llvm-project/llvm/lib/Target/SPIRV/SPIRVISelLowering.cpp
Vyacheslav Levytskyy f768083516
[SPIR-V] Update type inference and instruction selection (#88254)
This PR contains a series of fixes which are to improve type inference
and instruction selection.

Namely, it includes:
* fix OpSelect to support operands of a pointer type, according to the
SPIR-V specification (previously only integer/float/vectors of integer
or float were supported) -- a new test case is added and existing test
case is updated;
* fix TableGen typo's in definition of register classes and introduce a
new reg class that is a vector of pointers;
* fix usage of a machine function context when there is a need to switch
between different machine functions to infer/validate correct types;
* add usage of TypedPointerType instead of PointerType so that later
stages of type inference are able to distinguish pointer types by their
element types, effectively supporting hierarchy of pointer/pointee types
and avoiding more complicated recursive type matching on level of
machine instructions in favor of direct pointer comparison using LLVM's
`Type *` values;
* extracting detailed information about operand types using known type
rules for some llvm instructions (for instance, by deducing PHI's
operand pointee types if PHI's results type was deducted on previous
stages of type inference), and adding correspondent
`Intrinsic::spv_assign_ptr_type` to keep type info along consequent
passes,
* ensure that OpConstantComposite reuses a constant when it's already
created and available in the same machine function -- otherwise there is
a crash while building a dependency graph, the corresponding test case
is attached,
* implement deduction of function's return type for opaque pointers, a
new test case is attached,
* make 'emit intrinsics' a module pass to resolve function return types
over the module -- first types for all functions of the module must be
calculated, and only after that it's feasible to deduct function return
types on this earlier stage of translation.
2024-04-15 09:59:47 +02:00

323 lines
14 KiB
C++

//===- SPIRVISelLowering.cpp - SPIR-V DAG Lowering Impl ---------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file implements the SPIRVTargetLowering class.
//
//===----------------------------------------------------------------------===//
#include "SPIRVISelLowering.h"
#include "SPIRV.h"
#include "SPIRVInstrInfo.h"
#include "SPIRVRegisterBankInfo.h"
#include "SPIRVRegisterInfo.h"
#include "SPIRVSubtarget.h"
#include "SPIRVTargetMachine.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/IR/IntrinsicsSPIRV.h"
#define DEBUG_TYPE "spirv-lower"
using namespace llvm;
unsigned SPIRVTargetLowering::getNumRegistersForCallingConv(
LLVMContext &Context, CallingConv::ID CC, EVT VT) const {
// This code avoids CallLowering fail inside getVectorTypeBreakdown
// on v3i1 arguments. Maybe we need to return 1 for all types.
// TODO: remove it once this case is supported by the default implementation.
if (VT.isVector() && VT.getVectorNumElements() == 3 &&
(VT.getVectorElementType() == MVT::i1 ||
VT.getVectorElementType() == MVT::i8))
return 1;
if (!VT.isVector() && VT.isInteger() && VT.getSizeInBits() <= 64)
return 1;
return getNumRegisters(Context, VT);
}
MVT SPIRVTargetLowering::getRegisterTypeForCallingConv(LLVMContext &Context,
CallingConv::ID CC,
EVT VT) const {
// This code avoids CallLowering fail inside getVectorTypeBreakdown
// on v3i1 arguments. Maybe we need to return i32 for all types.
// TODO: remove it once this case is supported by the default implementation.
if (VT.isVector() && VT.getVectorNumElements() == 3) {
if (VT.getVectorElementType() == MVT::i1)
return MVT::v4i1;
else if (VT.getVectorElementType() == MVT::i8)
return MVT::v4i8;
}
return getRegisterType(Context, VT);
}
bool SPIRVTargetLowering::getTgtMemIntrinsic(IntrinsicInfo &Info,
const CallInst &I,
MachineFunction &MF,
unsigned Intrinsic) const {
unsigned AlignIdx = 3;
switch (Intrinsic) {
case Intrinsic::spv_load:
AlignIdx = 2;
[[fallthrough]];
case Intrinsic::spv_store: {
if (I.getNumOperands() >= AlignIdx + 1) {
auto *AlignOp = cast<ConstantInt>(I.getOperand(AlignIdx));
Info.align = Align(AlignOp->getZExtValue());
}
Info.flags = static_cast<MachineMemOperand::Flags>(
cast<ConstantInt>(I.getOperand(AlignIdx - 1))->getZExtValue());
Info.memVT = MVT::i64;
// TODO: take into account opaque pointers (don't use getElementType).
// MVT::getVT(PtrTy->getElementType());
return true;
break;
}
default:
break;
}
return false;
}
// Insert a bitcast before the instruction to keep SPIR-V code valid
// when there is a type mismatch between results and operand types.
static void validatePtrTypes(const SPIRVSubtarget &STI,
MachineRegisterInfo *MRI, SPIRVGlobalRegistry &GR,
MachineInstr &I, unsigned OpIdx,
SPIRVType *ResType, const Type *ResTy = nullptr) {
// Get operand type
MachineFunction *MF = I.getParent()->getParent();
Register OpReg = I.getOperand(OpIdx).getReg();
SPIRVType *TypeInst = MRI->getVRegDef(OpReg);
Register OpTypeReg =
TypeInst && TypeInst->getOpcode() == SPIRV::OpFunctionParameter
? TypeInst->getOperand(1).getReg()
: OpReg;
SPIRVType *OpType = GR.getSPIRVTypeForVReg(OpTypeReg, MF);
if (!ResType || !OpType || OpType->getOpcode() != SPIRV::OpTypePointer)
return;
// Get operand's pointee type
Register ElemTypeReg = OpType->getOperand(2).getReg();
SPIRVType *ElemType = GR.getSPIRVTypeForVReg(ElemTypeReg, MF);
if (!ElemType)
return;
// Check if we need a bitcast to make a statement valid
bool IsSameMF = MF == ResType->getParent()->getParent();
bool IsEqualTypes = IsSameMF ? ElemType == ResType
: GR.getTypeForSPIRVType(ElemType) == ResTy;
if (IsEqualTypes)
return;
// There is a type mismatch between results and operand types
// and we insert a bitcast before the instruction to keep SPIR-V code valid
SPIRV::StorageClass::StorageClass SC =
static_cast<SPIRV::StorageClass::StorageClass>(
OpType->getOperand(1).getImm());
MachineIRBuilder MIB(I);
SPIRVType *NewBaseType =
IsSameMF ? ResType
: GR.getOrCreateSPIRVType(
ResTy, MIB, SPIRV::AccessQualifier::ReadWrite, false);
SPIRVType *NewPtrType = GR.getOrCreateSPIRVPointerType(NewBaseType, MIB, SC);
if (!GR.isBitcastCompatible(NewPtrType, OpType))
report_fatal_error(
"insert validation bitcast: incompatible result and operand types");
Register NewReg = MRI->createGenericVirtualRegister(LLT::scalar(32));
bool Res = MIB.buildInstr(SPIRV::OpBitcast)
.addDef(NewReg)
.addUse(GR.getSPIRVTypeID(NewPtrType))
.addUse(OpReg)
.constrainAllUses(*STI.getInstrInfo(), *STI.getRegisterInfo(),
*STI.getRegBankInfo());
if (!Res)
report_fatal_error("insert validation bitcast: cannot constrain all uses");
MRI->setRegClass(NewReg, &SPIRV::IDRegClass);
GR.assignSPIRVTypeToVReg(NewPtrType, NewReg, MIB.getMF());
I.getOperand(OpIdx).setReg(NewReg);
}
// Insert a bitcast before the function call instruction to keep SPIR-V code
// valid when there is a type mismatch between actual and expected types of an
// argument:
// %formal = OpFunctionParameter %formal_type
// ...
// %res = OpFunctionCall %ty %fun %actual ...
// implies that %actual is of %formal_type, and in case of opaque pointers.
// We may need to insert a bitcast to ensure this.
void validateFunCallMachineDef(const SPIRVSubtarget &STI,
MachineRegisterInfo *DefMRI,
MachineRegisterInfo *CallMRI,
SPIRVGlobalRegistry &GR, MachineInstr &FunCall,
MachineInstr *FunDef) {
if (FunDef->getOpcode() != SPIRV::OpFunction)
return;
unsigned OpIdx = 3;
for (FunDef = FunDef->getNextNode();
FunDef && FunDef->getOpcode() == SPIRV::OpFunctionParameter &&
OpIdx < FunCall.getNumOperands();
FunDef = FunDef->getNextNode(), OpIdx++) {
SPIRVType *DefPtrType = DefMRI->getVRegDef(FunDef->getOperand(1).getReg());
SPIRVType *DefElemType =
DefPtrType && DefPtrType->getOpcode() == SPIRV::OpTypePointer
? GR.getSPIRVTypeForVReg(DefPtrType->getOperand(2).getReg(),
DefPtrType->getParent()->getParent())
: nullptr;
if (DefElemType) {
const Type *DefElemTy = GR.getTypeForSPIRVType(DefElemType);
// validatePtrTypes() works in the context if the call site
// When we process historical records about forward calls
// we need to switch context to the (forward) call site and
// then restore it back to the current machine function.
MachineFunction *CurMF =
GR.setCurrentFunc(*FunCall.getParent()->getParent());
validatePtrTypes(STI, CallMRI, GR, FunCall, OpIdx, DefElemType,
DefElemTy);
GR.setCurrentFunc(*CurMF);
}
}
}
// Ensure there is no mismatch between actual and expected arg types: calls
// with a processed definition. Return Function pointer if it's a forward
// call (ahead of definition), and nullptr otherwise.
const Function *validateFunCall(const SPIRVSubtarget &STI,
MachineRegisterInfo *CallMRI,
SPIRVGlobalRegistry &GR,
MachineInstr &FunCall) {
const GlobalValue *GV = FunCall.getOperand(2).getGlobal();
const Function *F = dyn_cast<Function>(GV);
MachineInstr *FunDef =
const_cast<MachineInstr *>(GR.getFunctionDefinition(F));
if (!FunDef)
return F;
MachineRegisterInfo *DefMRI = &FunDef->getParent()->getParent()->getRegInfo();
validateFunCallMachineDef(STI, DefMRI, CallMRI, GR, FunCall, FunDef);
return nullptr;
}
// Ensure there is no mismatch between actual and expected arg types: calls
// ahead of a processed definition.
void validateForwardCalls(const SPIRVSubtarget &STI,
MachineRegisterInfo *DefMRI, SPIRVGlobalRegistry &GR,
MachineInstr &FunDef) {
const Function *F = GR.getFunctionByDefinition(&FunDef);
if (SmallPtrSet<MachineInstr *, 8> *FwdCalls = GR.getForwardCalls(F))
for (MachineInstr *FunCall : *FwdCalls) {
MachineRegisterInfo *CallMRI =
&FunCall->getParent()->getParent()->getRegInfo();
validateFunCallMachineDef(STI, DefMRI, CallMRI, GR, *FunCall, &FunDef);
}
}
// Validation of an access chain.
void validateAccessChain(const SPIRVSubtarget &STI, MachineRegisterInfo *MRI,
SPIRVGlobalRegistry &GR, MachineInstr &I) {
SPIRVType *BaseTypeInst = GR.getSPIRVTypeForVReg(I.getOperand(0).getReg());
if (BaseTypeInst && BaseTypeInst->getOpcode() == SPIRV::OpTypePointer) {
SPIRVType *BaseElemType =
GR.getSPIRVTypeForVReg(BaseTypeInst->getOperand(2).getReg());
validatePtrTypes(STI, MRI, GR, I, 2, BaseElemType);
}
}
// TODO: the logic of inserting additional bitcast's is to be moved
// to pre-IRTranslation passes eventually
void SPIRVTargetLowering::finalizeLowering(MachineFunction &MF) const {
// finalizeLowering() is called twice (see GlobalISel/InstructionSelect.cpp)
// We'd like to avoid the needless second processing pass.
if (ProcessedMF.find(&MF) != ProcessedMF.end())
return;
MachineRegisterInfo *MRI = &MF.getRegInfo();
SPIRVGlobalRegistry &GR = *STI.getSPIRVGlobalRegistry();
GR.setCurrentFunc(MF);
for (MachineFunction::iterator I = MF.begin(), E = MF.end(); I != E; ++I) {
MachineBasicBlock *MBB = &*I;
for (MachineBasicBlock::iterator MBBI = MBB->begin(), MBBE = MBB->end();
MBBI != MBBE;) {
MachineInstr &MI = *MBBI++;
switch (MI.getOpcode()) {
case SPIRV::OpAtomicLoad:
case SPIRV::OpAtomicExchange:
case SPIRV::OpAtomicCompareExchange:
case SPIRV::OpAtomicCompareExchangeWeak:
case SPIRV::OpAtomicIIncrement:
case SPIRV::OpAtomicIDecrement:
case SPIRV::OpAtomicIAdd:
case SPIRV::OpAtomicISub:
case SPIRV::OpAtomicSMin:
case SPIRV::OpAtomicUMin:
case SPIRV::OpAtomicSMax:
case SPIRV::OpAtomicUMax:
case SPIRV::OpAtomicAnd:
case SPIRV::OpAtomicOr:
case SPIRV::OpAtomicXor:
// for the above listed instructions
// OpAtomicXXX <ResType>, ptr %Op, ...
// implies that %Op is a pointer to <ResType>
case SPIRV::OpLoad:
// OpLoad <ResType>, ptr %Op implies that %Op is a pointer to <ResType>
validatePtrTypes(STI, MRI, GR, MI, 2,
GR.getSPIRVTypeForVReg(MI.getOperand(0).getReg()));
break;
case SPIRV::OpAtomicStore:
// OpAtomicStore ptr %Op, <Scope>, <Mem>, <Obj>
// implies that %Op points to the <Obj>'s type
validatePtrTypes(STI, MRI, GR, MI, 0,
GR.getSPIRVTypeForVReg(MI.getOperand(3).getReg()));
break;
case SPIRV::OpStore:
// OpStore ptr %Op, <Obj> implies that %Op points to the <Obj>'s type
validatePtrTypes(STI, MRI, GR, MI, 0,
GR.getSPIRVTypeForVReg(MI.getOperand(1).getReg()));
break;
case SPIRV::OpPtrCastToGeneric:
validateAccessChain(STI, MRI, GR, MI);
break;
case SPIRV::OpInBoundsPtrAccessChain:
if (MI.getNumOperands() == 4)
validateAccessChain(STI, MRI, GR, MI);
break;
case SPIRV::OpFunctionCall:
// ensure there is no mismatch between actual and expected arg types:
// calls with a processed definition
if (MI.getNumOperands() > 3)
if (const Function *F = validateFunCall(STI, MRI, GR, MI))
GR.addForwardCall(F, &MI);
break;
case SPIRV::OpFunction:
// ensure there is no mismatch between actual and expected arg types:
// calls ahead of a processed definition
validateForwardCalls(STI, MRI, GR, MI);
break;
// ensure that LLVM IR bitwise instructions result in logical SPIR-V
// instructions when applied to bool type
case SPIRV::OpBitwiseOrS:
case SPIRV::OpBitwiseOrV:
if (GR.isScalarOrVectorOfType(MI.getOperand(1).getReg(),
SPIRV::OpTypeBool))
MI.setDesc(STI.getInstrInfo()->get(SPIRV::OpLogicalOr));
break;
case SPIRV::OpBitwiseAndS:
case SPIRV::OpBitwiseAndV:
if (GR.isScalarOrVectorOfType(MI.getOperand(1).getReg(),
SPIRV::OpTypeBool))
MI.setDesc(STI.getInstrInfo()->get(SPIRV::OpLogicalAnd));
break;
case SPIRV::OpBitwiseXorS:
case SPIRV::OpBitwiseXorV:
if (GR.isScalarOrVectorOfType(MI.getOperand(1).getReg(),
SPIRV::OpTypeBool))
MI.setDesc(STI.getInstrInfo()->get(SPIRV::OpLogicalNotEqual));
break;
}
}
}
ProcessedMF.insert(&MF);
TargetLowering::finalizeLowering(MF);
}