diff --git a/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp b/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp --- a/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp +++ b/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp @@ -141,8 +141,14 @@ .legalFor({s32}) .clampScalar(0, s32, s32); - getActionDefinitionsBuilder({G_SDIV, G_SREM, G_UREM, G_UDIV}) - .legalFor({s32}) + getActionDefinitionsBuilder({G_SDIV, G_SREM, G_UDIV, G_UREM}) + .legalIf([=, &ST](const LegalityQuery &Query) { + if (CheckTyN(0, Query, {s32})) + return true; + if (ST.hasMSA() && CheckTyN(0, Query, {v16s8, v8s16, v4s32, v2s64})) + return true; + return false; + }) .minScalar(0, s32) .libcallFor({s64}); @@ -378,6 +384,26 @@ case Intrinsic::mips_mulv_w: case Intrinsic::mips_mulv_d: return MSA3OpIntrinsicToGeneric(MI, TargetOpcode::G_MUL, MIRBuilder, ST); + case Intrinsic::mips_div_s_b: + case Intrinsic::mips_div_s_h: + case Intrinsic::mips_div_s_w: + case Intrinsic::mips_div_s_d: + return MSA3OpIntrinsicToGeneric(MI, TargetOpcode::G_SDIV, MIRBuilder, ST); + case Intrinsic::mips_mod_s_b: + case Intrinsic::mips_mod_s_h: + case Intrinsic::mips_mod_s_w: + case Intrinsic::mips_mod_s_d: + return MSA3OpIntrinsicToGeneric(MI, TargetOpcode::G_SREM, MIRBuilder, ST); + case Intrinsic::mips_div_u_b: + case Intrinsic::mips_div_u_h: + case Intrinsic::mips_div_u_w: + case Intrinsic::mips_div_u_d: + return MSA3OpIntrinsicToGeneric(MI, TargetOpcode::G_UDIV, MIRBuilder, ST); + case Intrinsic::mips_mod_u_b: + case Intrinsic::mips_mod_u_h: + case Intrinsic::mips_mod_u_w: + case Intrinsic::mips_mod_u_d: + return MSA3OpIntrinsicToGeneric(MI, TargetOpcode::G_UREM, MIRBuilder, ST); default: break; } diff --git a/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp b/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp --- a/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp +++ b/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp @@ -449,10 +449,6 @@ case G_SHL: case G_ASHR: case G_LSHR: - case G_SDIV: - case G_UDIV: - case G_SREM: - case G_UREM: case G_BRINDIRECT: case G_VASTART: OperandsMapping = &Mips::ValueMappings[Mips::GPRIdx]; @@ -460,6 +456,10 @@ case G_ADD: case G_SUB: case G_MUL: + case G_SDIV: + case G_SREM: + case G_UDIV: + case G_UREM: OperandsMapping = &Mips::ValueMappings[Mips::GPRIdx]; if (Op0Size == 128) OperandsMapping = getMSAMapping(MF); diff --git a/llvm/test/CodeGen/Mips/GlobalISel/instruction-select/rem_and_div_vec.mir b/llvm/test/CodeGen/Mips/GlobalISel/instruction-select/rem_and_div_vec.mir new file mode 100644 --- /dev/null +++ b/llvm/test/CodeGen/Mips/GlobalISel/instruction-select/rem_and_div_vec.mir @@ -0,0 +1,505 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=mipsel-linux-gnu -mcpu=mips32r5 -mattr=+msa,+fp64,+nan2008 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=P5600 +--- | + + define void @sdiv_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @sdiv_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @sdiv_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @sdiv_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @srem_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @srem_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @srem_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @srem_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @udiv_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @udiv_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @udiv_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @udiv_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @urem_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @urem_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @urem_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @urem_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + +... +--- +name: sdiv_v16i8 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v16i8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_B:%[0-9]+]]:msa128b = LD_B [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_B1:%[0-9]+]]:msa128b = LD_B [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_S_B:%[0-9]+]]:msa128b = DIV_S_B [[LD_B]], [[LD_B1]] + ; P5600: ST_B [[DIV_S_B]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<16 x s8>) = G_SDIV %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v8i16 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v8i16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_H:%[0-9]+]]:msa128h = LD_H [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_H1:%[0-9]+]]:msa128h = LD_H [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_S_H:%[0-9]+]]:msa128h = DIV_S_H [[LD_H]], [[LD_H1]] + ; P5600: ST_H [[DIV_S_H]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<8 x s16>) = G_SDIV %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v4i32 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v4i32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_W:%[0-9]+]]:msa128w = LD_W [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_W1:%[0-9]+]]:msa128w = LD_W [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_S_W:%[0-9]+]]:msa128w = DIV_S_W [[LD_W]], [[LD_W1]] + ; P5600: ST_W [[DIV_S_W]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<4 x s32>) = G_SDIV %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v2i64 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v2i64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_D:%[0-9]+]]:msa128d = LD_D [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_D1:%[0-9]+]]:msa128d = LD_D [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_S_D:%[0-9]+]]:msa128d = DIV_S_D [[LD_D]], [[LD_D1]] + ; P5600: ST_D [[DIV_S_D]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<2 x s64>) = G_SDIV %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v16i8 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v16i8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_B:%[0-9]+]]:msa128b = LD_B [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_B1:%[0-9]+]]:msa128b = LD_B [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_S_B:%[0-9]+]]:msa128b = MOD_S_B [[LD_B]], [[LD_B1]] + ; P5600: ST_B [[MOD_S_B]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<16 x s8>) = G_SREM %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v8i16 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v8i16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_H:%[0-9]+]]:msa128h = LD_H [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_H1:%[0-9]+]]:msa128h = LD_H [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_S_H:%[0-9]+]]:msa128h = MOD_S_H [[LD_H]], [[LD_H1]] + ; P5600: ST_H [[MOD_S_H]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<8 x s16>) = G_SREM %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v4i32 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v4i32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_W:%[0-9]+]]:msa128w = LD_W [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_W1:%[0-9]+]]:msa128w = LD_W [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_S_W:%[0-9]+]]:msa128w = MOD_S_W [[LD_W]], [[LD_W1]] + ; P5600: ST_W [[MOD_S_W]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<4 x s32>) = G_SREM %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v2i64 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v2i64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_D:%[0-9]+]]:msa128d = LD_D [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_D1:%[0-9]+]]:msa128d = LD_D [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_S_D:%[0-9]+]]:msa128d = MOD_S_D [[LD_D]], [[LD_D1]] + ; P5600: ST_D [[MOD_S_D]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<2 x s64>) = G_SREM %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v16u8 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v16u8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_B:%[0-9]+]]:msa128b = LD_B [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_B1:%[0-9]+]]:msa128b = LD_B [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_U_B:%[0-9]+]]:msa128b = DIV_U_B [[LD_B]], [[LD_B1]] + ; P5600: ST_B [[DIV_U_B]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<16 x s8>) = G_UDIV %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v8u16 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v8u16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_H:%[0-9]+]]:msa128h = LD_H [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_H1:%[0-9]+]]:msa128h = LD_H [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_U_H:%[0-9]+]]:msa128h = DIV_U_H [[LD_H]], [[LD_H1]] + ; P5600: ST_H [[DIV_U_H]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<8 x s16>) = G_UDIV %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v4u32 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v4u32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_W:%[0-9]+]]:msa128w = LD_W [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_W1:%[0-9]+]]:msa128w = LD_W [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_U_W:%[0-9]+]]:msa128w = DIV_U_W [[LD_W]], [[LD_W1]] + ; P5600: ST_W [[DIV_U_W]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<4 x s32>) = G_UDIV %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v2u64 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v2u64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_D:%[0-9]+]]:msa128d = LD_D [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_D1:%[0-9]+]]:msa128d = LD_D [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[DIV_U_D:%[0-9]+]]:msa128d = DIV_U_D [[LD_D]], [[LD_D1]] + ; P5600: ST_D [[DIV_U_D]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<2 x s64>) = G_UDIV %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v16u8 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v16u8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_B:%[0-9]+]]:msa128b = LD_B [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_B1:%[0-9]+]]:msa128b = LD_B [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_U_B:%[0-9]+]]:msa128b = MOD_U_B [[LD_B]], [[LD_B1]] + ; P5600: ST_B [[MOD_U_B]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<16 x s8>) = G_UREM %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v8u16 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v8u16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_H:%[0-9]+]]:msa128h = LD_H [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_H1:%[0-9]+]]:msa128h = LD_H [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_U_H:%[0-9]+]]:msa128h = MOD_U_H [[LD_H]], [[LD_H1]] + ; P5600: ST_H [[MOD_U_H]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<8 x s16>) = G_UREM %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v4u32 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v4u32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_W:%[0-9]+]]:msa128w = LD_W [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_W1:%[0-9]+]]:msa128w = LD_W [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_U_W:%[0-9]+]]:msa128w = MOD_U_W [[LD_W]], [[LD_W1]] + ; P5600: ST_W [[MOD_U_W]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<4 x s32>) = G_UREM %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v2u64 +alignment: 4 +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v2u64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gpr32 = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gpr32 = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gpr32 = COPY $a2 + ; P5600: [[LD_D:%[0-9]+]]:msa128d = LD_D [[COPY]], 0 :: (load 16 from %ir.a) + ; P5600: [[LD_D1:%[0-9]+]]:msa128d = LD_D [[COPY1]], 0 :: (load 16 from %ir.b) + ; P5600: [[MOD_U_D:%[0-9]+]]:msa128d = MOD_U_D [[LD_D]], [[LD_D1]] + ; P5600: ST_D [[MOD_U_D]], [[COPY2]], 0 :: (store 16 into %ir.c) + ; P5600: RetRA + %0:gprb(p0) = COPY $a0 + %1:gprb(p0) = COPY $a1 + %2:gprb(p0) = COPY $a2 + %3:fprb(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:fprb(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:fprb(<2 x s64>) = G_UREM %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... diff --git a/llvm/test/CodeGen/Mips/GlobalISel/legalizer/rem_and_div_vec.mir b/llvm/test/CodeGen/Mips/GlobalISel/legalizer/rem_and_div_vec.mir new file mode 100644 --- /dev/null +++ b/llvm/test/CodeGen/Mips/GlobalISel/legalizer/rem_and_div_vec.mir @@ -0,0 +1,473 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=mipsel-linux-gnu -mcpu=mips32r5 -mattr=+msa,+fp64,+nan2008 -run-pass=legalizer -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=P5600 +--- | + + define void @sdiv_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @sdiv_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @sdiv_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @sdiv_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @srem_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @srem_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @srem_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @srem_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @udiv_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @udiv_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @udiv_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @udiv_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @urem_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @urem_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @urem_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @urem_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + +... +--- +name: sdiv_v16i8 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v16i8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<16 x s8>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_SDIV %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v8i16 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v8i16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<8 x s16>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_SDIV %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v4i32 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v4i32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<4 x s32>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_SDIV %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v2i64 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v2i64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<2 x s64>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_SDIV %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v16i8 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v16i8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<16 x s8>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_SREM %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v8i16 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v8i16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<8 x s16>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_SREM %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v4i32 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v4i32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<4 x s32>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_SREM %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v2i64 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v2i64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<2 x s64>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_SREM %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v16u8 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v16u8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<16 x s8>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_UDIV %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v8u16 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v8u16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<8 x s16>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_UDIV %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v4u32 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v4u32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<4 x s32>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_UDIV %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v2u64 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v2u64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<2 x s64>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_UDIV %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v16u8 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v16u8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<16 x s8>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_UREM %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v8u16 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v8u16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<8 x s16>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_UREM %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v4u32 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v4u32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<4 x s32>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_UREM %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v2u64 +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v2u64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<2 x s64>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_UREM %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... diff --git a/llvm/test/CodeGen/Mips/GlobalISel/legalizer/rem_and_div_vec_builtin.mir b/llvm/test/CodeGen/Mips/GlobalISel/legalizer/rem_and_div_vec_builtin.mir new file mode 100644 --- /dev/null +++ b/llvm/test/CodeGen/Mips/GlobalISel/legalizer/rem_and_div_vec_builtin.mir @@ -0,0 +1,501 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=mipsel-linux-gnu -mcpu=mips32r5 -mattr=+msa,+fp64,+nan2008 -run-pass=legalizer -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=P5600 +--- | + + declare <16 x i8> @llvm.mips.div.s.b(<16 x i8>, <16 x i8>) + define void @sdiv_v16i8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + + declare <8 x i16> @llvm.mips.div.s.h(<8 x i16>, <8 x i16>) + define void @sdiv_v8i16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + + declare <4 x i32> @llvm.mips.div.s.w(<4 x i32>, <4 x i32>) + define void @sdiv_v4i32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + + declare <2 x i64> @llvm.mips.div.s.d(<2 x i64>, <2 x i64>) + define void @sdiv_v2i64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + declare <16 x i8> @llvm.mips.mod.s.b(<16 x i8>, <16 x i8>) + define void @smod_v16i8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + + declare <8 x i16> @llvm.mips.mod.s.h(<8 x i16>, <8 x i16>) + define void @smod_v8i16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + + declare <4 x i32> @llvm.mips.mod.s.w(<4 x i32>, <4 x i32>) + define void @smod_v4i32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + + declare <2 x i64> @llvm.mips.mod.s.d(<2 x i64>, <2 x i64>) + define void @smod_v2i64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + declare <16 x i8> @llvm.mips.div.u.b(<16 x i8>, <16 x i8>) + define void @udiv_v16i8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + + declare <8 x i16> @llvm.mips.div.u.h(<8 x i16>, <8 x i16>) + define void @udiv_v8i16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + + declare <4 x i32> @llvm.mips.div.u.w(<4 x i32>, <4 x i32>) + define void @udiv_v4i32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + + declare <2 x i64> @llvm.mips.div.u.d(<2 x i64>, <2 x i64>) + define void @udiv_v2i64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + declare <16 x i8> @llvm.mips.mod.u.b(<16 x i8>, <16 x i8>) + define void @umod_v16i8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + + declare <8 x i16> @llvm.mips.mod.u.h(<8 x i16>, <8 x i16>) + define void @umod_v8i16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + + declare <4 x i32> @llvm.mips.mod.u.w(<4 x i32>, <4 x i32>) + define void @umod_v4i32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + + declare <2 x i64> @llvm.mips.mod.u.d(<2 x i64>, <2 x i64>) + define void @umod_v2i64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + +... +--- +name: sdiv_v16i8_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v16i8_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<16 x s8>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_INTRINSIC intrinsic(@llvm.mips.div.s.b), %3(<16 x s8>), %4(<16 x s8>) + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v8i16_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v8i16_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<8 x s16>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_INTRINSIC intrinsic(@llvm.mips.div.s.h), %3(<8 x s16>), %4(<8 x s16>) + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v4i32_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v4i32_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<4 x s32>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.mips.div.s.w), %3(<4 x s32>), %4(<4 x s32>) + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v2i64_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v2i64_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:_(<2 x s64>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_INTRINSIC intrinsic(@llvm.mips.div.s.d), %3(<2 x s64>), %4(<2 x s64>) + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: smod_v16i8_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: smod_v16i8_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<16 x s8>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_INTRINSIC intrinsic(@llvm.mips.mod.s.b), %3(<16 x s8>), %4(<16 x s8>) + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: smod_v8i16_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: smod_v8i16_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<8 x s16>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_INTRINSIC intrinsic(@llvm.mips.mod.s.h), %3(<8 x s16>), %4(<8 x s16>) + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: smod_v4i32_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: smod_v4i32_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<4 x s32>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.mips.mod.s.w), %3(<4 x s32>), %4(<4 x s32>) + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: smod_v2i64_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: smod_v2i64_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:_(<2 x s64>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_INTRINSIC intrinsic(@llvm.mips.mod.s.d), %3(<2 x s64>), %4(<2 x s64>) + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v16i8_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v16i8_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<16 x s8>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_INTRINSIC intrinsic(@llvm.mips.div.u.b), %3(<16 x s8>), %4(<16 x s8>) + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v8i16_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v8i16_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<8 x s16>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_INTRINSIC intrinsic(@llvm.mips.div.u.h), %3(<8 x s16>), %4(<8 x s16>) + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v4i32_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v4i32_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<4 x s32>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.mips.div.u.w), %3(<4 x s32>), %4(<4 x s32>) + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v2i64_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v2i64_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:_(<2 x s64>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_INTRINSIC intrinsic(@llvm.mips.div.u.d), %3(<2 x s64>), %4(<2 x s64>) + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: umod_v16i8_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: umod_v16i8_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<16 x s8>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_INTRINSIC intrinsic(@llvm.mips.mod.u.b), %3(<16 x s8>), %4(<16 x s8>) + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: umod_v8i16_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: umod_v8i16_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<8 x s16>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_INTRINSIC intrinsic(@llvm.mips.mod.u.h), %3(<8 x s16>), %4(<8 x s16>) + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: umod_v4i32_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: umod_v4i32_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<4 x s32>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_INTRINSIC intrinsic(@llvm.mips.mod.u.w), %3(<4 x s32>), %4(<4 x s32>) + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: umod_v2i64_builtin +alignment: 4 +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: umod_v2i64_builtin + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:_(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:_(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:_(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:_(<2 x s64>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_INTRINSIC intrinsic(@llvm.mips.mod.u.d), %3(<2 x s64>), %4(<2 x s64>) + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... diff --git a/llvm/test/CodeGen/Mips/GlobalISel/llvm-ir/rem_and_div_vec.ll b/llvm/test/CodeGen/Mips/GlobalISel/llvm-ir/rem_and_div_vec.ll new file mode 100644 --- /dev/null +++ b/llvm/test/CodeGen/Mips/GlobalISel/llvm-ir/rem_and_div_vec.ll @@ -0,0 +1,274 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -O0 -mtriple=mipsel-linux-gnu -global-isel -mcpu=mips32r5 -mattr=+msa,+fp64,+nan2008 -verify-machineinstrs %s -o -| FileCheck %s -check-prefixes=P5600 + +define void @sdiv_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: sdiv_v16i8: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: div_s.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %div = sdiv <16 x i8> %0, %1 + store <16 x i8> %div, <16 x i8>* %c, align 16 + ret void +} + +define void @sdiv_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: sdiv_v8i16: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: div_s.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %div = sdiv <8 x i16> %0, %1 + store <8 x i16> %div, <8 x i16>* %c, align 16 + ret void +} + +define void @sdiv_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: sdiv_v4i32: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: div_s.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %div = sdiv <4 x i32> %0, %1 + store <4 x i32> %div, <4 x i32>* %c, align 16 + ret void +} + +define void @sdiv_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: sdiv_v2i64: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: div_s.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %div = sdiv <2 x i64> %0, %1 + store <2 x i64> %div, <2 x i64>* %c, align 16 + ret void +} + +define void @srem_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: srem_v16i8: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: mod_s.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %rem = srem <16 x i8> %0, %1 + store <16 x i8> %rem, <16 x i8>* %c, align 16 + ret void +} + +define void @srem_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: srem_v8i16: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: mod_s.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %rem = srem <8 x i16> %0, %1 + store <8 x i16> %rem, <8 x i16>* %c, align 16 + ret void +} + +define void @srem_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: srem_v4i32: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: mod_s.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %rem = srem <4 x i32> %0, %1 + store <4 x i32> %rem, <4 x i32>* %c, align 16 + ret void +} + +define void @srem_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: srem_v2i64: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: mod_s.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %rem = srem <2 x i64> %0, %1 + store <2 x i64> %rem, <2 x i64>* %c, align 16 + ret void +} + +define void @udiv_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: udiv_v16u8: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: div_u.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %div = udiv <16 x i8> %0, %1 + store <16 x i8> %div, <16 x i8>* %c, align 16 + ret void +} + +define void @udiv_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: udiv_v8u16: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: div_u.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %div = udiv <8 x i16> %0, %1 + store <8 x i16> %div, <8 x i16>* %c, align 16 + ret void +} + +define void @udiv_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: udiv_v4u32: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: div_u.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %div = udiv <4 x i32> %0, %1 + store <4 x i32> %div, <4 x i32>* %c, align 16 + ret void +} + +define void @udiv_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: udiv_v2u64: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: div_u.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %div = udiv <2 x i64> %0, %1 + store <2 x i64> %div, <2 x i64>* %c, align 16 + ret void +} + +define void @urem_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: urem_v16u8: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: mod_u.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %rem = urem <16 x i8> %0, %1 + store <16 x i8> %rem, <16 x i8>* %c, align 16 + ret void +} + +define void @urem_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: urem_v8u16: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: mod_u.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %rem = urem <8 x i16> %0, %1 + store <8 x i16> %rem, <8 x i16>* %c, align 16 + ret void +} + +define void @urem_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: urem_v4u32: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: mod_u.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %rem = urem <4 x i32> %0, %1 + store <4 x i32> %rem, <4 x i32>* %c, align 16 + ret void +} + +define void @urem_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: urem_v2u64: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: mod_u.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %rem = urem <2 x i64> %0, %1 + store <2 x i64> %rem, <2 x i64>* %c, align 16 + ret void +} diff --git a/llvm/test/CodeGen/Mips/GlobalISel/llvm-ir/rem_and_div_vec_builtin.ll b/llvm/test/CodeGen/Mips/GlobalISel/llvm-ir/rem_and_div_vec_builtin.ll new file mode 100644 --- /dev/null +++ b/llvm/test/CodeGen/Mips/GlobalISel/llvm-ir/rem_and_div_vec_builtin.ll @@ -0,0 +1,290 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -O0 -mtriple=mipsel-linux-gnu -global-isel -mcpu=mips32r5 -mattr=+msa,+fp64,+nan2008 -verify-machineinstrs %s -o -| FileCheck %s -check-prefixes=P5600 + +declare <16 x i8> @llvm.mips.div.s.b(<16 x i8>, <16 x i8>) +define void @sdiv_v16i8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: sdiv_v16i8_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: div_s.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %2 = tail call <16 x i8> @llvm.mips.div.s.b(<16 x i8> %0, <16 x i8> %1) + store <16 x i8> %2, <16 x i8>* %c, align 16 + ret void +} + +declare <8 x i16> @llvm.mips.div.s.h(<8 x i16>, <8 x i16>) +define void @sdiv_v8i16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: sdiv_v8i16_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: div_s.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %2 = tail call <8 x i16> @llvm.mips.div.s.h(<8 x i16> %0, <8 x i16> %1) + store <8 x i16> %2, <8 x i16>* %c, align 16 + ret void +} + +declare <4 x i32> @llvm.mips.div.s.w(<4 x i32>, <4 x i32>) +define void @sdiv_v4i32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: sdiv_v4i32_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: div_s.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %2 = tail call <4 x i32> @llvm.mips.div.s.w(<4 x i32> %0, <4 x i32> %1) + store <4 x i32> %2, <4 x i32>* %c, align 16 + ret void +} + +declare <2 x i64> @llvm.mips.div.s.d(<2 x i64>, <2 x i64>) +define void @sdiv_v2i64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: sdiv_v2i64_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: div_s.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %2 = tail call <2 x i64> @llvm.mips.div.s.d(<2 x i64> %0, <2 x i64> %1) + store <2 x i64> %2, <2 x i64>* %c, align 16 + ret void +} + +declare <16 x i8> @llvm.mips.mod.s.b(<16 x i8>, <16 x i8>) +define void @smod_v16i8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: smod_v16i8_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: mod_s.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %2 = tail call <16 x i8> @llvm.mips.mod.s.b(<16 x i8> %0, <16 x i8> %1) + store <16 x i8> %2, <16 x i8>* %c, align 16 + ret void +} + +declare <8 x i16> @llvm.mips.mod.s.h(<8 x i16>, <8 x i16>) +define void @smod_v8i16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: smod_v8i16_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: mod_s.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %2 = tail call <8 x i16> @llvm.mips.mod.s.h(<8 x i16> %0, <8 x i16> %1) + store <8 x i16> %2, <8 x i16>* %c, align 16 + ret void +} + +declare <4 x i32> @llvm.mips.mod.s.w(<4 x i32>, <4 x i32>) +define void @smod_v4i32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: smod_v4i32_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: mod_s.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %2 = tail call <4 x i32> @llvm.mips.mod.s.w(<4 x i32> %0, <4 x i32> %1) + store <4 x i32> %2, <4 x i32>* %c, align 16 + ret void +} + +declare <2 x i64> @llvm.mips.mod.s.d(<2 x i64>, <2 x i64>) +define void @smod_v2i64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: smod_v2i64_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: mod_s.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %2 = tail call <2 x i64> @llvm.mips.mod.s.d(<2 x i64> %0, <2 x i64> %1) + store <2 x i64> %2, <2 x i64>* %c, align 16 + ret void +} + +declare <16 x i8> @llvm.mips.div.u.b(<16 x i8>, <16 x i8>) +define void @udiv_v16u8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: udiv_v16u8_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: div_u.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %2 = tail call <16 x i8> @llvm.mips.div.u.b(<16 x i8> %0, <16 x i8> %1) + store <16 x i8> %2, <16 x i8>* %c, align 16 + ret void +} + +declare <8 x i16> @llvm.mips.div.u.h(<8 x i16>, <8 x i16>) +define void @udiv_v8u16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: udiv_v8u16_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: div_u.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %2 = tail call <8 x i16> @llvm.mips.div.u.h(<8 x i16> %0, <8 x i16> %1) + store <8 x i16> %2, <8 x i16>* %c, align 16 + ret void +} + +declare <4 x i32> @llvm.mips.div.u.w(<4 x i32>, <4 x i32>) +define void @udiv_v4u32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: udiv_v4u32_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: div_u.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %2 = tail call <4 x i32> @llvm.mips.div.u.w(<4 x i32> %0, <4 x i32> %1) + store <4 x i32> %2, <4 x i32>* %c, align 16 + ret void +} + +declare <2 x i64> @llvm.mips.div.u.d(<2 x i64>, <2 x i64>) +define void @udiv_v2u64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: udiv_v2u64_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: div_u.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %2 = tail call <2 x i64> @llvm.mips.div.u.d(<2 x i64> %0, <2 x i64> %1) + store <2 x i64> %2, <2 x i64>* %c, align 16 + ret void +} + +declare <16 x i8> @llvm.mips.mod.u.b(<16 x i8>, <16 x i8>) +define void @umod_v16u8_builtin(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { +; P5600-LABEL: umod_v16u8_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.b $w0, 0($4) +; P5600-NEXT: ld.b $w1, 0($5) +; P5600-NEXT: mod_u.b $w0, $w0, $w1 +; P5600-NEXT: st.b $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <16 x i8>, <16 x i8>* %a, align 16 + %1 = load <16 x i8>, <16 x i8>* %b, align 16 + %2 = tail call <16 x i8> @llvm.mips.mod.u.b(<16 x i8> %0, <16 x i8> %1) + store <16 x i8> %2, <16 x i8>* %c, align 16 + ret void +} + +declare <8 x i16> @llvm.mips.mod.u.h(<8 x i16>, <8 x i16>) +define void @umod_v8u16_builtin(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { +; P5600-LABEL: umod_v8u16_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.h $w0, 0($4) +; P5600-NEXT: ld.h $w1, 0($5) +; P5600-NEXT: mod_u.h $w0, $w0, $w1 +; P5600-NEXT: st.h $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <8 x i16>, <8 x i16>* %a, align 16 + %1 = load <8 x i16>, <8 x i16>* %b, align 16 + %2 = tail call <8 x i16> @llvm.mips.mod.u.h(<8 x i16> %0, <8 x i16> %1) + store <8 x i16> %2, <8 x i16>* %c, align 16 + ret void +} + +declare <4 x i32> @llvm.mips.mod.u.w(<4 x i32>, <4 x i32>) +define void @umod_v4u32_builtin(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { +; P5600-LABEL: umod_v4u32_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.w $w0, 0($4) +; P5600-NEXT: ld.w $w1, 0($5) +; P5600-NEXT: mod_u.w $w0, $w0, $w1 +; P5600-NEXT: st.w $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <4 x i32>, <4 x i32>* %a, align 16 + %1 = load <4 x i32>, <4 x i32>* %b, align 16 + %2 = tail call <4 x i32> @llvm.mips.mod.u.w(<4 x i32> %0, <4 x i32> %1) + store <4 x i32> %2, <4 x i32>* %c, align 16 + ret void +} + +declare <2 x i64> @llvm.mips.mod.u.d(<2 x i64>, <2 x i64>) +define void @umod_v2u64_builtin(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { +; P5600-LABEL: umod_v2u64_builtin: +; P5600: # %bb.0: # %entry +; P5600-NEXT: ld.d $w0, 0($4) +; P5600-NEXT: ld.d $w1, 0($5) +; P5600-NEXT: mod_u.d $w0, $w0, $w1 +; P5600-NEXT: st.d $w0, 0($6) +; P5600-NEXT: jr $ra +; P5600-NEXT: nop +entry: + %0 = load <2 x i64>, <2 x i64>* %a, align 16 + %1 = load <2 x i64>, <2 x i64>* %b, align 16 + %2 = tail call <2 x i64> @llvm.mips.mod.u.d(<2 x i64> %0, <2 x i64> %1) + store <2 x i64> %2, <2 x i64>* %c, align 16 + ret void +} diff --git a/llvm/test/CodeGen/Mips/GlobalISel/regbankselect/rem_and_div_vec.mir b/llvm/test/CodeGen/Mips/GlobalISel/regbankselect/rem_and_div_vec.mir new file mode 100644 --- /dev/null +++ b/llvm/test/CodeGen/Mips/GlobalISel/regbankselect/rem_and_div_vec.mir @@ -0,0 +1,489 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=mipsel-linux-gnu -mcpu=mips32r5 -mattr=+msa,+fp64,+nan2008 -run-pass=regbankselect -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=P5600 +--- | + + define void @sdiv_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @sdiv_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @sdiv_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @sdiv_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @srem_v16i8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @srem_v8i16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @srem_v4i32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @srem_v2i64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @udiv_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @udiv_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @udiv_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @udiv_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + + define void @urem_v16u8(<16 x i8>* %a, <16 x i8>* %b, <16 x i8>* %c) { entry: ret void } + define void @urem_v8u16(<8 x i16>* %a, <8 x i16>* %b, <8 x i16>* %c) { entry: ret void } + define void @urem_v4u32(<4 x i32>* %a, <4 x i32>* %b, <4 x i32>* %c) { entry: ret void } + define void @urem_v2u64(<2 x i64>* %a, <2 x i64>* %b, <2 x i64>* %c) { entry: ret void } + +... +--- +name: sdiv_v16i8 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v16i8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:fprb(<16 x s8>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_SDIV %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v8i16 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v8i16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:fprb(<8 x s16>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_SDIV %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v4i32 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v4i32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:fprb(<4 x s32>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_SDIV %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: sdiv_v2i64 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: sdiv_v2i64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SDIV:%[0-9]+]]:fprb(<2 x s64>) = G_SDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SDIV]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_SDIV %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v16i8 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v16i8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:fprb(<16 x s8>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_SREM %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v8i16 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v8i16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:fprb(<8 x s16>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_SREM %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v4i32 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v4i32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:fprb(<4 x s32>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_SREM %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: srem_v2i64 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: srem_v2i64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[SREM:%[0-9]+]]:fprb(<2 x s64>) = G_SREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[SREM]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_SREM %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v16u8 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v16u8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:fprb(<16 x s8>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_UDIV %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v8u16 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v8u16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:fprb(<8 x s16>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_UDIV %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v4u32 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v4u32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:fprb(<4 x s32>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_UDIV %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: udiv_v2u64 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: udiv_v2u64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UDIV:%[0-9]+]]:fprb(<2 x s64>) = G_UDIV [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UDIV]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_UDIV %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v16u8 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v16u8 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<16 x s8>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:fprb(<16 x s8>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<16 x s8>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<16 x s8>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<16 x s8>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<16 x s8>) = G_UREM %3, %4 + G_STORE %5(<16 x s8>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v8u16 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v8u16 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<8 x s16>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:fprb(<8 x s16>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<8 x s16>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<8 x s16>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<8 x s16>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<8 x s16>) = G_UREM %3, %4 + G_STORE %5(<8 x s16>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v4u32 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v4u32 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<4 x s32>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:fprb(<4 x s32>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<4 x s32>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<4 x s32>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<4 x s32>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<4 x s32>) = G_UREM %3, %4 + G_STORE %5(<4 x s32>), %2(p0) :: (store 16 into %ir.c) + RetRA + +... +--- +name: urem_v2u64 +alignment: 4 +legalized: true +tracksRegLiveness: true +body: | + bb.1.entry: + liveins: $a0, $a1, $a2 + + ; P5600-LABEL: name: urem_v2u64 + ; P5600: liveins: $a0, $a1, $a2 + ; P5600: [[COPY:%[0-9]+]]:gprb(p0) = COPY $a0 + ; P5600: [[COPY1:%[0-9]+]]:gprb(p0) = COPY $a1 + ; P5600: [[COPY2:%[0-9]+]]:gprb(p0) = COPY $a2 + ; P5600: [[LOAD:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load 16 from %ir.a) + ; P5600: [[LOAD1:%[0-9]+]]:fprb(<2 x s64>) = G_LOAD [[COPY1]](p0) :: (load 16 from %ir.b) + ; P5600: [[UREM:%[0-9]+]]:fprb(<2 x s64>) = G_UREM [[LOAD]], [[LOAD1]] + ; P5600: G_STORE [[UREM]](<2 x s64>), [[COPY2]](p0) :: (store 16 into %ir.c) + ; P5600: RetRA + %0:_(p0) = COPY $a0 + %1:_(p0) = COPY $a1 + %2:_(p0) = COPY $a2 + %3:_(<2 x s64>) = G_LOAD %0(p0) :: (load 16 from %ir.a) + %4:_(<2 x s64>) = G_LOAD %1(p0) :: (load 16 from %ir.b) + %5:_(<2 x s64>) = G_UREM %3, %4 + G_STORE %5(<2 x s64>), %2(p0) :: (store 16 into %ir.c) + RetRA + +...