Index: llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp =================================================================== --- llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp +++ llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp @@ -1315,6 +1315,7 @@ case TargetOpcode::G_FLOG10: case TargetOpcode::G_FLOG: case TargetOpcode::G_FLOG2: + case TargetOpcode::G_FNEARBYINT: case TargetOpcode::G_FSQRT: case TargetOpcode::G_FEXP: case TargetOpcode::G_FEXP2: @@ -2181,6 +2182,7 @@ case G_FLOG: case G_FLOG2: case G_FLOG10: + case G_FNEARBYINT: case G_FCEIL: case G_FFLOOR: case G_INTRINSIC_ROUND: Index: llvm/lib/Target/AArch64/AArch64LegalizerInfo.cpp =================================================================== --- llvm/lib/Target/AArch64/AArch64LegalizerInfo.cpp +++ llvm/lib/Target/AArch64/AArch64LegalizerInfo.cpp @@ -130,7 +130,8 @@ getActionDefinitionsBuilder(G_FREM).libcallFor({s32, s64}); - getActionDefinitionsBuilder({G_FCEIL, G_FABS, G_FSQRT, G_FFLOOR}) + getActionDefinitionsBuilder( + {G_FCEIL, G_FABS, G_FSQRT, G_FFLOOR, G_FNEARBYINT}) // If we don't have full FP16 support, then scalarize the elements of // vectors containing fp16 types. .fewerElementsIf( Index: llvm/lib/Target/AArch64/AArch64RegisterBankInfo.cpp =================================================================== --- llvm/lib/Target/AArch64/AArch64RegisterBankInfo.cpp +++ llvm/lib/Target/AArch64/AArch64RegisterBankInfo.cpp @@ -395,6 +395,7 @@ case TargetOpcode::G_FPTRUNC: case TargetOpcode::G_FCEIL: case TargetOpcode::G_FFLOOR: + case TargetOpcode::G_FNEARBYINT: case TargetOpcode::G_FNEG: case TargetOpcode::G_FCOS: case TargetOpcode::G_FSIN: Index: llvm/test/CodeGen/AArch64/GlobalISel/legalize-nearbyint.mir =================================================================== --- /dev/null +++ llvm/test/CodeGen/AArch64/GlobalISel/legalize-nearbyint.mir @@ -0,0 +1,234 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple aarch64-unknown-unknown -run-pass=legalizer -mattr=+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=FP16 +# RUN: llc -mtriple aarch64-unknown-unknown -run-pass=legalizer -mattr=-fullfp16 -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=NO-FP16 + +... +--- +name: test_v4f16.nearbyint +alignment: 2 +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; FP16-LABEL: name: test_v4f16.nearbyint + ; FP16: liveins: $d0 + ; FP16: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $d0 + ; FP16: [[FNEARBYINT:%[0-9]+]]:_(<4 x s16>) = G_FNEARBYINT [[COPY]] + ; FP16: $d0 = COPY [[FNEARBYINT]](<4 x s16>) + ; FP16: RET_ReallyLR implicit $d0 + ; NO-FP16-LABEL: name: test_v4f16.nearbyint + ; NO-FP16: liveins: $d0 + ; NO-FP16: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $d0 + ; NO-FP16: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<4 x s16>) + ; NO-FP16: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16) + ; NO-FP16: [[FNEARBYINT:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT]] + ; NO-FP16: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT]](s32) + ; NO-FP16: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16) + ; NO-FP16: [[FNEARBYINT1:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT1]] + ; NO-FP16: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT1]](s32) + ; NO-FP16: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16) + ; NO-FP16: [[FNEARBYINT2:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT2]] + ; NO-FP16: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT2]](s32) + ; NO-FP16: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16) + ; NO-FP16: [[FNEARBYINT3:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT3]] + ; NO-FP16: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT3]](s32) + ; NO-FP16: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16), [[FPTRUNC2]](s16), [[FPTRUNC3]](s16) + ; NO-FP16: $d0 = COPY [[BUILD_VECTOR]](<4 x s16>) + ; NO-FP16: RET_ReallyLR implicit $d0 + %0:_(<4 x s16>) = COPY $d0 + %1:_(<4 x s16>) = G_FNEARBYINT %0 + $d0 = COPY %1(<4 x s16>) + RET_ReallyLR implicit $d0 + +... +--- +name: test_v8f16.nearbyint +alignment: 2 +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $q0 + + ; FP16-LABEL: name: test_v8f16.nearbyint + ; FP16: liveins: $q0 + ; FP16: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0 + ; FP16: [[FNEARBYINT:%[0-9]+]]:_(<8 x s16>) = G_FNEARBYINT [[COPY]] + ; FP16: $q0 = COPY [[FNEARBYINT]](<8 x s16>) + ; FP16: RET_ReallyLR implicit $q0 + ; NO-FP16-LABEL: name: test_v8f16.nearbyint + ; NO-FP16: liveins: $q0 + ; NO-FP16: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $q0 + ; NO-FP16: [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[COPY]](<8 x s16>) + ; NO-FP16: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[UV]](s16) + ; NO-FP16: [[FNEARBYINT:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT]] + ; NO-FP16: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT]](s32) + ; NO-FP16: [[FPEXT1:%[0-9]+]]:_(s32) = G_FPEXT [[UV1]](s16) + ; NO-FP16: [[FNEARBYINT1:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT1]] + ; NO-FP16: [[FPTRUNC1:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT1]](s32) + ; NO-FP16: [[FPEXT2:%[0-9]+]]:_(s32) = G_FPEXT [[UV2]](s16) + ; NO-FP16: [[FNEARBYINT2:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT2]] + ; NO-FP16: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT2]](s32) + ; NO-FP16: [[FPEXT3:%[0-9]+]]:_(s32) = G_FPEXT [[UV3]](s16) + ; NO-FP16: [[FNEARBYINT3:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT3]] + ; NO-FP16: [[FPTRUNC3:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT3]](s32) + ; NO-FP16: [[FPEXT4:%[0-9]+]]:_(s32) = G_FPEXT [[UV4]](s16) + ; NO-FP16: [[FNEARBYINT4:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT4]] + ; NO-FP16: [[FPTRUNC4:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT4]](s32) + ; NO-FP16: [[FPEXT5:%[0-9]+]]:_(s32) = G_FPEXT [[UV5]](s16) + ; NO-FP16: [[FNEARBYINT5:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT5]] + ; NO-FP16: [[FPTRUNC5:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT5]](s32) + ; NO-FP16: [[FPEXT6:%[0-9]+]]:_(s32) = G_FPEXT [[UV6]](s16) + ; NO-FP16: [[FNEARBYINT6:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT6]] + ; NO-FP16: [[FPTRUNC6:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT6]](s32) + ; NO-FP16: [[FPEXT7:%[0-9]+]]:_(s32) = G_FPEXT [[UV7]](s16) + ; NO-FP16: [[FNEARBYINT7:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT7]] + ; NO-FP16: [[FPTRUNC7:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT7]](s32) + ; NO-FP16: [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s16>) = G_BUILD_VECTOR [[FPTRUNC]](s16), [[FPTRUNC1]](s16), [[FPTRUNC2]](s16), [[FPTRUNC3]](s16), [[FPTRUNC4]](s16), [[FPTRUNC5]](s16), [[FPTRUNC6]](s16), [[FPTRUNC7]](s16) + ; NO-FP16: $q0 = COPY [[BUILD_VECTOR]](<8 x s16>) + ; NO-FP16: RET_ReallyLR implicit $q0 + %0:_(<8 x s16>) = COPY $q0 + %1:_(<8 x s16>) = G_FNEARBYINT %0 + $q0 = COPY %1(<8 x s16>) + RET_ReallyLR implicit $q0 + +... +--- +name: test_v2f32.nearbyint +alignment: 2 +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; FP16-LABEL: name: test_v2f32.nearbyint + ; FP16: liveins: $d0 + ; FP16: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0 + ; FP16: [[FNEARBYINT:%[0-9]+]]:_(<2 x s32>) = G_FNEARBYINT [[COPY]] + ; FP16: $d0 = COPY [[FNEARBYINT]](<2 x s32>) + ; FP16: RET_ReallyLR implicit $d0 + ; NO-FP16-LABEL: name: test_v2f32.nearbyint + ; NO-FP16: liveins: $d0 + ; NO-FP16: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0 + ; NO-FP16: [[FNEARBYINT:%[0-9]+]]:_(<2 x s32>) = G_FNEARBYINT [[COPY]] + ; NO-FP16: $d0 = COPY [[FNEARBYINT]](<2 x s32>) + ; NO-FP16: RET_ReallyLR implicit $d0 + %0:_(<2 x s32>) = COPY $d0 + %1:_(<2 x s32>) = G_FNEARBYINT %0 + $d0 = COPY %1(<2 x s32>) + RET_ReallyLR implicit $d0 + +... +--- +name: test_v2f64.nearbyint +alignment: 2 +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $q0 + + ; FP16-LABEL: name: test_v2f64.nearbyint + ; FP16: liveins: $q0 + ; FP16: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0 + ; FP16: [[FNEARBYINT:%[0-9]+]]:_(<2 x s64>) = G_FNEARBYINT [[COPY]] + ; FP16: $q0 = COPY [[FNEARBYINT]](<2 x s64>) + ; FP16: RET_ReallyLR implicit $q0 + ; NO-FP16-LABEL: name: test_v2f64.nearbyint + ; NO-FP16: liveins: $q0 + ; NO-FP16: [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0 + ; NO-FP16: [[FNEARBYINT:%[0-9]+]]:_(<2 x s64>) = G_FNEARBYINT [[COPY]] + ; NO-FP16: $q0 = COPY [[FNEARBYINT]](<2 x s64>) + ; NO-FP16: RET_ReallyLR implicit $q0 + %0:_(<2 x s64>) = COPY $q0 + %1:_(<2 x s64>) = G_FNEARBYINT %0 + $q0 = COPY %1(<2 x s64>) + RET_ReallyLR implicit $q0 + +... +--- +name: test_f32.nearbyint +alignment: 2 +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $s0 + + ; FP16-LABEL: name: test_f32.nearbyint + ; FP16: liveins: $s0 + ; FP16: [[COPY:%[0-9]+]]:_(s32) = COPY $s0 + ; FP16: [[FNEARBYINT:%[0-9]+]]:_(s32) = G_FNEARBYINT [[COPY]] + ; FP16: $s0 = COPY [[FNEARBYINT]](s32) + ; FP16: RET_ReallyLR implicit $s0 + ; NO-FP16-LABEL: name: test_f32.nearbyint + ; NO-FP16: liveins: $s0 + ; NO-FP16: [[COPY:%[0-9]+]]:_(s32) = COPY $s0 + ; NO-FP16: [[FNEARBYINT:%[0-9]+]]:_(s32) = G_FNEARBYINT [[COPY]] + ; NO-FP16: $s0 = COPY [[FNEARBYINT]](s32) + ; NO-FP16: RET_ReallyLR implicit $s0 + %0:_(s32) = COPY $s0 + %1:_(s32) = G_FNEARBYINT %0 + $s0 = COPY %1(s32) + RET_ReallyLR implicit $s0 + +... +--- +name: test_f64.nearbyint +alignment: 2 +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; FP16-LABEL: name: test_f64.nearbyint + ; FP16: liveins: $d0 + ; FP16: [[COPY:%[0-9]+]]:_(s64) = COPY $d0 + ; FP16: [[FNEARBYINT:%[0-9]+]]:_(s64) = G_FNEARBYINT [[COPY]] + ; FP16: $d0 = COPY [[FNEARBYINT]](s64) + ; FP16: RET_ReallyLR implicit $d0 + ; NO-FP16-LABEL: name: test_f64.nearbyint + ; NO-FP16: liveins: $d0 + ; NO-FP16: [[COPY:%[0-9]+]]:_(s64) = COPY $d0 + ; NO-FP16: [[FNEARBYINT:%[0-9]+]]:_(s64) = G_FNEARBYINT [[COPY]] + ; NO-FP16: $d0 = COPY [[FNEARBYINT]](s64) + ; NO-FP16: RET_ReallyLR implicit $d0 + %0:_(s64) = COPY $d0 + %1:_(s64) = G_FNEARBYINT %0 + $d0 = COPY %1(s64) + RET_ReallyLR implicit $d0 + +... +--- +name: test_f16.nearbyint +alignment: 2 +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $h0 + + ; FP16-LABEL: name: test_f16.nearbyint + ; FP16: liveins: $h0 + ; FP16: [[COPY:%[0-9]+]]:_(s16) = COPY $h0 + ; FP16: [[FNEARBYINT:%[0-9]+]]:_(s16) = G_FNEARBYINT [[COPY]] + ; FP16: $h0 = COPY [[FNEARBYINT]](s16) + ; FP16: RET_ReallyLR implicit $h0 + ; NO-FP16-LABEL: name: test_f16.nearbyint + ; NO-FP16: liveins: $h0 + ; NO-FP16: [[COPY:%[0-9]+]]:_(s16) = COPY $h0 + ; NO-FP16: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[COPY]](s16) + ; NO-FP16: [[FNEARBYINT:%[0-9]+]]:_(s32) = G_FNEARBYINT [[FPEXT]] + ; NO-FP16: [[FPTRUNC:%[0-9]+]]:_(s16) = G_FPTRUNC [[FNEARBYINT]](s32) + ; NO-FP16: $h0 = COPY [[FPTRUNC]](s16) + ; NO-FP16: RET_ReallyLR implicit $h0 + %0:_(s16) = COPY $h0 + %1:_(s16) = G_FNEARBYINT %0 + $h0 = COPY %1(s16) + RET_ReallyLR implicit $h0 + +... Index: llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir =================================================================== --- llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir +++ llvm/test/CodeGen/AArch64/GlobalISel/legalizer-info-validation.mir @@ -349,7 +349,7 @@ # DEBUG: .. type index coverage check SKIPPED: user-defined predicate detected # # DEBUG-NEXT: G_FNEARBYINT (opcode {{[0-9]+}}): 1 type index -# DEBUG: .. type index coverage check SKIPPED: no rules defined +# DEBUG: .. type index coverage check SKIPPED: user-defined predicate detected # CHECK-NOT: ill-defined Index: llvm/test/CodeGen/AArch64/GlobalISel/regbank-nearbyint.mir =================================================================== --- /dev/null +++ llvm/test/CodeGen/AArch64/GlobalISel/regbank-nearbyint.mir @@ -0,0 +1,158 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple aarch64-unknown-unknown -run-pass=regbankselect -mattr=+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s + +... +--- +name: test_v4f16.nearbyint +alignment: 2 +legalized: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; CHECK-LABEL: name: test_v4f16.nearbyint + ; CHECK: liveins: $d0 + ; CHECK: [[COPY:%[0-9]+]]:fpr(<4 x s16>) = COPY $d0 + ; CHECK: [[FNEARBYINT:%[0-9]+]]:fpr(<4 x s16>) = G_FNEARBYINT [[COPY]] + ; CHECK: $d0 = COPY [[FNEARBYINT]](<4 x s16>) + ; CHECK: RET_ReallyLR implicit $d0 + %0:_(<4 x s16>) = COPY $d0 + %1:_(<4 x s16>) = G_FNEARBYINT %0 + $d0 = COPY %1(<4 x s16>) + RET_ReallyLR implicit $d0 + +... +--- +name: test_v8f16.nearbyint +alignment: 2 +legalized: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $q0 + + ; CHECK-LABEL: name: test_v8f16.nearbyint + ; CHECK: liveins: $q0 + ; CHECK: [[COPY:%[0-9]+]]:fpr(<8 x s16>) = COPY $q0 + ; CHECK: [[FNEARBYINT:%[0-9]+]]:fpr(<8 x s16>) = G_FNEARBYINT [[COPY]] + ; CHECK: $q0 = COPY [[FNEARBYINT]](<8 x s16>) + ; CHECK: RET_ReallyLR implicit $q0 + %0:_(<8 x s16>) = COPY $q0 + %1:_(<8 x s16>) = G_FNEARBYINT %0 + $q0 = COPY %1(<8 x s16>) + RET_ReallyLR implicit $q0 + +... +--- +name: test_v2f32.nearbyint +alignment: 2 +legalized: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; CHECK-LABEL: name: test_v2f32.nearbyint + ; CHECK: liveins: $d0 + ; CHECK: [[COPY:%[0-9]+]]:fpr(<2 x s32>) = COPY $d0 + ; CHECK: [[FNEARBYINT:%[0-9]+]]:fpr(<2 x s32>) = G_FNEARBYINT [[COPY]] + ; CHECK: $d0 = COPY [[FNEARBYINT]](<2 x s32>) + ; CHECK: RET_ReallyLR implicit $d0 + %0:_(<2 x s32>) = COPY $d0 + %1:_(<2 x s32>) = G_FNEARBYINT %0 + $d0 = COPY %1(<2 x s32>) + RET_ReallyLR implicit $d0 + +... +--- +name: test_v2f64.nearbyint +alignment: 2 +legalized: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $q0 + + ; CHECK-LABEL: name: test_v2f64.nearbyint + ; CHECK: liveins: $q0 + ; CHECK: [[COPY:%[0-9]+]]:fpr(<2 x s64>) = COPY $q0 + ; CHECK: [[FNEARBYINT:%[0-9]+]]:fpr(<2 x s64>) = G_FNEARBYINT [[COPY]] + ; CHECK: $q0 = COPY [[FNEARBYINT]](<2 x s64>) + ; CHECK: RET_ReallyLR implicit $q0 + %0:_(<2 x s64>) = COPY $q0 + %1:_(<2 x s64>) = G_FNEARBYINT %0 + $q0 = COPY %1(<2 x s64>) + RET_ReallyLR implicit $q0 + +... +--- +name: test_f32.nearbyint +alignment: 2 +legalized: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $s0 + + ; CHECK-LABEL: name: test_f32.nearbyint + ; CHECK: liveins: $s0 + ; CHECK: [[COPY:%[0-9]+]]:fpr(s32) = COPY $s0 + ; CHECK: [[FNEARBYINT:%[0-9]+]]:fpr(s32) = G_FNEARBYINT [[COPY]] + ; CHECK: $s0 = COPY [[FNEARBYINT]](s32) + ; CHECK: RET_ReallyLR implicit $s0 + %0:_(s32) = COPY $s0 + %1:_(s32) = G_FNEARBYINT %0 + $s0 = COPY %1(s32) + RET_ReallyLR implicit $s0 + +... +--- +name: test_f64.nearbyint +alignment: 2 +legalized: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; CHECK-LABEL: name: test_f64.nearbyint + ; CHECK: liveins: $d0 + ; CHECK: [[COPY:%[0-9]+]]:fpr(s64) = COPY $d0 + ; CHECK: [[FNEARBYINT:%[0-9]+]]:fpr(s64) = G_FNEARBYINT [[COPY]] + ; CHECK: $d0 = COPY [[FNEARBYINT]](s64) + ; CHECK: RET_ReallyLR implicit $d0 + %0:_(s64) = COPY $d0 + %1:_(s64) = G_FNEARBYINT %0 + $d0 = COPY %1(s64) + RET_ReallyLR implicit $d0 + +... +--- +name: test_f16.nearbyint +alignment: 2 +legalized: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $h0 + + ; CHECK-LABEL: name: test_f16.nearbyint + ; CHECK: liveins: $h0 + ; CHECK: [[COPY:%[0-9]+]]:fpr(s16) = COPY $h0 + ; CHECK: [[FNEARBYINT:%[0-9]+]]:fpr(s16) = G_FNEARBYINT [[COPY]] + ; CHECK: $h0 = COPY [[FNEARBYINT]](s16) + ; CHECK: RET_ReallyLR implicit $h0 + %0:_(s16) = COPY $h0 + %1:_(s16) = G_FNEARBYINT %0 + $h0 = COPY %1(s16) + RET_ReallyLR implicit $h0 + +... Index: llvm/test/CodeGen/AArch64/GlobalISel/select-nearbyint.mir =================================================================== --- /dev/null +++ llvm/test/CodeGen/AArch64/GlobalISel/select-nearbyint.mir @@ -0,0 +1,165 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple aarch64-unknown-unknown -run-pass=instruction-select -mattr=+fullfp16 -verify-machineinstrs %s -o - | FileCheck %s + +... +--- +name: test_v4f16.nearbyint +alignment: 2 +legalized: true +regBankSelected: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; CHECK-LABEL: name: test_v4f16.nearbyint + ; CHECK: liveins: $d0 + ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0 + ; CHECK: [[FRINTIv4f16_:%[0-9]+]]:fpr64 = FRINTIv4f16 [[COPY]] + ; CHECK: $d0 = COPY [[FRINTIv4f16_]] + ; CHECK: RET_ReallyLR implicit $d0 + %0:fpr(<4 x s16>) = COPY $d0 + %1:fpr(<4 x s16>) = G_FNEARBYINT %0 + $d0 = COPY %1(<4 x s16>) + RET_ReallyLR implicit $d0 + +... +--- +name: test_v8f16.nearbyint +alignment: 2 +legalized: true +regBankSelected: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $q0 + + ; CHECK-LABEL: name: test_v8f16.nearbyint + ; CHECK: liveins: $q0 + ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0 + ; CHECK: [[FRINTIv8f16_:%[0-9]+]]:fpr128 = FRINTIv8f16 [[COPY]] + ; CHECK: $q0 = COPY [[FRINTIv8f16_]] + ; CHECK: RET_ReallyLR implicit $q0 + %0:fpr(<8 x s16>) = COPY $q0 + %1:fpr(<8 x s16>) = G_FNEARBYINT %0 + $q0 = COPY %1(<8 x s16>) + RET_ReallyLR implicit $q0 + +... +--- +name: test_v2f32.nearbyint +alignment: 2 +legalized: true +regBankSelected: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; CHECK-LABEL: name: test_v2f32.nearbyint + ; CHECK: liveins: $d0 + ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0 + ; CHECK: [[FRINTIv2f32_:%[0-9]+]]:fpr64 = FRINTIv2f32 [[COPY]] + ; CHECK: $d0 = COPY [[FRINTIv2f32_]] + ; CHECK: RET_ReallyLR implicit $d0 + %0:fpr(<2 x s32>) = COPY $d0 + %1:fpr(<2 x s32>) = G_FNEARBYINT %0 + $d0 = COPY %1(<2 x s32>) + RET_ReallyLR implicit $d0 + +... +--- +name: test_v2f64.nearbyint +alignment: 2 +legalized: true +regBankSelected: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $q0 + + ; CHECK-LABEL: name: test_v2f64.nearbyint + ; CHECK: liveins: $q0 + ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0 + ; CHECK: [[FRINTIv2f64_:%[0-9]+]]:fpr128 = FRINTIv2f64 [[COPY]] + ; CHECK: $q0 = COPY [[FRINTIv2f64_]] + ; CHECK: RET_ReallyLR implicit $q0 + %0:fpr(<2 x s64>) = COPY $q0 + %1:fpr(<2 x s64>) = G_FNEARBYINT %0 + $q0 = COPY %1(<2 x s64>) + RET_ReallyLR implicit $q0 + +... +--- +name: test_f32.nearbyint +alignment: 2 +legalized: true +regBankSelected: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $s0 + + ; CHECK-LABEL: name: test_f32.nearbyint + ; CHECK: liveins: $s0 + ; CHECK: [[COPY:%[0-9]+]]:fpr32 = COPY $s0 + ; CHECK: [[FRINTISr:%[0-9]+]]:fpr32 = FRINTISr [[COPY]] + ; CHECK: $s0 = COPY [[FRINTISr]] + ; CHECK: RET_ReallyLR implicit $s0 + %0:fpr(s32) = COPY $s0 + %1:fpr(s32) = G_FNEARBYINT %0 + $s0 = COPY %1(s32) + RET_ReallyLR implicit $s0 + +... +--- +name: test_f64.nearbyint +alignment: 2 +legalized: true +regBankSelected: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $d0 + + ; CHECK-LABEL: name: test_f64.nearbyint + ; CHECK: liveins: $d0 + ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0 + ; CHECK: [[FRINTIDr:%[0-9]+]]:fpr64 = FRINTIDr [[COPY]] + ; CHECK: $d0 = COPY [[FRINTIDr]] + ; CHECK: RET_ReallyLR implicit $d0 + %0:fpr(s64) = COPY $d0 + %1:fpr(s64) = G_FNEARBYINT %0 + $d0 = COPY %1(s64) + RET_ReallyLR implicit $d0 + +... +--- +name: test_f16.nearbyint +alignment: 2 +legalized: true +regBankSelected: true +tracksRegLiveness: true +machineFunctionInfo: {} +body: | + bb.0: + liveins: $h0 + + ; CHECK-LABEL: name: test_f16.nearbyint + ; CHECK: liveins: $h0 + ; CHECK: [[COPY:%[0-9]+]]:fpr16 = COPY $h0 + ; CHECK: [[FRINTIHr:%[0-9]+]]:fpr16 = FRINTIHr [[COPY]] + ; CHECK: $h0 = COPY [[FRINTIHr]] + ; CHECK: RET_ReallyLR implicit $h0 + %0:fpr(s16) = COPY $h0 + %1:fpr(s16) = G_FNEARBYINT %0 + $h0 = COPY %1(s16) + RET_ReallyLR implicit $h0 + +... Index: llvm/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll =================================================================== --- llvm/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll +++ llvm/test/CodeGen/AArch64/arm64-vfloatintrinsics.ll @@ -195,12 +195,19 @@ %1 = call %v4f16 @llvm.rint.v4f16(%v4f16 %a) ret %v4f16 %1 } + +; FALLBACK-NOT: remark{{.*}}test_v4f16.nearbyint define %v4f16 @test_v4f16.nearbyint(%v4f16 %a) { ; CHECK-LABEL: test_v4f16.nearbyint: ; CHECK-NOFP16-COUNT-4: frinti s{{[0-9]+}}, s{{[0-9]+}} ; CHECK-FP16-NOT: fcvt ; CHECK-FP16: frinti.4h ; CHECK-FP16-NEXT: ret + ; GISEL-LABEL: test_v4f16.nearbyint: + ; GISEL-NOFP16-COUNT-4: frinti s{{[0-9]+}}, s{{[0-9]+}} + ; GISEL-FP16-NOT: fcvt + ; GISEL-FP16: frinti.4h + ; GISEL-FP16-NEXT: ret %1 = call %v4f16 @llvm.nearbyint.v4f16(%v4f16 %a) ret %v4f16 %1 } @@ -420,12 +427,19 @@ %1 = call %v8f16 @llvm.rint.v8f16(%v8f16 %a) ret %v8f16 %1 } + +; FALLBACK-NOT: remark{{.*}}test_v8f16.nearbyint define %v8f16 @test_v8f16.nearbyint(%v8f16 %a) { ; CHECK-LABEL: test_v8f16.nearbyint: ; CHECK-NOFP16-COUNT-8: frinti s{{[0-9]+}}, s{{[0-9]+}} ; CHECK-FP16-NOT: fcvt ; CHECK-FP16: frinti.8h ; CHECK-FP16-NEXT: ret + ; GISEL-LABEL: test_v8f16.nearbyint: + ; GISEL-NOFP16-COUNT-8: frinti s{{[0-9]+}}, s{{[0-9]+}} + ; GISEL-FP16-NOT: fcvt + ; GISEL-FP16: frinti.8h + ; GISEL-FP16-NEXT: ret %1 = call %v8f16 @llvm.nearbyint.v8f16(%v8f16 %a) ret %v8f16 %1 } @@ -597,9 +611,13 @@ %1 = call %v2f32 @llvm.rint.v2f32(%v2f32 %a) ret %v2f32 %1 } + +; FALLBACK-NOT: remark{{.*}}test_v2f32.nearbyint ; CHECK-LABEL: test_v2f32.nearbyint: +; GISEL-LABEL: test_v2f32.nearbyint: define %v2f32 @test_v2f32.nearbyint(%v2f32 %a) { ; CHECK: frinti.2s + ; GISEL: frinti.2s %1 = call %v2f32 @llvm.nearbyint.v2f32(%v2f32 %a) ret %v2f32 %1 } @@ -760,9 +778,13 @@ %1 = call %v4f32 @llvm.rint.v4f32(%v4f32 %a) ret %v4f32 %1 } + +; FALLBACK-NOT: remark{{.*}}test_v4f32.nearbyint ; CHECK: test_v4f32.nearbyint: +; GISEL: test_v4f32.nearbyint: define %v4f32 @test_v4f32.nearbyint(%v4f32 %a) { ; CHECK: frinti.4s + ; GISEL: frinti.4s %1 = call %v4f32 @llvm.nearbyint.v4f32(%v4f32 %a) ret %v4f32 %1 } @@ -923,9 +945,13 @@ %1 = call %v2f64 @llvm.rint.v2f64(%v2f64 %a) ret %v2f64 %1 } + +; FALLBACK-NOT: remark{{.*}}test_v2f64.nearbyint ; CHECK: test_v2f64.nearbyint: +; GISEL: test_v2f64.nearbyint: define %v2f64 @test_v2f64.nearbyint(%v2f64 %a) { ; CHECK: frinti.2d + ; GISEL: frinti.2d %1 = call %v2f64 @llvm.nearbyint.v2f64(%v2f64 %a) ret %v2f64 %1 }