aboutsummaryrefslogtreecommitdiff
path: root/llvm/lib/Target
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/lib/Target')
-rw-r--r--llvm/lib/Target/AArch64/AArch64ISelLowering.cpp40
-rw-r--r--llvm/lib/Target/AArch64/AArch64ISelLowering.h1
-rw-r--r--llvm/lib/Target/AArch64/AArch64InstrGISel.td7
-rw-r--r--llvm/lib/Target/AArch64/AArch64InstrInfo.cpp3
-rw-r--r--llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp2
-rw-r--r--llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp2
-rw-r--r--llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp129
-rw-r--r--llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.h8
-rw-r--r--llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp180
-rw-r--r--llvm/lib/Target/AMDGPU/GCNSchedStrategy.h19
-rw-r--r--llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp2
-rw-r--r--llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h3
-rw-r--r--llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp4
-rw-r--r--llvm/lib/Target/DirectX/DXIL.td8
-rw-r--r--llvm/lib/Target/DirectX/DXILOpBuilder.cpp8
-rw-r--r--llvm/lib/Target/DirectX/DXILOpLowering.cpp25
-rw-r--r--llvm/lib/Target/DirectX/DXILWriter/DXILBitcodeWriter.cpp15
-rw-r--r--llvm/lib/Target/DirectX/DXILWriter/DXILWriterPass.cpp8
-rw-r--r--llvm/lib/Target/Hexagon/Disassembler/HexagonDisassembler.cpp16
-rw-r--r--llvm/lib/Target/Hexagon/HexagonBitSimplify.cpp4
-rw-r--r--llvm/lib/Target/Hexagon/HexagonCommonGEP.cpp12
-rw-r--r--llvm/lib/Target/Hexagon/HexagonEarlyIfConv.cpp3
-rw-r--r--llvm/lib/Target/Hexagon/HexagonGenPredicate.cpp10
-rw-r--r--llvm/lib/Target/Hexagon/MCTargetDesc/HexagonMCTargetDesc.cpp4
-rw-r--r--llvm/lib/Target/Mips/MipsFastISel.cpp7
-rw-r--r--llvm/lib/Target/Mips/MipsISelLowering.cpp7
-rw-r--r--llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp6
-rw-r--r--llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp12
-rw-r--r--llvm/lib/Target/RISCV/MCA/RISCVCustomBehaviour.cpp2
-rw-r--r--llvm/lib/Target/RISCV/RISCVISelLowering.cpp2
-rw-r--r--llvm/lib/Target/RISCV/RISCVSchedSiFive7.td20
-rw-r--r--llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp1
-rw-r--r--llvm/lib/Target/Sparc/SparcFrameLowering.cpp3
-rw-r--r--llvm/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTypeUtilities.cpp2
-rw-r--r--llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td32
-rw-r--r--llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp11
-rw-r--r--llvm/lib/Target/X86/GISel/X86LegalizerInfo.cpp2
-rw-r--r--llvm/lib/Target/X86/X86.td7
-rw-r--r--llvm/lib/Target/X86/X86FloatingPoint.cpp3
-rw-r--r--llvm/lib/Target/X86/X86ISelLowering.cpp7
-rw-r--r--llvm/lib/Target/X86/X86InstrCompiler.td13
-rw-r--r--llvm/lib/Target/X86/X86MCInstLower.cpp36
42 files changed, 512 insertions, 174 deletions
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index be2f2e4..662d84b 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -1561,6 +1561,7 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
setOperationAction(ISD::VECREDUCE_ADD, VT, Custom);
setOperationAction(ISD::VECREDUCE_AND, VT, Custom);
setOperationAction(ISD::VECREDUCE_OR, VT, Custom);
+ setOperationAction(ISD::VECREDUCE_MUL, VT, Custom);
setOperationAction(ISD::VECREDUCE_XOR, VT, Custom);
setOperationAction(ISD::VECREDUCE_UMIN, VT, Custom);
setOperationAction(ISD::VECREDUCE_UMAX, VT, Custom);
@@ -1717,6 +1718,7 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
setOperationAction(ISD::VECREDUCE_FMIN, VT, Custom);
setOperationAction(ISD::VECREDUCE_FMAXIMUM, VT, Custom);
setOperationAction(ISD::VECREDUCE_FMINIMUM, VT, Custom);
+ setOperationAction(ISD::VECREDUCE_FMUL, VT, Custom);
setOperationAction(ISD::VECTOR_SPLICE, VT, Custom);
setOperationAction(ISD::VECTOR_DEINTERLEAVE, VT, Custom);
setOperationAction(ISD::VECTOR_INTERLEAVE, VT, Custom);
@@ -7775,6 +7777,9 @@ SDValue AArch64TargetLowering::LowerOperation(SDValue Op,
case ISD::VECREDUCE_FMAXIMUM:
case ISD::VECREDUCE_FMINIMUM:
return LowerVECREDUCE(Op, DAG);
+ case ISD::VECREDUCE_MUL:
+ case ISD::VECREDUCE_FMUL:
+ return LowerVECREDUCE_MUL(Op, DAG);
case ISD::ATOMIC_LOAD_AND:
return LowerATOMIC_LOAD_AND(Op, DAG);
case ISD::DYNAMIC_STACKALLOC:
@@ -16794,6 +16799,33 @@ SDValue AArch64TargetLowering::LowerVECREDUCE(SDValue Op,
}
}
+SDValue AArch64TargetLowering::LowerVECREDUCE_MUL(SDValue Op,
+ SelectionDAG &DAG) const {
+ SDLoc DL(Op);
+ SDValue Src = Op.getOperand(0);
+ EVT SrcVT = Src.getValueType();
+ assert(SrcVT.isScalableVector() && "Unexpected operand type!");
+
+ SDVTList SrcVTs = DAG.getVTList(SrcVT, SrcVT);
+ unsigned BaseOpc = ISD::getVecReduceBaseOpcode(Op.getOpcode());
+ SDValue Identity = DAG.getNeutralElement(BaseOpc, DL, SrcVT, Op->getFlags());
+
+ // Whilst we don't know the size of the vector we do know the maximum size so
+ // can perform a tree reduction with an identity vector, which means once we
+ // arrive at the result the remaining stages (when the vector is smaller than
+ // the maximum) have no affect.
+
+ unsigned Segments = AArch64::SVEMaxBitsPerVector / AArch64::SVEBitsPerBlock;
+ unsigned Stages = llvm::Log2_32(Segments * SrcVT.getVectorMinNumElements());
+
+ for (unsigned I = 0; I < Stages; ++I) {
+ Src = DAG.getNode(ISD::VECTOR_DEINTERLEAVE, DL, SrcVTs, Src, Identity);
+ Src = DAG.getNode(BaseOpc, DL, SrcVT, Src.getValue(0), Src.getValue(1));
+ }
+
+ return DAG.getExtractVectorElt(DL, Op.getValueType(), Src, 0);
+}
+
SDValue AArch64TargetLowering::LowerATOMIC_LOAD_AND(SDValue Op,
SelectionDAG &DAG) const {
auto &Subtarget = DAG.getSubtarget<AArch64Subtarget>();
@@ -18144,8 +18176,8 @@ bool AArch64TargetLowering::lowerInterleavedStore(Instruction *Store,
bool AArch64TargetLowering::lowerDeinterleaveIntrinsicToLoad(
Instruction *Load, Value *Mask, IntrinsicInst *DI) const {
const unsigned Factor = getDeinterleaveIntrinsicFactor(DI->getIntrinsicID());
- if (Factor != 2 && Factor != 4) {
- LLVM_DEBUG(dbgs() << "Matching ld2 and ld4 patterns failed\n");
+ if (Factor != 2 && Factor != 3 && Factor != 4) {
+ LLVM_DEBUG(dbgs() << "Matching ld2, ld3 and ld4 patterns failed\n");
return false;
}
auto *LI = dyn_cast<LoadInst>(Load);
@@ -18223,8 +18255,8 @@ bool AArch64TargetLowering::lowerInterleaveIntrinsicToStore(
Instruction *Store, Value *Mask,
ArrayRef<Value *> InterleavedValues) const {
unsigned Factor = InterleavedValues.size();
- if (Factor != 2 && Factor != 4) {
- LLVM_DEBUG(dbgs() << "Matching st2 and st4 patterns failed\n");
+ if (Factor != 2 && Factor != 3 && Factor != 4) {
+ LLVM_DEBUG(dbgs() << "Matching st2, st3 and st4 patterns failed\n");
return false;
}
StoreInst *SI = dyn_cast<StoreInst>(Store);
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.h b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
index 00956fd..9495c9f 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
@@ -752,6 +752,7 @@ private:
SDValue LowerVSCALE(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerTRUNCATE(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerVECREDUCE(SDValue Op, SelectionDAG &DAG) const;
+ SDValue LowerVECREDUCE_MUL(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerATOMIC_LOAD_AND(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerWindowsDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerInlineDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG) const;
diff --git a/llvm/lib/Target/AArch64/AArch64InstrGISel.td b/llvm/lib/Target/AArch64/AArch64InstrGISel.td
index 7322212..fe84193 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrGISel.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrGISel.td
@@ -233,6 +233,12 @@ def G_SDOT : AArch64GenericInstruction {
let hasSideEffects = 0;
}
+def G_USDOT : AArch64GenericInstruction {
+ let OutOperandList = (outs type0:$dst);
+ let InOperandList = (ins type0:$src1, type0:$src2, type0:$src3);
+ let hasSideEffects = 0;
+}
+
// Generic instruction for the BSP pseudo. It is expanded into BSP, which
// expands into BSL/BIT/BIF after register allocation.
def G_BSP : AArch64GenericInstruction {
@@ -278,6 +284,7 @@ def : GINodeEquiv<G_UADDLV, AArch64uaddlv>;
def : GINodeEquiv<G_UDOT, AArch64udot>;
def : GINodeEquiv<G_SDOT, AArch64sdot>;
+def : GINodeEquiv<G_USDOT, AArch64usdot>;
def : GINodeEquiv<G_EXTRACT_VECTOR_ELT, vector_extract>;
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index 30dfcf2b..12c600f 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -10600,6 +10600,9 @@ describeORRLoadedValue(const MachineInstr &MI, Register DescribedReg,
Register DestReg = DestSrc->Destination->getReg();
Register SrcReg = DestSrc->Source->getReg();
+ if (!DestReg.isValid() || !SrcReg.isValid())
+ return std::nullopt;
+
auto Expr = DIExpression::get(MI.getMF()->getFunction().getContext(), {});
// If the described register is the destination, just return the source.
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
index 9e2d698..05a4313 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
@@ -1855,6 +1855,8 @@ bool AArch64LegalizerInfo::legalizeIntrinsic(LegalizerHelper &Helper,
return LowerTriOp(AArch64::G_UDOT);
case Intrinsic::aarch64_neon_sdot:
return LowerTriOp(AArch64::G_SDOT);
+ case Intrinsic::aarch64_neon_usdot:
+ return LowerTriOp(AArch64::G_USDOT);
case Intrinsic::aarch64_neon_sqxtn:
return LowerUnaryOp(TargetOpcode::G_TRUNC_SSAT_S);
case Intrinsic::aarch64_neon_sqxtun:
diff --git a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
index bfe2c80..a67b12a 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPURegBankLegalizeRules.cpp
@@ -901,6 +901,8 @@ RegBankLegalizeRules::RegBankLegalizeRules(const GCNSubtarget &_ST,
addRulesForGOpcs({G_ABS}, Standard).Uni(S16, {{Sgpr32Trunc}, {Sgpr32SExt}});
+ addRulesForGOpcs({G_READSTEADYCOUNTER}, Standard).Uni(S64, {{Sgpr64}, {}});
+
bool hasSALUFloat = ST->hasSALUFloatInsts();
addRulesForGOpcs({G_FADD}, Standard)
diff --git a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
index be62395..e0375ea 100644
--- a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
+++ b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
@@ -513,8 +513,7 @@ void AMDGPUDisassembler::decodeImmOperands(MCInst &MI,
}
if (Imm == AMDGPU::EncValues::LITERAL_CONST) {
- Op = decodeLiteralConstant(
- Desc, OpDesc, OpDesc.OperandType == AMDGPU::OPERAND_REG_IMM_FP64);
+ Op = decodeLiteralConstant(Desc, OpDesc);
continue;
}
@@ -1545,21 +1544,21 @@ AMDGPUDisassembler::decodeMandatoryLiteralConstant(unsigned Val) const {
MCOperand
AMDGPUDisassembler::decodeMandatoryLiteral64Constant(uint64_t Val) const {
if (HasLiteral) {
- if (Literal64 != Val)
+ if (Literal != Val)
return errOperand(Val, "More than one unique literal is illegal");
}
HasLiteral = true;
- Literal = Literal64 = Val;
+ Literal = Val;
- bool UseLit64 = Hi_32(Literal64) == 0;
+ bool UseLit64 = Hi_32(Literal) == 0;
return UseLit64 ? MCOperand::createExpr(AMDGPUMCExpr::createLit(
- LitModifier::Lit64, Literal64, getContext()))
- : MCOperand::createImm(Literal64);
+ LitModifier::Lit64, Literal, getContext()))
+ : MCOperand::createImm(Literal);
}
-MCOperand AMDGPUDisassembler::decodeLiteralConstant(const MCInstrDesc &Desc,
- const MCOperandInfo &OpDesc,
- bool ExtendFP64) const {
+MCOperand
+AMDGPUDisassembler::decodeLiteralConstant(const MCInstrDesc &Desc,
+ const MCOperandInfo &OpDesc) const {
// For now all literal constants are supposed to be unsigned integer
// ToDo: deal with signed/unsigned 64-bit integer constants
// ToDo: deal with float/double constants
@@ -1569,35 +1568,79 @@ MCOperand AMDGPUDisassembler::decodeLiteralConstant(const MCInstrDesc &Desc,
Twine(Bytes.size()));
}
HasLiteral = true;
- Literal = Literal64 = eatBytes<uint32_t>(Bytes);
- if (ExtendFP64)
- Literal64 <<= 32;
+ Literal = eatBytes<uint32_t>(Bytes);
}
- int64_t Val = ExtendFP64 ? Literal64 : Literal;
+ // For disassembling always assume all inline constants are available.
+ bool HasInv2Pi = true;
- bool CanUse64BitLiterals =
- STI.hasFeature(AMDGPU::Feature64BitLiterals) &&
- !(Desc.TSFlags & (SIInstrFlags::VOP3 | SIInstrFlags::VOP3P));
-
- bool UseLit64 = false;
- if (CanUse64BitLiterals) {
- if (OpDesc.OperandType == AMDGPU::OPERAND_REG_IMM_INT64 ||
- OpDesc.OperandType == AMDGPU::OPERAND_REG_INLINE_C_INT64)
- UseLit64 = false;
- else if (OpDesc.OperandType == AMDGPU::OPERAND_REG_IMM_FP64 ||
- OpDesc.OperandType == AMDGPU::OPERAND_REG_INLINE_C_FP64 ||
- OpDesc.OperandType == AMDGPU::OPERAND_REG_INLINE_AC_FP64)
- UseLit64 = Hi_32(Literal64) == 0;
+ // Invalid instruction codes may contain literals for inline-only
+ // operands, so we support them here as well.
+ int64_t Val = Literal;
+ bool UseLit = false;
+ switch (OpDesc.OperandType) {
+ default:
+ llvm_unreachable("Unexpected operand type!");
+ case AMDGPU::OPERAND_REG_IMM_BF16:
+ case AMDGPU::OPERAND_REG_INLINE_C_BF16:
+ case AMDGPU::OPERAND_REG_INLINE_C_V2BF16:
+ UseLit = AMDGPU::isInlinableLiteralBF16(Val, HasInv2Pi);
+ break;
+ case AMDGPU::OPERAND_REG_IMM_V2BF16:
+ UseLit = AMDGPU::isInlinableLiteralV2BF16(Val);
+ break;
+ case AMDGPU::OPERAND_REG_IMM_FP16:
+ case AMDGPU::OPERAND_REG_INLINE_C_FP16:
+ case AMDGPU::OPERAND_REG_INLINE_C_V2FP16:
+ UseLit = AMDGPU::isInlinableLiteralFP16(Val, HasInv2Pi);
+ break;
+ case AMDGPU::OPERAND_REG_IMM_V2FP16:
+ UseLit = AMDGPU::isInlinableLiteralV2F16(Val);
+ break;
+ case AMDGPU::OPERAND_REG_IMM_NOINLINE_V2FP16:
+ break;
+ case AMDGPU::OPERAND_REG_IMM_INT16:
+ case AMDGPU::OPERAND_REG_INLINE_C_INT16:
+ case AMDGPU::OPERAND_REG_INLINE_C_V2INT16:
+ UseLit = AMDGPU::isInlinableLiteralI16(Val, HasInv2Pi);
+ break;
+ case AMDGPU::OPERAND_REG_IMM_V2INT16:
+ UseLit = AMDGPU::isInlinableLiteralV2I16(Val);
+ break;
+ case AMDGPU::OPERAND_REG_IMM_FP32:
+ case AMDGPU::OPERAND_REG_INLINE_C_FP32:
+ case AMDGPU::OPERAND_REG_INLINE_AC_FP32:
+ case AMDGPU::OPERAND_REG_IMM_INT32:
+ case AMDGPU::OPERAND_REG_INLINE_C_INT32:
+ case AMDGPU::OPERAND_REG_INLINE_AC_INT32:
+ case AMDGPU::OPERAND_REG_IMM_V2FP32:
+ case AMDGPU::OPERAND_REG_IMM_V2INT32:
+ case AMDGPU::OPERAND_KIMM32:
+ UseLit = AMDGPU::isInlinableLiteral32(Val, HasInv2Pi);
+ break;
+ case AMDGPU::OPERAND_REG_IMM_FP64:
+ case AMDGPU::OPERAND_REG_INLINE_C_FP64:
+ case AMDGPU::OPERAND_REG_INLINE_AC_FP64:
+ Val <<= 32;
+ break;
+ case AMDGPU::OPERAND_REG_IMM_INT64:
+ case AMDGPU::OPERAND_REG_INLINE_C_INT64:
+ UseLit = AMDGPU::isInlinableLiteral64(Val, HasInv2Pi);
+ break;
+ case MCOI::OPERAND_REGISTER:
+ // TODO: Disassembling V_DUAL_FMAMK_F32_X_FMAMK_F32_gfx11 hits
+ // decoding a literal in a position of a register operand. Give
+ // it special handling in the caller, decodeImmOperands(), instead
+ // of quietly allowing it here.
+ break;
}
- return UseLit64 ? MCOperand::createExpr(AMDGPUMCExpr::createLit(
- LitModifier::Lit64, Val, getContext()))
- : MCOperand::createImm(Val);
+ return UseLit ? MCOperand::createExpr(AMDGPUMCExpr::createLit(
+ LitModifier::Lit, Val, getContext()))
+ : MCOperand::createImm(Val);
}
-MCOperand
-AMDGPUDisassembler::decodeLiteral64Constant(const MCInst &Inst) const {
+MCOperand AMDGPUDisassembler::decodeLiteral64Constant() const {
assert(STI.hasFeature(AMDGPU::Feature64BitLiterals));
if (!HasLiteral) {
@@ -1606,25 +1649,13 @@ AMDGPUDisassembler::decodeLiteral64Constant(const MCInst &Inst) const {
Twine(Bytes.size()));
}
HasLiteral = true;
- Literal64 = eatBytes<uint64_t>(Bytes);
- }
-
- bool UseLit64 = false;
- const MCInstrDesc &Desc = MCII->get(Inst.getOpcode());
- const MCOperandInfo &OpDesc = Desc.operands()[Inst.getNumOperands()];
- if (OpDesc.OperandType == AMDGPU::OPERAND_REG_IMM_INT64 ||
- OpDesc.OperandType == AMDGPU::OPERAND_REG_INLINE_C_INT64) {
- UseLit64 = false;
- } else {
- assert(OpDesc.OperandType == AMDGPU::OPERAND_REG_IMM_FP64 ||
- OpDesc.OperandType == AMDGPU::OPERAND_REG_INLINE_C_FP64 ||
- OpDesc.OperandType == AMDGPU::OPERAND_REG_INLINE_AC_FP64);
- UseLit64 = Hi_32(Literal64) == 0;
+ Literal = eatBytes<uint64_t>(Bytes);
}
+ bool UseLit64 = Hi_32(Literal) == 0;
return UseLit64 ? MCOperand::createExpr(AMDGPUMCExpr::createLit(
- LitModifier::Lit64, Literal64, getContext()))
- : MCOperand::createImm(Literal64);
+ LitModifier::Lit64, Literal, getContext()))
+ : MCOperand::createImm(Literal);
}
MCOperand AMDGPUDisassembler::decodeIntImmed(unsigned Imm) {
@@ -1913,7 +1944,7 @@ MCOperand AMDGPUDisassembler::decodeNonVGPRSrcOp(const MCInst &Inst,
return MCOperand::createImm(Val);
if (Val == LITERAL64_CONST && STI.hasFeature(AMDGPU::Feature64BitLiterals)) {
- return decodeLiteral64Constant(Inst);
+ return decodeLiteral64Constant();
}
switch (Width) {
diff --git a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.h b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.h
index 2751857..d103d79 100644
--- a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.h
+++ b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.h
@@ -44,8 +44,7 @@ private:
const unsigned HwModeRegClass;
const unsigned TargetMaxInstBytes;
mutable ArrayRef<uint8_t> Bytes;
- mutable uint32_t Literal;
- mutable uint64_t Literal64;
+ mutable uint64_t Literal;
mutable bool HasLiteral;
mutable std::optional<bool> EnableWavefrontSize32;
unsigned CodeObjectVersion;
@@ -144,9 +143,8 @@ public:
MCOperand decodeMandatoryLiteralConstant(unsigned Imm) const;
MCOperand decodeMandatoryLiteral64Constant(uint64_t Imm) const;
MCOperand decodeLiteralConstant(const MCInstrDesc &Desc,
- const MCOperandInfo &OpDesc,
- bool ExtendFP64) const;
- MCOperand decodeLiteral64Constant(const MCInst &Inst) const;
+ const MCOperandInfo &OpDesc) const;
+ MCOperand decodeLiteral64Constant() const;
MCOperand decodeSrcOp(const MCInst &Inst, unsigned Width, unsigned Val) const;
diff --git a/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp b/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp
index 58482ea..9fbf9e5 100644
--- a/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp
@@ -69,6 +69,12 @@ static cl::opt<bool> GCNTrackers(
cl::desc("Use the AMDGPU specific RPTrackers during scheduling"),
cl::init(false));
+static cl::opt<unsigned> PendingQueueLimit(
+ "amdgpu-scheduler-pending-queue-limit", cl::Hidden,
+ cl::desc(
+ "Max (Available+Pending) size to inspect pending queue (0 disables)"),
+ cl::init(256));
+
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
#define DUMP_MAX_REG_PRESSURE
static cl::opt<bool> PrintMaxRPRegUsageBeforeScheduler(
@@ -335,17 +341,52 @@ void GCNSchedStrategy::initCandidate(SchedCandidate &Cand, SUnit *SU,
}
}
+static bool shouldCheckPending(SchedBoundary &Zone,
+ const TargetSchedModel *SchedModel) {
+ bool HasBufferedModel =
+ SchedModel->hasInstrSchedModel() && SchedModel->getMicroOpBufferSize();
+ unsigned Combined = Zone.Available.size() + Zone.Pending.size();
+ return Combined <= PendingQueueLimit && HasBufferedModel;
+}
+
+static SUnit *pickOnlyChoice(SchedBoundary &Zone,
+ const TargetSchedModel *SchedModel) {
+ // pickOnlyChoice() releases pending instructions and checks for new hazards.
+ SUnit *OnlyChoice = Zone.pickOnlyChoice();
+ if (!shouldCheckPending(Zone, SchedModel) || Zone.Pending.empty())
+ return OnlyChoice;
+
+ return nullptr;
+}
+
+void GCNSchedStrategy::printCandidateDecision(const SchedCandidate &Current,
+ const SchedCandidate &Preferred) {
+ LLVM_DEBUG({
+ dbgs() << "Prefer:\t\t";
+ DAG->dumpNode(*Preferred.SU);
+
+ if (Current.SU) {
+ dbgs() << "Not:\t";
+ DAG->dumpNode(*Current.SU);
+ }
+
+ dbgs() << "Reason:\t\t";
+ traceCandidate(Preferred);
+ });
+}
+
// This function is mostly cut and pasted from
// GenericScheduler::pickNodeFromQueue()
void GCNSchedStrategy::pickNodeFromQueue(SchedBoundary &Zone,
const CandPolicy &ZonePolicy,
const RegPressureTracker &RPTracker,
- SchedCandidate &Cand,
+ SchedCandidate &Cand, bool &IsPending,
bool IsBottomUp) {
const SIRegisterInfo *SRI = static_cast<const SIRegisterInfo *>(TRI);
ArrayRef<unsigned> Pressure = RPTracker.getRegSetPressureAtPos();
unsigned SGPRPressure = 0;
unsigned VGPRPressure = 0;
+ IsPending = false;
if (DAG->isTrackingPressure()) {
if (!GCNTrackers) {
SGPRPressure = Pressure[AMDGPU::RegisterPressureSets::SReg_32];
@@ -358,8 +399,9 @@ void GCNSchedStrategy::pickNodeFromQueue(SchedBoundary &Zone,
VGPRPressure = T->getPressure().getArchVGPRNum();
}
}
- ReadyQueue &Q = Zone.Available;
- for (SUnit *SU : Q) {
+ LLVM_DEBUG(dbgs() << "Available Q:\n");
+ ReadyQueue &AQ = Zone.Available;
+ for (SUnit *SU : AQ) {
SchedCandidate TryCand(ZonePolicy);
initCandidate(TryCand, SU, Zone.isTop(), RPTracker, SRI, SGPRPressure,
@@ -371,27 +413,55 @@ void GCNSchedStrategy::pickNodeFromQueue(SchedBoundary &Zone,
// Initialize resource delta if needed in case future heuristics query it.
if (TryCand.ResDelta == SchedResourceDelta())
TryCand.initResourceDelta(Zone.DAG, SchedModel);
+ LLVM_DEBUG(printCandidateDecision(Cand, TryCand));
Cand.setBest(TryCand);
- LLVM_DEBUG(traceCandidate(Cand));
+ } else {
+ printCandidateDecision(TryCand, Cand);
+ }
+ }
+
+ if (!shouldCheckPending(Zone, SchedModel))
+ return;
+
+ LLVM_DEBUG(dbgs() << "Pending Q:\n");
+ ReadyQueue &PQ = Zone.Pending;
+ for (SUnit *SU : PQ) {
+
+ SchedCandidate TryCand(ZonePolicy);
+ initCandidate(TryCand, SU, Zone.isTop(), RPTracker, SRI, SGPRPressure,
+ VGPRPressure, IsBottomUp);
+ // Pass SchedBoundary only when comparing nodes from the same boundary.
+ SchedBoundary *ZoneArg = Cand.AtTop == TryCand.AtTop ? &Zone : nullptr;
+ tryPendingCandidate(Cand, TryCand, ZoneArg);
+ if (TryCand.Reason != NoCand) {
+ // Initialize resource delta if needed in case future heuristics query it.
+ if (TryCand.ResDelta == SchedResourceDelta())
+ TryCand.initResourceDelta(Zone.DAG, SchedModel);
+ LLVM_DEBUG(printCandidateDecision(Cand, TryCand));
+ IsPending = true;
+ Cand.setBest(TryCand);
+ } else {
+ printCandidateDecision(TryCand, Cand);
}
}
}
// This function is mostly cut and pasted from
// GenericScheduler::pickNodeBidirectional()
-SUnit *GCNSchedStrategy::pickNodeBidirectional(bool &IsTopNode) {
+SUnit *GCNSchedStrategy::pickNodeBidirectional(bool &IsTopNode,
+ bool &PickedPending) {
// Schedule as far as possible in the direction of no choice. This is most
// efficient, but also provides the best heuristics for CriticalPSets.
- if (SUnit *SU = Bot.pickOnlyChoice()) {
+ if (SUnit *SU = pickOnlyChoice(Bot, SchedModel)) {
IsTopNode = false;
return SU;
}
- if (SUnit *SU = Top.pickOnlyChoice()) {
+ if (SUnit *SU = pickOnlyChoice(Top, SchedModel)) {
IsTopNode = true;
return SU;
}
- // Set the bottom-up policy based on the state of the current bottom zone and
- // the instructions outside the zone, including the top zone.
+ // Set the bottom-up policy based on the state of the current bottom zone
+ // and the instructions outside the zone, including the top zone.
CandPolicy BotPolicy;
setPolicy(BotPolicy, /*IsPostRA=*/false, Bot, &Top);
// Set the top-down policy based on the state of the current top zone and
@@ -399,12 +469,14 @@ SUnit *GCNSchedStrategy::pickNodeBidirectional(bool &IsTopNode) {
CandPolicy TopPolicy;
setPolicy(TopPolicy, /*IsPostRA=*/false, Top, &Bot);
+ bool BotPending = false;
// See if BotCand is still valid (because we previously scheduled from Top).
LLVM_DEBUG(dbgs() << "Picking from Bot:\n");
if (!BotCand.isValid() || BotCand.SU->isScheduled ||
BotCand.Policy != BotPolicy) {
BotCand.reset(CandPolicy());
pickNodeFromQueue(Bot, BotPolicy, DAG->getBotRPTracker(), BotCand,
+ BotPending,
/*IsBottomUp=*/true);
assert(BotCand.Reason != NoCand && "failed to find the first candidate");
} else {
@@ -414,6 +486,7 @@ SUnit *GCNSchedStrategy::pickNodeBidirectional(bool &IsTopNode) {
SchedCandidate TCand;
TCand.reset(CandPolicy());
pickNodeFromQueue(Bot, BotPolicy, DAG->getBotRPTracker(), TCand,
+ BotPending,
/*IsBottomUp=*/true);
assert(TCand.SU == BotCand.SU &&
"Last pick result should correspond to re-picking right now");
@@ -421,12 +494,14 @@ SUnit *GCNSchedStrategy::pickNodeBidirectional(bool &IsTopNode) {
#endif
}
+ bool TopPending = false;
// Check if the top Q has a better candidate.
LLVM_DEBUG(dbgs() << "Picking from Top:\n");
if (!TopCand.isValid() || TopCand.SU->isScheduled ||
TopCand.Policy != TopPolicy) {
TopCand.reset(CandPolicy());
pickNodeFromQueue(Top, TopPolicy, DAG->getTopRPTracker(), TopCand,
+ TopPending,
/*IsBottomUp=*/false);
assert(TopCand.Reason != NoCand && "failed to find the first candidate");
} else {
@@ -436,6 +511,7 @@ SUnit *GCNSchedStrategy::pickNodeBidirectional(bool &IsTopNode) {
SchedCandidate TCand;
TCand.reset(CandPolicy());
pickNodeFromQueue(Top, TopPolicy, DAG->getTopRPTracker(), TCand,
+ TopPending,
/*IsBottomUp=*/false);
assert(TCand.SU == TopCand.SU &&
"Last pick result should correspond to re-picking right now");
@@ -446,12 +522,21 @@ SUnit *GCNSchedStrategy::pickNodeBidirectional(bool &IsTopNode) {
// Pick best from BotCand and TopCand.
LLVM_DEBUG(dbgs() << "Top Cand: "; traceCandidate(TopCand);
dbgs() << "Bot Cand: "; traceCandidate(BotCand););
- SchedCandidate Cand = BotCand;
- TopCand.Reason = NoCand;
- tryCandidate(Cand, TopCand, nullptr);
- if (TopCand.Reason != NoCand) {
- Cand.setBest(TopCand);
+ SchedCandidate Cand = BotPending ? TopCand : BotCand;
+ SchedCandidate TryCand = BotPending ? BotCand : TopCand;
+ PickedPending = BotPending && TopPending;
+
+ TryCand.Reason = NoCand;
+ if (BotPending || TopPending) {
+ PickedPending |= tryPendingCandidate(Cand, TopCand, nullptr);
+ } else {
+ tryCandidate(Cand, TryCand, nullptr);
+ }
+
+ if (TryCand.Reason != NoCand) {
+ Cand.setBest(TryCand);
}
+
LLVM_DEBUG(dbgs() << "Picking: "; traceCandidate(Cand););
IsTopNode = Cand.AtTop;
@@ -466,35 +551,55 @@ SUnit *GCNSchedStrategy::pickNode(bool &IsTopNode) {
Bot.Available.empty() && Bot.Pending.empty() && "ReadyQ garbage");
return nullptr;
}
+ bool PickedPending;
SUnit *SU;
do {
+ PickedPending = false;
if (RegionPolicy.OnlyTopDown) {
- SU = Top.pickOnlyChoice();
+ SU = pickOnlyChoice(Top, SchedModel);
if (!SU) {
CandPolicy NoPolicy;
TopCand.reset(NoPolicy);
pickNodeFromQueue(Top, NoPolicy, DAG->getTopRPTracker(), TopCand,
+ PickedPending,
/*IsBottomUp=*/false);
assert(TopCand.Reason != NoCand && "failed to find a candidate");
SU = TopCand.SU;
}
IsTopNode = true;
} else if (RegionPolicy.OnlyBottomUp) {
- SU = Bot.pickOnlyChoice();
+ SU = pickOnlyChoice(Bot, SchedModel);
if (!SU) {
CandPolicy NoPolicy;
BotCand.reset(NoPolicy);
pickNodeFromQueue(Bot, NoPolicy, DAG->getBotRPTracker(), BotCand,
+ PickedPending,
/*IsBottomUp=*/true);
assert(BotCand.Reason != NoCand && "failed to find a candidate");
SU = BotCand.SU;
}
IsTopNode = false;
} else {
- SU = pickNodeBidirectional(IsTopNode);
+ SU = pickNodeBidirectional(IsTopNode, PickedPending);
}
} while (SU->isScheduled);
+ if (PickedPending) {
+ unsigned ReadyCycle = IsTopNode ? SU->TopReadyCycle : SU->BotReadyCycle;
+ SchedBoundary &Zone = IsTopNode ? Top : Bot;
+ unsigned CurrentCycle = Zone.getCurrCycle();
+ if (ReadyCycle > CurrentCycle)
+ Zone.bumpCycle(ReadyCycle);
+
+ // FIXME: checkHazard() doesn't give information about which cycle the
+ // hazard will resolve so just keep bumping the cycle by 1. This could be
+ // made more efficient if checkHazard() returned more details.
+ while (Zone.checkHazard(SU))
+ Zone.bumpCycle(Zone.getCurrCycle() + 1);
+
+ Zone.releasePending();
+ }
+
if (SU->isTopReady())
Top.removeReady(SU);
if (SU->isBottomReady())
@@ -540,6 +645,47 @@ GCNSchedStageID GCNSchedStrategy::getNextStage() const {
return *std::next(CurrentStage);
}
+bool GCNSchedStrategy::tryPendingCandidate(SchedCandidate &Cand,
+ SchedCandidate &TryCand,
+ SchedBoundary *Zone) const {
+ // Initialize the candidate if needed.
+ if (!Cand.isValid()) {
+ TryCand.Reason = NodeOrder;
+ return true;
+ }
+
+ // Bias PhysReg Defs and copies to their uses and defined respectively.
+ if (tryGreater(biasPhysReg(TryCand.SU, TryCand.AtTop),
+ biasPhysReg(Cand.SU, Cand.AtTop), TryCand, Cand, PhysReg))
+ return TryCand.Reason != NoCand;
+
+ // Avoid exceeding the target's limit.
+ if (DAG->isTrackingPressure() &&
+ tryPressure(TryCand.RPDelta.Excess, Cand.RPDelta.Excess, TryCand, Cand,
+ RegExcess, TRI, DAG->MF))
+ return TryCand.Reason != NoCand;
+
+ // Avoid increasing the max critical pressure in the scheduled region.
+ if (DAG->isTrackingPressure() &&
+ tryPressure(TryCand.RPDelta.CriticalMax, Cand.RPDelta.CriticalMax,
+ TryCand, Cand, RegCritical, TRI, DAG->MF))
+ return TryCand.Reason != NoCand;
+
+ bool SameBoundary = Zone != nullptr;
+ if (SameBoundary) {
+ TryCand.initResourceDelta(DAG, SchedModel);
+ if (tryLess(TryCand.ResDelta.CritResources, Cand.ResDelta.CritResources,
+ TryCand, Cand, ResourceReduce))
+ return TryCand.Reason != NoCand;
+ if (tryGreater(TryCand.ResDelta.DemandedResources,
+ Cand.ResDelta.DemandedResources, TryCand, Cand,
+ ResourceDemand))
+ return TryCand.Reason != NoCand;
+ }
+
+ return false;
+}
+
GCNMaxOccupancySchedStrategy::GCNMaxOccupancySchedStrategy(
const MachineSchedContext *C, bool IsLegacyScheduler)
: GCNSchedStrategy(C) {
diff --git a/llvm/lib/Target/AMDGPU/GCNSchedStrategy.h b/llvm/lib/Target/AMDGPU/GCNSchedStrategy.h
index 8ea4267..975781f 100644
--- a/llvm/lib/Target/AMDGPU/GCNSchedStrategy.h
+++ b/llvm/lib/Target/AMDGPU/GCNSchedStrategy.h
@@ -44,17 +44,32 @@ raw_ostream &operator<<(raw_ostream &OS, const GCNSchedStageID &StageID);
/// heuristics to determine excess/critical pressure sets.
class GCNSchedStrategy : public GenericScheduler {
protected:
- SUnit *pickNodeBidirectional(bool &IsTopNode);
+ SUnit *pickNodeBidirectional(bool &IsTopNode, bool &PickedPending);
void pickNodeFromQueue(SchedBoundary &Zone, const CandPolicy &ZonePolicy,
const RegPressureTracker &RPTracker,
- SchedCandidate &Cand, bool IsBottomUp);
+ SchedCandidate &Cand, bool &IsPending,
+ bool IsBottomUp);
void initCandidate(SchedCandidate &Cand, SUnit *SU, bool AtTop,
const RegPressureTracker &RPTracker,
const SIRegisterInfo *SRI, unsigned SGPRPressure,
unsigned VGPRPressure, bool IsBottomUp);
+ /// Evaluates instructions in the pending queue using a subset of scheduling
+ /// heuristics.
+ ///
+ /// Instructions that cannot be issued due to hardware constraints are placed
+ /// in the pending queue rather than the available queue, making them normally
+ /// invisible to scheduling heuristics. However, in certain scenarios (such as
+ /// avoiding register spilling), it may be beneficial to consider scheduling
+ /// these not-yet-ready instructions.
+ bool tryPendingCandidate(SchedCandidate &Cand, SchedCandidate &TryCand,
+ SchedBoundary *Zone) const;
+
+ void printCandidateDecision(const SchedCandidate &Current,
+ const SchedCandidate &Preferred);
+
std::vector<unsigned> Pressure;
std::vector<unsigned> MaxPressure;
diff --git a/llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp b/llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp
index c684f9e..01a40c1 100644
--- a/llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp
+++ b/llvm/lib/Target/AMDGPU/SIPreEmitPeephole.cpp
@@ -654,7 +654,6 @@ void SIPreEmitPeephole::collectUnpackingCandidates(
if (TotalCyclesBetweenCandidates < NumMFMACycles - 1)
InstrsToUnpack.insert(&Instr);
}
- return;
}
void SIPreEmitPeephole::performF32Unpacking(MachineInstr &I) {
@@ -681,7 +680,6 @@ void SIPreEmitPeephole::performF32Unpacking(MachineInstr &I) {
HiDstOp.setIsRenamable(DstOp.isRenamable());
I.eraseFromParent();
- return;
}
MachineInstrBuilder SIPreEmitPeephole::createUnpackedMI(MachineInstr &I,
diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
index a01a5fd..5e3195b 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -1697,9 +1697,6 @@ LLVM_READNONE
bool isInlinableLiteralFP16(int16_t Literal, bool HasInv2Pi);
LLVM_READNONE
-bool isInlinableLiteralBF16(int16_t Literal, bool HasInv2Pi);
-
-LLVM_READNONE
bool isInlinableLiteralI16(int32_t Literal, bool HasInv2Pi);
LLVM_READNONE
diff --git a/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp b/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp
index 9945ecc..0d7b6d1 100644
--- a/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp
+++ b/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp
@@ -161,8 +161,8 @@ namespace {
friend bool operator<(const NEONLdStTableEntry &TE, unsigned PseudoOpc) {
return TE.PseudoOpc < PseudoOpc;
}
- friend bool LLVM_ATTRIBUTE_UNUSED operator<(unsigned PseudoOpc,
- const NEONLdStTableEntry &TE) {
+ [[maybe_unused]] friend bool operator<(unsigned PseudoOpc,
+ const NEONLdStTableEntry &TE) {
return PseudoOpc < TE.PseudoOpc;
}
};
diff --git a/llvm/lib/Target/DirectX/DXIL.td b/llvm/lib/Target/DirectX/DXIL.td
index 228114c..44c4830 100644
--- a/llvm/lib/Target/DirectX/DXIL.td
+++ b/llvm/lib/Target/DirectX/DXIL.td
@@ -57,6 +57,7 @@ def ResBindTy : DXILOpParamType;
def ResPropsTy : DXILOpParamType;
def SplitDoubleTy : DXILOpParamType;
def BinaryWithCarryTy : DXILOpParamType;
+def DimensionsTy : DXILOpParamType;
class DXILOpClass;
@@ -901,6 +902,13 @@ def CheckAccessFullyMapped : DXILOp<71, checkAccessFullyMapped> {
let attributes = [Attributes<DXIL1_0, [ReadOnly]>];
}
+def GetDimensions : DXILOp<72, getDimensions> {
+ let Doc = "gets the dimensions of a buffer or texture";
+ let arguments = [HandleTy, Int32Ty];
+ let result = DimensionsTy;
+ let stages = [Stages<DXIL1_0, [all_stages]>];
+}
+
def Barrier : DXILOp<80, barrier> {
let Doc = "inserts a memory barrier in the shader";
let intrinsics = [
diff --git a/llvm/lib/Target/DirectX/DXILOpBuilder.cpp b/llvm/lib/Target/DirectX/DXILOpBuilder.cpp
index 1aed8f9..944b2e6 100644
--- a/llvm/lib/Target/DirectX/DXILOpBuilder.cpp
+++ b/llvm/lib/Target/DirectX/DXILOpBuilder.cpp
@@ -261,6 +261,12 @@ static StructType *getBinaryWithCarryType(LLVMContext &Context) {
return StructType::create({Int32Ty, Int1Ty}, "dx.types.i32c");
}
+static StructType *getDimensionsType(LLVMContext &Ctx) {
+ Type *Int32Ty = Type::getInt32Ty(Ctx);
+ return getOrCreateStructType("dx.types.Dimensions",
+ {Int32Ty, Int32Ty, Int32Ty, Int32Ty}, Ctx);
+}
+
static Type *getTypeFromOpParamType(OpParamType Kind, LLVMContext &Ctx,
Type *OverloadTy) {
switch (Kind) {
@@ -318,6 +324,8 @@ static Type *getTypeFromOpParamType(OpParamType Kind, LLVMContext &Ctx,
return getSplitDoubleType(Ctx);
case OpParamType::BinaryWithCarryTy:
return getBinaryWithCarryType(Ctx);
+ case OpParamType::DimensionsTy:
+ return getDimensionsType(Ctx);
}
llvm_unreachable("Invalid parameter kind");
return nullptr;
diff --git a/llvm/lib/Target/DirectX/DXILOpLowering.cpp b/llvm/lib/Target/DirectX/DXILOpLowering.cpp
index 610d8b6..e46a393 100644
--- a/llvm/lib/Target/DirectX/DXILOpLowering.cpp
+++ b/llvm/lib/Target/DirectX/DXILOpLowering.cpp
@@ -627,6 +627,28 @@ public:
});
}
+ [[nodiscard]] bool lowerGetDimensionsX(Function &F) {
+ IRBuilder<> &IRB = OpBuilder.getIRB();
+ Type *Int32Ty = IRB.getInt32Ty();
+
+ return replaceFunction(F, [&](CallInst *CI) -> Error {
+ IRB.SetInsertPoint(CI);
+ Value *Handle =
+ createTmpHandleCast(CI->getArgOperand(0), OpBuilder.getHandleType());
+ Value *Undef = UndefValue::get(Int32Ty);
+
+ Expected<CallInst *> OpCall = OpBuilder.tryCreateOp(
+ OpCode::GetDimensions, {Handle, Undef}, CI->getName(), Int32Ty);
+ if (Error E = OpCall.takeError())
+ return E;
+ Value *Dim = IRB.CreateExtractValue(*OpCall, 0);
+
+ CI->replaceAllUsesWith(Dim);
+ CI->eraseFromParent();
+ return Error::success();
+ });
+ }
+
[[nodiscard]] bool lowerGetPointer(Function &F) {
// These should have already been handled in DXILResourceAccess, so we can
// just clean up the dead prototype.
@@ -934,6 +956,9 @@ public:
case Intrinsic::dx_resource_updatecounter:
HasErrors |= lowerUpdateCounter(F);
break;
+ case Intrinsic::dx_resource_getdimensions_x:
+ HasErrors |= lowerGetDimensionsX(F);
+ break;
case Intrinsic::ctpop:
HasErrors |= lowerCtpopToCountBits(F);
break;
diff --git a/llvm/lib/Target/DirectX/DXILWriter/DXILBitcodeWriter.cpp b/llvm/lib/Target/DirectX/DXILWriter/DXILBitcodeWriter.cpp
index 82c43ff..26a8728 100644
--- a/llvm/lib/Target/DirectX/DXILWriter/DXILBitcodeWriter.cpp
+++ b/llvm/lib/Target/DirectX/DXILWriter/DXILBitcodeWriter.cpp
@@ -1165,12 +1165,15 @@ void DXILBitcodeWriter::writeValueSymbolTableForwardDecl() {}
/// Returns the bit offset to backpatch with the location of the real VST.
void DXILBitcodeWriter::writeModuleInfo() {
// Emit various pieces of data attached to a module.
- if (!M.getTargetTriple().empty())
- writeStringRecord(Stream, bitc::MODULE_CODE_TRIPLE,
- M.getTargetTriple().str(), 0 /*TODO*/);
- const std::string &DL = M.getDataLayoutStr();
- if (!DL.empty())
- writeStringRecord(Stream, bitc::MODULE_CODE_DATALAYOUT, DL, 0 /*TODO*/);
+
+ // We need to hardcode a triple and datalayout that's compatible with the
+ // historical DXIL triple and datalayout from DXC.
+ StringRef Triple = "dxil-ms-dx";
+ StringRef DL = "e-m:e-p:32:32-i1:8-i8:8-i16:32-i32:32-i64:64-"
+ "f16:32-f32:32-f64:64-n8:16:32:64";
+ writeStringRecord(Stream, bitc::MODULE_CODE_TRIPLE, Triple, 0 /*TODO*/);
+ writeStringRecord(Stream, bitc::MODULE_CODE_DATALAYOUT, DL, 0 /*TODO*/);
+
if (!M.getModuleInlineAsm().empty())
writeStringRecord(Stream, bitc::MODULE_CODE_ASM, M.getModuleInlineAsm(),
0 /*TODO*/);
diff --git a/llvm/lib/Target/DirectX/DXILWriter/DXILWriterPass.cpp b/llvm/lib/Target/DirectX/DXILWriter/DXILWriterPass.cpp
index 1eb03bf..725f2b1 100644
--- a/llvm/lib/Target/DirectX/DXILWriter/DXILWriterPass.cpp
+++ b/llvm/lib/Target/DirectX/DXILWriter/DXILWriterPass.cpp
@@ -149,11 +149,6 @@ public:
std::string Data;
llvm::raw_string_ostream OS(Data);
- Triple OriginalTriple = M.getTargetTriple();
- // Set to DXIL triple when write to bitcode.
- // Only the output bitcode need to be DXIL triple.
- M.setTargetTriple(Triple("dxil-ms-dx"));
-
// Perform late legalization of lifetime intrinsics that would otherwise
// fail the Module Verifier if performed in an earlier pass
legalizeLifetimeIntrinsics(M);
@@ -165,9 +160,6 @@ public:
// not-so-legal legalizations
removeLifetimeIntrinsics(M);
- // Recover triple.
- M.setTargetTriple(OriginalTriple);
-
Constant *ModuleConstant =
ConstantDataArray::get(M.getContext(), arrayRefFromStringRef(Data));
auto *GV = new llvm::GlobalVariable(M, ModuleConstant->getType(), true,
diff --git a/llvm/lib/Target/Hexagon/Disassembler/HexagonDisassembler.cpp b/llvm/lib/Target/Hexagon/Disassembler/HexagonDisassembler.cpp
index 5f180d6..3bd6ed4 100644
--- a/llvm/lib/Target/Hexagon/Disassembler/HexagonDisassembler.cpp
+++ b/llvm/lib/Target/Hexagon/Disassembler/HexagonDisassembler.cpp
@@ -66,6 +66,10 @@ public:
void remapInstruction(MCInst &Instr) const;
+ Expected<bool> onSymbolStart(SymbolInfoTy &Symbol, uint64_t &Size,
+ ArrayRef<uint8_t> Bytes,
+ uint64_t Address) const override;
+
private:
bool makeBundle(ArrayRef<uint8_t> Bytes, uint64_t Address,
uint64_t &BytesToSkip, raw_ostream &CS) const;
@@ -567,6 +571,18 @@ DecodeStatus HexagonDisassembler::getSingleInstruction(MCInst &MI, MCInst &MCB,
return Result;
}
+Expected<bool> HexagonDisassembler::onSymbolStart(SymbolInfoTy &Symbol,
+ uint64_t &Size,
+ ArrayRef<uint8_t> Bytes,
+ uint64_t Address) const {
+ // At the start of a symbol, force a fresh packet by resetting any
+ // in-progress bundle state. This prevents packets from straddling label
+ // boundaries when data (e.g. jump tables) appears in between.
+ Size = 0;
+ resetBundle();
+ return true;
+}
+
static DecodeStatus DecodeRegisterClass(MCInst &Inst, unsigned RegNo,
ArrayRef<MCPhysReg> Table) {
if (RegNo < Table.size()) {
diff --git a/llvm/lib/Target/Hexagon/HexagonBitSimplify.cpp b/llvm/lib/Target/Hexagon/HexagonBitSimplify.cpp
index 52e6b0b..68f5312 100644
--- a/llvm/lib/Target/Hexagon/HexagonBitSimplify.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonBitSimplify.cpp
@@ -174,8 +174,8 @@ namespace {
const TargetRegisterInfo *TRI;
};
- raw_ostream &operator<< (raw_ostream &OS, const PrintRegSet &P)
- LLVM_ATTRIBUTE_UNUSED;
+ [[maybe_unused]] raw_ostream &operator<<(raw_ostream &OS,
+ const PrintRegSet &P);
raw_ostream &operator<< (raw_ostream &OS, const PrintRegSet &P) {
OS << '{';
for (unsigned R = P.RS.find_first(); R; R = P.RS.find_next(R))
diff --git a/llvm/lib/Target/Hexagon/HexagonCommonGEP.cpp b/llvm/lib/Target/Hexagon/HexagonCommonGEP.cpp
index 14b6bb3..9087f9d 100644
--- a/llvm/lib/Target/Hexagon/HexagonCommonGEP.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonCommonGEP.cpp
@@ -272,15 +272,14 @@ namespace {
OS << *I << ' ' << **I << '\n';
}
- raw_ostream &operator<< (raw_ostream &OS,
- const NodeVect &S) LLVM_ATTRIBUTE_UNUSED;
+ [[maybe_unused]] raw_ostream &operator<<(raw_ostream &OS, const NodeVect &S);
raw_ostream &operator<< (raw_ostream &OS, const NodeVect &S) {
dump_node_container(OS, S);
return OS;
}
- raw_ostream &operator<< (raw_ostream &OS,
- const NodeToUsesMap &M) LLVM_ATTRIBUTE_UNUSED;
+ [[maybe_unused]] raw_ostream &operator<<(raw_ostream &OS,
+ const NodeToUsesMap &M);
raw_ostream &operator<< (raw_ostream &OS, const NodeToUsesMap &M){
for (const auto &I : M) {
const UseSet &Us = I.second;
@@ -914,9 +913,8 @@ namespace {
const NodeToValueMap &Map;
};
- raw_ostream &operator<< (raw_ostream &OS,
- const LocationAsBlock &Loc) LLVM_ATTRIBUTE_UNUSED ;
- raw_ostream &operator<< (raw_ostream &OS, const LocationAsBlock &Loc) {
+ [[maybe_unused]] raw_ostream &operator<<(raw_ostream &OS,
+ const LocationAsBlock &Loc) {
for (const auto &I : Loc.Map) {
OS << I.first << " -> ";
if (BasicBlock *B = cast_or_null<BasicBlock>(I.second))
diff --git a/llvm/lib/Target/Hexagon/HexagonEarlyIfConv.cpp b/llvm/lib/Target/Hexagon/HexagonEarlyIfConv.cpp
index 14a7ae7..3900aac 100644
--- a/llvm/lib/Target/Hexagon/HexagonEarlyIfConv.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonEarlyIfConv.cpp
@@ -132,8 +132,7 @@ namespace {
const TargetRegisterInfo &TRI;
friend raw_ostream &operator<< (raw_ostream &OS, const PrintFP &P);
};
- raw_ostream &operator<<(raw_ostream &OS,
- const PrintFP &P) LLVM_ATTRIBUTE_UNUSED;
+ [[maybe_unused]] raw_ostream &operator<<(raw_ostream &OS, const PrintFP &P);
raw_ostream &operator<<(raw_ostream &OS, const PrintFP &P) {
OS << "{ SplitB:" << PrintMB(P.FP.SplitB)
<< ", PredR:" << printReg(P.FP.PredR, &P.TRI)
diff --git a/llvm/lib/Target/Hexagon/HexagonGenPredicate.cpp b/llvm/lib/Target/Hexagon/HexagonGenPredicate.cpp
index f9fdab4..9c81e96 100644
--- a/llvm/lib/Target/Hexagon/HexagonGenPredicate.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonGenPredicate.cpp
@@ -51,11 +51,11 @@ private:
const TargetRegisterInfo &TRI;
};
- raw_ostream &operator<< (raw_ostream &OS, const PrintRegister &PR)
- LLVM_ATTRIBUTE_UNUSED;
- raw_ostream &operator<< (raw_ostream &OS, const PrintRegister &PR) {
- return OS << printReg(PR.Reg.Reg, &PR.TRI, PR.Reg.SubReg);
- }
+[[maybe_unused]] raw_ostream &operator<<(raw_ostream &OS,
+ const PrintRegister &PR);
+raw_ostream &operator<<(raw_ostream &OS, const PrintRegister &PR) {
+ return OS << printReg(PR.Reg.Reg, &PR.TRI, PR.Reg.SubReg);
+}
class HexagonGenPredicate : public MachineFunctionPass {
public:
diff --git a/llvm/lib/Target/Hexagon/MCTargetDesc/HexagonMCTargetDesc.cpp b/llvm/lib/Target/Hexagon/MCTargetDesc/HexagonMCTargetDesc.cpp
index bfea50e..6b48a21 100644
--- a/llvm/lib/Target/Hexagon/MCTargetDesc/HexagonMCTargetDesc.cpp
+++ b/llvm/lib/Target/Hexagon/MCTargetDesc/HexagonMCTargetDesc.cpp
@@ -422,12 +422,12 @@ static MCTargetStreamer *createHexagonNullTargetStreamer(MCStreamer &S) {
return new HexagonTargetStreamer(S);
}
-static void LLVM_ATTRIBUTE_UNUSED clearFeature(MCSubtargetInfo* STI, uint64_t F) {
+[[maybe_unused]] static void clearFeature(MCSubtargetInfo *STI, uint64_t F) {
if (STI->hasFeature(F))
STI->ToggleFeature(F);
}
-static bool LLVM_ATTRIBUTE_UNUSED checkFeature(MCSubtargetInfo* STI, uint64_t F) {
+[[maybe_unused]] static bool checkFeature(MCSubtargetInfo *STI, uint64_t F) {
return STI->hasFeature(F);
}
diff --git a/llvm/lib/Target/Mips/MipsFastISel.cpp b/llvm/lib/Target/Mips/MipsFastISel.cpp
index 1ce8d7e3..df0c8c1 100644
--- a/llvm/lib/Target/Mips/MipsFastISel.cpp
+++ b/llvm/lib/Target/Mips/MipsFastISel.cpp
@@ -264,9 +264,10 @@ public:
} // end anonymous namespace
-static bool CC_Mips(unsigned ValNo, MVT ValVT, MVT LocVT,
- CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags,
- Type *OrigTy, CCState &State) LLVM_ATTRIBUTE_UNUSED;
+[[maybe_unused]] static bool CC_Mips(unsigned ValNo, MVT ValVT, MVT LocVT,
+ CCValAssign::LocInfo LocInfo,
+ ISD::ArgFlagsTy ArgFlags, Type *OrigTy,
+ CCState &State);
static bool CC_MipsO32_FP32(unsigned ValNo, MVT ValVT, MVT LocVT,
CCValAssign::LocInfo LocInfo,
diff --git a/llvm/lib/Target/Mips/MipsISelLowering.cpp b/llvm/lib/Target/Mips/MipsISelLowering.cpp
index 7f1ff45..2fd7327 100644
--- a/llvm/lib/Target/Mips/MipsISelLowering.cpp
+++ b/llvm/lib/Target/Mips/MipsISelLowering.cpp
@@ -3176,9 +3176,10 @@ static bool CC_MipsO32_FP64(unsigned ValNo, MVT ValVT, MVT LocVT,
F64Regs);
}
-static bool CC_MipsO32(unsigned ValNo, MVT ValVT, MVT LocVT,
- CCValAssign::LocInfo LocInfo, ISD::ArgFlagsTy ArgFlags,
- Type *OrigTy, CCState &State) LLVM_ATTRIBUTE_UNUSED;
+[[maybe_unused]] static bool CC_MipsO32(unsigned ValNo, MVT ValVT, MVT LocVT,
+ CCValAssign::LocInfo LocInfo,
+ ISD::ArgFlagsTy ArgFlags, Type *OrigTy,
+ CCState &State);
#include "MipsGenCallingConv.inc"
diff --git a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
index 8851a0f..e857b2d 100644
--- a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
+++ b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
@@ -3356,10 +3356,10 @@ bool RISCVAsmParser::parseDirectiveAttribute() {
bool isValidInsnFormat(StringRef Format, const MCSubtargetInfo &STI) {
return StringSwitch<bool>(Format)
- .Cases("r", "r4", "i", "b", "sb", "u", "j", "uj", "s", true)
- .Cases("cr", "ci", "ciw", "css", "cl", "cs", "ca", "cb", "cj",
+ .Cases({"r", "r4", "i", "b", "sb", "u", "j", "uj", "s"}, true)
+ .Cases({"cr", "ci", "ciw", "css", "cl", "cs", "ca", "cb", "cj"},
STI.hasFeature(RISCV::FeatureStdExtZca))
- .Cases("qc.eai", "qc.ei", "qc.eb", "qc.ej", "qc.es",
+ .Cases({"qc.eai", "qc.ei", "qc.eb", "qc.ej", "qc.es"},
!STI.hasFeature(RISCV::Feature64Bit))
.Default(false);
}
diff --git a/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp b/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
index 662d3f6..b1794b7 100644
--- a/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
+++ b/llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
@@ -717,6 +717,18 @@ RISCVLegalizerInfo::RISCVLegalizerInfo(const RISCVSubtarget &ST)
.clampScalar(0, sXLen, sXLen)
.lower();
+ LegalityPredicate InsertVectorEltPred = [=](const LegalityQuery &Query) {
+ LLT VecTy = Query.Types[0];
+ LLT EltTy = Query.Types[1];
+ return VecTy.getElementType() == EltTy;
+ };
+
+ getActionDefinitionsBuilder(G_INSERT_VECTOR_ELT)
+ .legalIf(all(typeIsLegalIntOrFPVec(0, IntOrFPVecTys, ST),
+ InsertVectorEltPred, typeIs(2, sXLen)))
+ .legalIf(all(typeIsLegalBoolVec(0, BoolVecTys, ST), InsertVectorEltPred,
+ typeIs(2, sXLen)));
+
getLegacyLegalizerInfo().computeTables();
verify(*ST.getInstrInfo());
}
diff --git a/llvm/lib/Target/RISCV/MCA/RISCVCustomBehaviour.cpp b/llvm/lib/Target/RISCV/MCA/RISCVCustomBehaviour.cpp
index 50730c6..ab93bba 100644
--- a/llvm/lib/Target/RISCV/MCA/RISCVCustomBehaviour.cpp
+++ b/llvm/lib/Target/RISCV/MCA/RISCVCustomBehaviour.cpp
@@ -43,7 +43,7 @@ const llvm::StringRef RISCVLMULInstrument::DESC_NAME = "RISCV-LMUL";
bool RISCVLMULInstrument::isDataValid(llvm::StringRef Data) {
// Return true if not one of the valid LMUL strings
return StringSwitch<bool>(Data)
- .Cases("M1", "M2", "M4", "M8", "MF2", "MF4", "MF8", true)
+ .Cases({"M1", "M2", "M4", "M8", "MF2", "MF4", "MF8"}, true)
.Default(false);
}
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 7123a2d..eb87558 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1672,6 +1672,8 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
if (Subtarget.useRVVForFixedLengthVectors())
setTargetDAGCombine(ISD::BITCAST);
+ setMaxDivRemBitWidthSupported(Subtarget.is64Bit() ? 128 : 64);
+
// Disable strict node mutation.
IsStrictFPEnabled = true;
EnableExtLdPromotion = true;
diff --git a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
index f863392a..637d61fe 100644
--- a/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
+++ b/llvm/lib/Target/RISCV/RISCVSchedSiFive7.td
@@ -270,7 +270,7 @@ class SiFive7AnyToGPRBypass<SchedRead read, int cycles = 2>
// and floating point computation.
// The V pipeline is modeled by the VCQ, VA, VL, and VS resources. There can
// be one or two VA (Vector Arithmetic).
-multiclass SiFive7ProcResources<bit extraVALU = false> {
+multiclass SiFive7ProcResources<bit dualVALU = false> {
let BufferSize = 0 in {
def PipeA : ProcResource<1>;
def PipeB : ProcResource<1>;
@@ -279,7 +279,7 @@ multiclass SiFive7ProcResources<bit extraVALU = false> {
def FDiv : ProcResource<1>; // FP Division/Sqrt
// Arithmetic sequencer(s)
- if extraVALU then {
+ if dualVALU then {
// VA1 can handle any vector airthmetic instruction.
def VA1 : ProcResource<1>;
// VA2 generally can only handle simple vector arithmetic.
@@ -305,7 +305,7 @@ multiclass SiFive7ProcResources<bit extraVALU = false> {
def PipeAB : ProcResGroup<[!cast<ProcResource>(NAME#"PipeA"),
!cast<ProcResource>(NAME#"PipeB")]>;
- if extraVALU then
+ if dualVALU then
def VA1OrVA2 : ProcResGroup<[!cast<ProcResource>(NAME#"VA1"),
!cast<ProcResource>(NAME#"VA2")]>;
}
@@ -1550,10 +1550,10 @@ multiclass SiFive7ReadAdvance {
/// This multiclass is a "bundle" of (1) processor resources (i.e. pipes) and
/// (2) WriteRes entries. It's parameterized by config values that will
/// eventually be supplied by different SchedMachineModels.
-multiclass SiFive7SchedResources<int vlen, bit extraVALU,
+multiclass SiFive7SchedResources<int vlen, bit dualVALU,
SiFive7FPLatencies fpLatencies,
bit hasFastGather> {
- defm SiFive7 : SiFive7ProcResources<extraVALU>;
+ defm SiFive7 : SiFive7ProcResources<dualVALU>;
// Pull out defs from SiFive7ProcResources so we can refer to them by name.
defvar SiFive7PipeA = !cast<ProcResource>(NAME # SiFive7PipeA);
@@ -1562,10 +1562,10 @@ multiclass SiFive7SchedResources<int vlen, bit extraVALU,
defvar SiFive7IDiv = !cast<ProcResource>(NAME # SiFive7IDiv);
defvar SiFive7FDiv = !cast<ProcResource>(NAME # SiFive7FDiv);
// Pass SiFive7VA for VA1 and VA1OrVA2 if there is only 1 VALU.
- defvar SiFive7VA1 = !if (extraVALU,
+ defvar SiFive7VA1 = !if (dualVALU,
!cast<ProcResource>(NAME # SiFive7VA1),
!cast<ProcResource>(NAME # SiFive7VA));
- defvar SiFive7VA1OrVA2 = !if (extraVALU,
+ defvar SiFive7VA1OrVA2 = !if (dualVALU,
!cast<ProcResGroup>(NAME # SiFive7VA1OrVA2),
!cast<ProcResource>(NAME # SiFive7VA));
defvar SiFive7VA = !cast<ProcResource>(NAME # SiFive7VA);
@@ -1608,7 +1608,7 @@ class SiFive7SchedMachineModel<int vlen> : SchedMachineModel {
HasStdExtZknh, HasStdExtZksed, HasStdExtZksh,
HasStdExtZkr];
int VLEN = vlen;
- bit HasExtraVALU = false;
+ bit HasDualVALU = false;
SiFive7FPLatencies FPLatencies;
bit HasFastGather = false;
@@ -1635,7 +1635,7 @@ def SiFive7VLEN512Model : SiFive7SchedMachineModel<512> {
}
def SiFive7VLEN1024X300Model : SiFive7SchedMachineModel<1024> {
- let HasExtraVALU = true;
+ let HasDualVALU = true;
let FPLatencies = SiFive7LowFPLatencies;
let HasFastGather = true;
}
@@ -1643,7 +1643,7 @@ def SiFive7VLEN1024X300Model : SiFive7SchedMachineModel<1024> {
/// Binding models to their scheduling resources.
foreach model = [SiFive7VLEN512Model, SiFive7VLEN1024X300Model] in {
let SchedModel = model in
- defm model.Name : SiFive7SchedResources<model.VLEN, model.HasExtraVALU,
+ defm model.Name : SiFive7SchedResources<model.VLEN, model.HasDualVALU,
model.FPLatencies,
model.HasFastGather>;
}
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index a466ab2..a0cff4d 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -3765,7 +3765,6 @@ void SPIRVInstructionSelector::decorateUsesAsNonUniform(
SPIRV::Decoration::NonUniformEXT, {});
}
}
- return;
}
bool SPIRVInstructionSelector::extractSubvector(
diff --git a/llvm/lib/Target/Sparc/SparcFrameLowering.cpp b/llvm/lib/Target/Sparc/SparcFrameLowering.cpp
index 2934c88..fa08d44 100644
--- a/llvm/lib/Target/Sparc/SparcFrameLowering.cpp
+++ b/llvm/lib/Target/Sparc/SparcFrameLowering.cpp
@@ -246,8 +246,7 @@ SparcFrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI,
}
}
-static bool LLVM_ATTRIBUTE_UNUSED verifyLeafProcRegUse(MachineRegisterInfo *MRI)
-{
+[[maybe_unused]] static bool verifyLeafProcRegUse(MachineRegisterInfo *MRI) {
for (unsigned reg = SP::I0; reg <= SP::I7; ++reg)
if (MRI->isPhysRegUsed(reg))
diff --git a/llvm/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTypeUtilities.cpp b/llvm/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTypeUtilities.cpp
index d9c8e22..6e99fc3 100644
--- a/llvm/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTypeUtilities.cpp
+++ b/llvm/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTypeUtilities.cpp
@@ -23,7 +23,7 @@ std::optional<wasm::ValType> WebAssembly::parseType(StringRef Type) {
.Case("i64", wasm::ValType::I64)
.Case("f32", wasm::ValType::F32)
.Case("f64", wasm::ValType::F64)
- .Cases("v128", "i8x16", "i16x8", "i32x4", "i64x2", "f32x4", "f64x2",
+ .Cases({"v128", "i8x16", "i16x8", "i32x4", "i64x2", "f32x4", "f64x2"},
wasm::ValType::V128)
.Case("funcref", wasm::ValType::FUNCREF)
.Case("externref", wasm::ValType::EXTERNREF)
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td b/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td
index 0f6e1ca..ed54404d 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrSIMD.td
@@ -1763,6 +1763,26 @@ defm RELAXED_DOT :
"i16x8.relaxed_dot_i8x16_i7x16_s\t$dst, $lhs, $rhs",
"i16x8.relaxed_dot_i8x16_i7x16_s", 0x112>;
+def : Pat<
+ (v8i16 (add
+ (wasm_shuffle
+ (v8i16 (extmul_low_s v16i8:$lhs, v16i8:$rhs)),
+ (v8i16 (extmul_high_s v16i8:$lhs, v16i8:$rhs)),
+ (i32 0), (i32 1), (i32 4), (i32 5),
+ (i32 8), (i32 9), (i32 12), (i32 13),
+ (i32 16), (i32 17), (i32 20), (i32 21),
+ (i32 24), (i32 25), (i32 28), (i32 29)),
+ (wasm_shuffle
+ (v8i16 (extmul_low_s v16i8:$lhs, v16i8:$rhs)),
+ (v8i16 (extmul_high_s v16i8:$lhs, v16i8:$rhs)),
+ (i32 2), (i32 3), (i32 6), (i32 7),
+ (i32 10), (i32 11), (i32 14), (i32 15),
+ (i32 18), (i32 19), (i32 22), (i32 23),
+ (i32 26), (i32 27), (i32 30), (i32 31)))
+ ),
+ (v8i16 (RELAXED_DOT v16i8:$lhs, v16i8:$rhs))
+>;
+
defm RELAXED_DOT_ADD :
RELAXED_I<(outs V128:$dst), (ins V128:$lhs, V128:$rhs, V128:$acc),
(outs), (ins),
@@ -1771,6 +1791,18 @@ defm RELAXED_DOT_ADD :
"i32x4.relaxed_dot_i8x16_i7x16_add_s\t$dst, $lhs, $rhs, $acc",
"i32x4.relaxed_dot_i8x16_i7x16_add_s", 0x113>;
+def : Pat<
+ (v4i32 (add
+ (v4i32 (int_wasm_extadd_pairwise_signed
+ (v8i16 (int_wasm_relaxed_dot_i8x16_i7x16_signed v16i8:$lhs, v16i8:$rhs)))),
+ (v4i32 V128:$acc))),
+ (v4i32 (RELAXED_DOT_ADD v16i8:$lhs, v16i8:$rhs, (v4i32 V128:$acc)))
+ >;
+
+def : Pat<(v4i32 (partial_reduce_smla (v4i32 V128:$acc), (v16i8 V128:$lhs),
+ (v16i8 V128:$rhs))),
+ (RELAXED_DOT_ADD $lhs, $rhs, $acc)>, Requires<[HasRelaxedSIMD]>;
+
//===----------------------------------------------------------------------===//
// Relaxed BFloat16 dot product
//===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
index a8908d4..ac251fd 100644
--- a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
+++ b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
@@ -3514,15 +3514,16 @@ bool X86AsmParser::parseInstruction(ParseInstructionInfo &Info, StringRef Name,
// xacquire <insn> ; xacquire must be accompanied by 'lock'
bool IsPrefix =
StringSwitch<bool>(Name)
- .Cases("cs", "ds", "es", "fs", "gs", "ss", true)
- .Cases("rex64", "data32", "data16", "addr32", "addr16", true)
- .Cases("xacquire", "xrelease", true)
- .Cases("acquire", "release", isParsingIntelSyntax())
+ .Cases({"cs", "ds", "es", "fs", "gs", "ss"}, true)
+ .Cases({"rex64", "data32", "data16", "addr32", "addr16"}, true)
+ .Cases({"xacquire", "xrelease"}, true)
+ .Cases({"acquire", "release"}, isParsingIntelSyntax())
.Default(false);
auto isLockRepeatNtPrefix = [](StringRef N) {
return StringSwitch<bool>(N)
- .Cases("lock", "rep", "repe", "repz", "repne", "repnz", "notrack", true)
+ .Cases({"lock", "rep", "repe", "repz", "repne", "repnz", "notrack"},
+ true)
.Default(false);
};
diff --git a/llvm/lib/Target/X86/GISel/X86LegalizerInfo.cpp b/llvm/lib/Target/X86/GISel/X86LegalizerInfo.cpp
index b81641f..28fa2cd 100644
--- a/llvm/lib/Target/X86/GISel/X86LegalizerInfo.cpp
+++ b/llvm/lib/Target/X86/GISel/X86LegalizerInfo.cpp
@@ -414,8 +414,6 @@ X86LegalizerInfo::X86LegalizerInfo(const X86Subtarget &STI,
getActionDefinitionsBuilder(G_SEXT_INREG).lower();
- getActionDefinitionsBuilder(G_IS_FPCLASS).lower();
-
// fp constants
getActionDefinitionsBuilder(G_FCONSTANT)
.legalFor({s32, s64})
diff --git a/llvm/lib/Target/X86/X86.td b/llvm/lib/Target/X86/X86.td
index 6db780f..8e08d16 100644
--- a/llvm/lib/Target/X86/X86.td
+++ b/llvm/lib/Target/X86/X86.td
@@ -1338,6 +1338,10 @@ def ProcessorFeatures {
list<SubtargetFeature> PTLFeatures =
!listremove(ARLSFeatures, [FeatureWIDEKL]);
+ // Novalake
+ list<SubtargetFeature> NVLFeatures =
+ !listconcat(PTLFeatures, [FeaturePREFETCHI]);
+
// Clearwaterforest
list<SubtargetFeature> CWFAdditionalFeatures = [FeaturePREFETCHI,
FeatureAVXVNNIINT16,
@@ -1883,6 +1887,9 @@ foreach P = ["pantherlake", "wildcatlake"] in {
def : ProcModel<P, AlderlakePModel,
ProcessorFeatures.PTLFeatures, ProcessorFeatures.ADLTuning>;
}
+def : ProcModel<"novalake", AlderlakePModel, ProcessorFeatures.NVLFeatures,
+ ProcessorFeatures.ADLTuning>;
+
def : ProcModel<"clearwaterforest", AlderlakePModel,
ProcessorFeatures.CWFFeatures, ProcessorFeatures.ADLTuning>;
def : ProcModel<"emeraldrapids", SapphireRapidsModel,
diff --git a/llvm/lib/Target/X86/X86FloatingPoint.cpp b/llvm/lib/Target/X86/X86FloatingPoint.cpp
index e0991aa..9f88fda 100644
--- a/llvm/lib/Target/X86/X86FloatingPoint.cpp
+++ b/llvm/lib/Target/X86/X86FloatingPoint.cpp
@@ -602,8 +602,7 @@ namespace {
friend bool operator<(const TableEntry &TE, unsigned V) {
return TE.from < V;
}
- friend bool LLVM_ATTRIBUTE_UNUSED operator<(unsigned V,
- const TableEntry &TE) {
+ [[maybe_unused]] friend bool operator<(unsigned V, const TableEntry &TE) {
return V < TE.from;
}
};
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index c32b1a6..a0b64ff 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -58342,11 +58342,12 @@ static SDValue combineX86CloadCstore(SDNode *N, SelectionDAG &DAG) {
} else if (Op1.getOpcode() == ISD::AND && Sub.getValue(0).use_empty()) {
SDValue Src = Op1;
SDValue Op10 = Op1.getOperand(0);
- if (Op10.getOpcode() == ISD::XOR && isAllOnesConstant(Op10.getOperand(1))) {
- // res, flags2 = sub 0, (and (xor X, -1), Y)
+ if (Op10.getOpcode() == ISD::XOR && isAllOnesConstant(Op10.getOperand(1)) &&
+ llvm::isOneConstant(Op1.getOperand(1))) {
+ // res, flags2 = sub 0, (and (xor X, -1), 1)
// cload/cstore ..., cond_ne, flag2
// ->
- // res, flags2 = sub 0, (and X, Y)
+ // res, flags2 = sub 0, (and X, 1)
// cload/cstore ..., cond_e, flag2
Src = DAG.getNode(ISD::AND, DL, Op1.getValueType(), Op10.getOperand(0),
Op1.getOperand(1));
diff --git a/llvm/lib/Target/X86/X86InstrCompiler.td b/llvm/lib/Target/X86/X86InstrCompiler.td
index 0fd44b7..ec31675 100644
--- a/llvm/lib/Target/X86/X86InstrCompiler.td
+++ b/llvm/lib/Target/X86/X86InstrCompiler.td
@@ -1256,8 +1256,17 @@ def : Pat<(i64 (X86Wrapper tconstpool :$dst)),
(MOV64ri32 tconstpool :$dst)>, Requires<[KernelCode]>;
def : Pat<(i64 (X86Wrapper tjumptable :$dst)),
(MOV64ri32 tjumptable :$dst)>, Requires<[KernelCode]>;
-def : Pat<(i64 (X86Wrapper tglobaladdr :$dst)),
- (MOV64ri32 tglobaladdr :$dst)>, Requires<[KernelCode]>;
+
+// If the globaladdr is an absolute_symbol, don't bother using the sign extending
+// instruction since there's no benefit to using it with absolute symbols.
+def globalAddrNoAbsSym : PatLeaf<(tglobaladdr:$dst), [{
+ auto *GA = cast<GlobalAddressSDNode>(N);
+ return !GA->getGlobal()->getAbsoluteSymbolRange();
+}]>;
+def : Pat<(i64 (X86Wrapper globalAddrNoAbsSym:$dst)),
+ (MOV64ri32 tglobaladdr:$dst)>,
+ Requires<[KernelCode]>;
+
def : Pat<(i64 (X86Wrapper texternalsym:$dst)),
(MOV64ri32 texternalsym:$dst)>, Requires<[KernelCode]>;
def : Pat<(i64 (X86Wrapper mcsym:$dst)),
diff --git a/llvm/lib/Target/X86/X86MCInstLower.cpp b/llvm/lib/Target/X86/X86MCInstLower.cpp
index 481a9be..1fca466f 100644
--- a/llvm/lib/Target/X86/X86MCInstLower.cpp
+++ b/llvm/lib/Target/X86/X86MCInstLower.cpp
@@ -1946,16 +1946,14 @@ static void addConstantComments(const MachineInstr *MI,
CASE_ARITH_RM(PMADDUBSW) {
unsigned SrcIdx = getSrcIdx(MI, 1);
if (auto *C = X86::getConstantFromPool(*MI, SrcIdx + 1)) {
- if (C->getType()->getScalarSizeInBits() == 8) {
- std::string Comment;
- raw_string_ostream CS(Comment);
- unsigned VectorWidth =
- X86::getVectorRegisterWidth(MI->getDesc().operands()[0]);
- CS << "[";
- printConstant(C, VectorWidth, CS);
- CS << "]";
- OutStreamer.AddComment(CS.str());
- }
+ std::string Comment;
+ raw_string_ostream CS(Comment);
+ unsigned VectorWidth =
+ X86::getVectorRegisterWidth(MI->getDesc().operands()[0]);
+ CS << "[";
+ printConstant(C, VectorWidth, CS);
+ CS << "]";
+ OutStreamer.AddComment(CS.str());
}
break;
}
@@ -1967,16 +1965,14 @@ static void addConstantComments(const MachineInstr *MI,
CASE_ARITH_RM(PMULHRSW) {
unsigned SrcIdx = getSrcIdx(MI, 1);
if (auto *C = X86::getConstantFromPool(*MI, SrcIdx + 1)) {
- if (C->getType()->getScalarSizeInBits() == 16) {
- std::string Comment;
- raw_string_ostream CS(Comment);
- unsigned VectorWidth =
- X86::getVectorRegisterWidth(MI->getDesc().operands()[0]);
- CS << "[";
- printConstant(C, VectorWidth, CS);
- CS << "]";
- OutStreamer.AddComment(CS.str());
- }
+ std::string Comment;
+ raw_string_ostream CS(Comment);
+ unsigned VectorWidth =
+ X86::getVectorRegisterWidth(MI->getDesc().operands()[0]);
+ CS << "[";
+ printConstant(C, VectorWidth, CS);
+ CS << "]";
+ OutStreamer.AddComment(CS.str());
}
break;
}