LSHR7]](s16) |
| ; GFX8: [[AND1:%[0-9]+]]:_(s16) = G_AND [[COPY3]], [[C4]] |
| ; GFX8: [[SHL4:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16) |
| ; GFX8: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL4]] |
| ; GFX8: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) |
| ; GFX8: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; GFX9-LABEL: name: usubsat_v2s8 |
| ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8 |
| ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C]](s32) |
| ; GFX9: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C1]](s32) |
| ; GFX9: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24 |
| ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[COPY]], [[C2]](s32) |
| ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32) |
| ; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C1]](s32) |
| ; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C2]](s32) |
| ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY]](s32) |
| ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32) |
| ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32) |
| ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32) |
| ; GFX9: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8 |
| ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[COPY6]](s32) |
| ; GFX9: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]](<2 x s16>) |
| ; GFX9: [[SHL1:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC2]](<2 x s16>) |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[SHL]], [[SHL1]] |
| ; GFX9: [[LSHR6:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[USUBSAT]], [[BUILD_VECTOR_TRUNC2]](<2 x s16>) |
| ; GFX9: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR6]](<2 x s16>) |
| ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX9: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32) |
| ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32) |
| ; GFX9: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 255 |
| ; GFX9: [[COPY7:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16) |
| ; GFX9: [[AND:%[0-9]+]]:_(s16) = G_AND [[COPY7]], [[C4]] |
| ; GFX9: [[COPY8:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16) |
| ; GFX9: [[AND1:%[0-9]+]]:_(s16) = G_AND [[COPY8]], [[C4]] |
| ; GFX9: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C3]](s16) |
| ; GFX9: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL2]] |
| ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR]](s16) |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[COPY]](s32) |
| ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32) |
| ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32) |
| ; GFX6: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C]](s32) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX6: $vgpr0 = COPY [[COPY4]](s32) |
| ; GFX8-LABEL: name: usubsat_s16 |
| ; GFX8: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC1]] |
| ; GFX8: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[USUBSAT]](s16) |
| ; GFX8: $vgpr0 = COPY [[ANYEXT]](s32) |
| ; GFX9-LABEL: name: usubsat_s16 |
| ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32) |
| ; GFX9: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32) |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC1]] |
| ; GFX9: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[USUBSAT]](s16) |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 |
| ; GFX6: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) |
| ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX6: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) |
| ; GFX6: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) |
| ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) |
| ; GFX6: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32) |
| ; GFX6: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C]](s32) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX6: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) |
| ; GFX6: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C]](s32) |
| ; GFX6: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[COPY5]], [[C]](s32) |
| ; GFX6: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C]](s32) |
| ; GFX6: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 |
| ; GFX6: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) |
| ; GFX6: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY6]], [[C1]] |
| ; GFX6: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) |
| ; GFX6: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C1]] |
| ; GFX6: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) |
| ; GFX6: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL4]] |
| ; GFX6: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX6: $vgpr0 = COPY [[BITCAST2]](<2 x s16>) |
| ; GFX8-LABEL: name: usubsat_v2s16 |
| ; GFX8: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 |
| ; GFX8: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>) |
| ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX8: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX8: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>) |
| ; GFX8: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) |
| ; GFX8: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX8: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC2]] |
| ; GFX8: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC1]], [[TRUNC3]] |
| ; GFX8: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT]](s16) |
| ; GFX8: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT1]](s16) |
| ; GFX8: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) |
| ; GFX8: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] |
| ; GFX8: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX8: $vgpr0 = COPY [[BITCAST2]](<2 x s16>) |
| ; GFX9-LABEL: name: usubsat_v2s16 |
| ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1 |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[COPY]], [[COPY1]] |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 |
| ; GFX6: [[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: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX6: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX6: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX6: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) |
| ; GFX6: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) |
| ; GFX6: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX6: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) |
| ; GFX6: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) |
| ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) |
| ; GFX6: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY1]], [[C]](s32) |
| ; GFX6: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) |
| ; GFX6: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C]](s32) |
| ; GFX6: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C]](s32) |
| ; GFX6: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C]](s32) |
| ; GFX6: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) |
| ; GFX6: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) |
| ; GFX6: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[COPY5]], [[C]](s32) |
| ; GFX6: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[COPY6]], [[C]](s32) |
| ; GFX6: [[UMIN2:%[0-9]+]]:_(s32) = G_UMIN [[SHL4]], [[SHL5]] |
| ; GFX6: [[SUB2:%[0-9]+]]:_(s32) = G_SUB [[SHL4]], [[UMIN2]] |
| ; GFX6: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[SUB2]], [[C]](s32) |
| ; GFX6: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF |
| ; GFX6: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF |
| ; GFX6: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX6: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX6: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) |
| ; GFX6: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) |
| ; GFX6: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) |
| ; GFX6: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) |
| ; GFX6: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) |
| ; GFX6: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 |
| ; GFX6: [[COPY7:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) |
| ; GFX6: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C1]] |
| ; GFX6: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) |
| ; GFX6: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY8]], [[C1]] |
| ; GFX6: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) |
| ; GFX6: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL6]] |
| ; GFX6: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX6: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) |
| ; GFX6: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY9]], [[C1]] |
| ; GFX6: [[COPY10:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) |
| ; GFX6: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY10]], [[C1]] |
| ; GFX6: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) |
| ; GFX6: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL7]] |
| ; GFX6: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX6: [[COPY11:%[0-9]+]]:_(s32) = COPY [[LSHR7]](s32) |
| ; GFX6: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] |
| ; GFX6: [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) |
| ; GFX6: [[AND5:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] |
| ; GFX6: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32) |
| ; GFX6: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL8]] |
| ; GFX6: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) |
| ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>) |
| ; GFX6: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) |
| ; GFX8-LABEL: name: usubsat_v3s16 |
| ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 |
| ; GFX8: [[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: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX8: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX8: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX8: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) |
| ; GFX8: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX8: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) |
| ; GFX8: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) |
| ; GFX8: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX8: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) |
| ; GFX8: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) |
| ; GFX8: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) |
| ; GFX8: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX8: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC3]] |
| ; GFX8: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC1]], [[TRUNC4]] |
| ; GFX8: [[USUBSAT2:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC2]], [[TRUNC5]] |
| ; GFX8: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF |
| ; GFX8: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF |
| ; GFX8: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX8: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX8: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) |
| ; GFX8: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) |
| ; GFX8: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) |
| ; GFX8: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) |
| ; GFX8: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) |
| ; GFX8: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT]](s16) |
| ; GFX8: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT1]](s16) |
| ; GFX8: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) |
| ; GFX8: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] |
| ; GFX8: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX8: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT2]](s16) |
| ; GFX8: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) |
| ; GFX8: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]] |
| ; GFX8: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32) |
| ; GFX8: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] |
| ; GFX8: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX8: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) |
| ; GFX8: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]] |
| ; GFX8: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) |
| ; GFX8: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]] |
| ; GFX8: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32) |
| ; GFX8: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]] |
| ; GFX8: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32) |
| ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>) |
| ; GFX8: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>) |
| ; GFX9-LABEL: name: usubsat_v3s16 |
| ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2 |
| ; GFX9: [[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: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX9: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX9: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX9: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX9: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) |
| ; GFX9: [[COPY2:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[COPY2]](s32) |
| ; GFX9: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) |
| ; GFX9: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF |
| ; GFX9: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY3]](s32), [[DEF]](s32) |
| ; GFX9: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF |
| ; GFX9: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>) |
| ; GFX9: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>) |
| ; GFX9: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX9: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>) |
| ; GFX9: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX9: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) |
| ; GFX9: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32) |
| ; GFX9: [[COPY6:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[DEF]](s32) |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]] |
| ; GFX9: [[USUBSAT1:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]] |
| ; GFX9: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX9: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF |
| ; GFX9: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT]](<2 x s16>) |
| ; GFX9: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32) |
| ; GFX9: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT1]](<2 x s16>) |
| ; GFX9: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32) |
| ; GFX9: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>) |
| ; GFX9: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>) |
| ; GFX9: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32) |
| ; GFX9: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>) |
| ; GFX9: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32) |
| ; GFX9: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST4]](s32) |
| ; GFX9: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32) |
| ; GFX9: [[COPY9:%[0-9]+]]:_(s32) = COPY [[BITCAST5]](s32) |
| ; GFX9: [[COPY10:%[0-9]+]]:_(s32) = COPY [[BITCAST6]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY9]](s32), [[COPY10]](s32) |
| ; GFX9: [[COPY11:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) |
| ; GFX9: [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST7]](s32) |
| ; GFX9: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY11]](s32), [[COPY12]](s32) |
| ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>) |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 |
| ; GFX6: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) |
| ; GFX6: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX6: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX6: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX6: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX6: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX6: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) |
| ; GFX6: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) |
| ; GFX6: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX6: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) |
| ; GFX6: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX6: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32) |
| ; GFX6: [[COPY3:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32) |
| ; GFX6: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32) |
| ; GFX6: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C]](s32) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]] |
| ; GFX6: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32) |
| ; GFX6: [[COPY4:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32) |
| ; GFX6: [[COPY5:%[0-9]+]]:_(s32) = COPY [[LSHR2]](s32) |
| ; GFX6: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C]](s32) |
| ; GFX6: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[COPY5]], [[C]](s32) |
| ; GFX6: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]] |
| ; GFX6: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]] |
| ; GFX6: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C]](s32) |
| ; GFX6: [[COPY6:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32) |
| ; GFX6: [[COPY7:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32) |
| ; GFX6: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[COPY6]], [[C]](s32) |
| ; GFX6: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[COPY7]], [[C]](s32) |
| ; GFX6: [[UMIN2:%[0-9]+]]:_(s32) = G_UMIN [[SHL4]], [[SHL5]] |
| ; GFX6: [[SUB2:%[0-9]+]]:_(s32) = G_SUB [[SHL4]], [[UMIN2]] |
| ; GFX6: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[SUB2]], [[C]](s32) |
| ; GFX6: [[COPY8:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32) |
| ; GFX6: [[COPY9:%[0-9]+]]:_(s32) = COPY [[LSHR3]](s32) |
| ; GFX6: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[COPY8]], [[C]](s32) |
| ; GFX6: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[COPY9]], [[C]](s32) |
| ; GFX6: [[UMIN3:%[0-9]+]]:_(s32) = G_UMIN [[SHL6]], [[SHL7]] |
| ; GFX6: [[SUB3:%[0-9]+]]:_(s32) = G_SUB [[SHL6]], [[UMIN3]] |
| ; GFX6: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[SUB3]], [[C]](s32) |
| ; GFX6: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535 |
| ; GFX6: [[COPY10:%[0-9]+]]:_(s32) = COPY [[LSHR4]](s32) |
| ; GFX6: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY10]], [[C1]] |
| ; GFX6: [[COPY11:%[0-9]+]]:_(s32) = COPY [[LSHR5]](s32) |
| ; GFX6: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY11]], [[C1]] |
| ; GFX6: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32) |
| ; GFX6: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL8]] |
| ; GFX6: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX6: [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR6]](s32) |
| ; GFX6: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY12]], [[C1]] |
| ; GFX6: [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR7]](s32) |
| ; GFX6: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C1]] |
| ; GFX6: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32) |
| ; GFX6: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL9]] |
| ; GFX6: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX6: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) |
| ; GFX6: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) |
| ; GFX8-LABEL: name: usubsat_v4s16 |
| ; GFX8: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 |
| ; GFX8: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) |
| ; GFX8: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>) |
| ; GFX8: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32) |
| ; GFX8: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16 |
| ; GFX8: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32) |
| ; GFX8: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32) |
| ; GFX8: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>) |
| ; GFX8: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32) |
| ; GFX8: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32) |
| ; GFX8: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32) |
| ; GFX8: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) |
| ; GFX8: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>) |
| ; GFX8: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32) |
| ; GFX8: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32) |
| ; GFX8: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32) |
| ; GFX8: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>) |
| ; GFX8: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32) |
| ; GFX8: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32) |
| ; GFX8: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC4]] |
| ; GFX8: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC1]], [[TRUNC5]] |
| ; GFX8: [[USUBSAT2:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC2]], [[TRUNC6]] |
| ; GFX8: [[USUBSAT3:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC3]], [[TRUNC7]] |
| ; GFX8: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT]](s16) |
| ; GFX8: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT1]](s16) |
| ; GFX8: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32) |
| ; GFX8: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]] |
| ; GFX8: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32) |
| ; GFX8: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT2]](s16) |
| ; GFX8: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT3]](s16) |
| ; GFX8: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32) |
| ; GFX8: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]] |
| ; GFX8: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32) |
| ; GFX8: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>) |
| ; GFX8: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>) |
| ; GFX9-LABEL: name: usubsat_v4s16 |
| ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3 |
| ; GFX9: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) |
| ; GFX9: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>) |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[UV]], [[UV2]] |
| ; GFX9: [[USUBSAT1:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[UV1]], [[UV3]] |
| ; GFX9: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[USUBSAT]](<2 x s16>), [[USUBSAT1]](<2 x s16>) |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[COPY]], [[COPY1]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[COPY]], [[UMIN]] |
| ; GFX6: $vgpr0 = COPY [[SUB]](s32) |
| ; GFX8-LABEL: name: usubsat_s32 |
| ; GFX8: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[COPY]], [[COPY1]] |
| ; GFX8: $vgpr0 = COPY [[USUBSAT]](s32) |
| ; GFX9-LABEL: name: usubsat_s32 |
| ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1 |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[COPY]], [[COPY1]] |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 |
| ; GFX6: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) |
| ; GFX6: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) |
| ; GFX6: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[UV]], [[UV2]] |
| ; GFX6: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[UV]], [[UMIN]] |
| ; GFX6: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[UV1]], [[UV3]] |
| ; GFX6: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[UV1]], [[UMIN1]] |
| ; GFX6: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[SUB]](s32), [[SUB1]](s32) |
| ; GFX6: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) |
| ; GFX8-LABEL: name: usubsat_v2s32 |
| ; GFX8: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 |
| ; GFX8: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) |
| ; GFX8: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) |
| ; GFX8: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[UV]], [[UV2]] |
| ; GFX8: [[USUBSAT1:%[0-9]+]]:_(s32) = G_USUBSAT [[UV1]], [[UV3]] |
| ; GFX8: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[USUBSAT]](s32), [[USUBSAT1]](s32) |
| ; GFX8: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>) |
| ; GFX9-LABEL: name: usubsat_v2s32 |
| ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr2_vgpr3 |
| ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>) |
| ; GFX9: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>) |
| ; GFX9: [[USUBSAT:%[0-9]+]]:_(s32) = G_USUBSAT [[UV]], [[UV2]] |
| ; GFX9: [[USUBSAT1:%[0-9]+]]:_(s32) = G_USUBSAT [[UV1]], [[UV3]] |
| ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[USUBSAT]](s32), [[USUBSAT1]](s32) |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 |
| ; GFX6: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) |
| ; GFX6: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64) |
| ; GFX6: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]] |
| ; GFX6: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]] |
| ; GFX6: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX6: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[COPY]](s64), [[COPY1]] |
| ; GFX6: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX6: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[ICMP]](s1), [[C]], [[MV]] |
| ; GFX6: $vgpr0_vgpr1 = COPY [[SELECT]](s64) |
| ; GFX8-LABEL: name: usubsat_s64 |
| ; GFX8: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 |
| ; GFX8: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) |
| ; GFX8: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64) |
| ; GFX8: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]] |
| ; GFX8: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]] |
| ; GFX8: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX8: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[COPY]](s64), [[COPY1]] |
| ; GFX8: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX8: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[ICMP]](s1), [[C]], [[MV]] |
| ; GFX8: $vgpr0_vgpr1 = COPY [[SELECT]](s64) |
| ; GFX9-LABEL: name: usubsat_s64 |
| ; GFX9: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr2_vgpr3 |
| ; GFX9: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s64) |
| ; GFX9: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64) |
| ; GFX9: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]] |
| ; GFX9: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]] |
| ; GFX9: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX9: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[COPY]](s64), [[COPY1]] |
| ; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX9: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[ICMP]](s1), [[C]], [[MV]] |
| ; GFX9: $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: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 |
| ; GFX6: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX6: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) |
| ; GFX6: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) |
| ; GFX6: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64) |
| ; GFX6: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64) |
| ; GFX6: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]] |
| ; GFX6: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]] |
| ; GFX6: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX6: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[UV]](s64), [[UV2]] |
| ; GFX6: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX6: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[ICMP]](s1), [[C]], [[MV]] |
| ; GFX6: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64) |
| ; GFX6: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64) |
| ; GFX6: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV8]], [[UV10]] |
| ; GFX6: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV9]], [[UV11]], [[USUBO3]] |
| ; GFX6: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32) |
| ; GFX6: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[UV1]](s64), [[UV3]] |
| ; GFX6: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[ICMP1]](s1), [[C]], [[MV1]] |
| ; GFX6: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64) |
| ; GFX6: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) |
| ; GFX8-LABEL: name: usubsat_v2s64 |
| ; GFX8: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 |
| ; GFX8: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX8: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) |
| ; GFX8: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) |
| ; GFX8: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64) |
| ; GFX8: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64) |
| ; GFX8: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]] |
| ; GFX8: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]] |
| ; GFX8: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX8: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[UV]](s64), [[UV2]] |
| ; GFX8: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX8: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[ICMP]](s1), [[C]], [[MV]] |
| ; GFX8: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64) |
| ; GFX8: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64) |
| ; GFX8: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV8]], [[UV10]] |
| ; GFX8: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV9]], [[UV11]], [[USUBO3]] |
| ; GFX8: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32) |
| ; GFX8: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[UV1]](s64), [[UV3]] |
| ; GFX8: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[ICMP1]](s1), [[C]], [[MV1]] |
| ; GFX8: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64) |
| ; GFX8: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>) |
| ; GFX9-LABEL: name: usubsat_v2s64 |
| ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3 |
| ; GFX9: [[COPY1:%[0-9]+]]:_(<2 x s64>) = COPY $vgpr4_vgpr5_vgpr6_vgpr7 |
| ; GFX9: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<2 x s64>) |
| ; GFX9: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY1]](<2 x s64>) |
| ; GFX9: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64) |
| ; GFX9: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64) |
| ; GFX9: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]] |
| ; GFX9: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]] |
| ; GFX9: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32) |
| ; GFX9: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[UV]](s64), [[UV2]] |
| ; GFX9: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0 |
| ; GFX9: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[ICMP]](s1), [[C]], [[MV]] |
| ; GFX9: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64) |
| ; GFX9: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64) |
| ; GFX9: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV8]], [[UV10]] |
| ; GFX9: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV9]], [[UV11]], [[USUBO3]] |
| ; GFX9: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32) |
| ; GFX9: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[UV1]](s64), [[UV3]] |
| ; GFX9: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[ICMP1]](s1), [[C]], [[MV1]] |
| ; GFX9: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64) |
| ; GFX9: $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 |
| ... |
