llvm-project/llvm/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp

597 lines
22 KiB
C++
Raw Normal View History

//===-- FunctionLoweringInfo.cpp ------------------------------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This implements routines for translating functions from LLVM IR into
// Machine IR.
//
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/Analysis.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/TargetFrameLowering.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetLowering.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/CodeGen/WasmEHFuncInfo.h"
#include "llvm/CodeGen/WinEHFuncInfo.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetOptions.h"
#include <algorithm>
using namespace llvm;
#define DEBUG_TYPE "function-lowering-info"
/// isUsedOutsideOfDefiningBlock - Return true if this instruction is used by
/// PHI nodes or outside of the basic block that defines it, or used by a
/// switch or atomic instruction, which may expand to multiple basic blocks.
static bool isUsedOutsideOfDefiningBlock(const Instruction *I) {
if (I->use_empty()) return false;
if (isa<PHINode>(I)) return true;
const BasicBlock *BB = I->getParent();
[C++11] Add range based accessors for the Use-Def chain of a Value. This requires a number of steps. 1) Move value_use_iterator into the Value class as an implementation detail 2) Change it to actually be a *Use* iterator rather than a *User* iterator. 3) Add an adaptor which is a User iterator that always looks through the Use to the User. 4) Wrap these in Value::use_iterator and Value::user_iterator typedefs. 5) Add the range adaptors as Value::uses() and Value::users(). 6) Update *all* of the callers to correctly distinguish between whether they wanted a use_iterator (and to explicitly dig out the User when needed), or a user_iterator which makes the Use itself totally opaque. Because #6 requires churning essentially everything that walked the Use-Def chains, I went ahead and added all of the range adaptors and switched them to range-based loops where appropriate. Also because the renaming requires at least churning every line of code, it didn't make any sense to split these up into multiple commits -- all of which would touch all of the same lies of code. The result is still not quite optimal. The Value::use_iterator is a nice regular iterator, but Value::user_iterator is an iterator over User*s rather than over the User objects themselves. As a consequence, it fits a bit awkwardly into the range-based world and it has the weird extra-dereferencing 'operator->' that so many of our iterators have. I think this could be fixed by providing something which transforms a range of T&s into a range of T*s, but that *can* be separated into another patch, and it isn't yet 100% clear whether this is the right move. However, this change gets us most of the benefit and cleans up a substantial amount of code around Use and User. =] llvm-svn: 203364
2014-03-09 11:16:01 +08:00
for (const User *U : I->users())
if (cast<Instruction>(U)->getParent() != BB || isa<PHINode>(U))
return true;
[C++11] Add range based accessors for the Use-Def chain of a Value. This requires a number of steps. 1) Move value_use_iterator into the Value class as an implementation detail 2) Change it to actually be a *Use* iterator rather than a *User* iterator. 3) Add an adaptor which is a User iterator that always looks through the Use to the User. 4) Wrap these in Value::use_iterator and Value::user_iterator typedefs. 5) Add the range adaptors as Value::uses() and Value::users(). 6) Update *all* of the callers to correctly distinguish between whether they wanted a use_iterator (and to explicitly dig out the User when needed), or a user_iterator which makes the Use itself totally opaque. Because #6 requires churning essentially everything that walked the Use-Def chains, I went ahead and added all of the range adaptors and switched them to range-based loops where appropriate. Also because the renaming requires at least churning every line of code, it didn't make any sense to split these up into multiple commits -- all of which would touch all of the same lies of code. The result is still not quite optimal. The Value::use_iterator is a nice regular iterator, but Value::user_iterator is an iterator over User*s rather than over the User objects themselves. As a consequence, it fits a bit awkwardly into the range-based world and it has the weird extra-dereferencing 'operator->' that so many of our iterators have. I think this could be fixed by providing something which transforms a range of T&s into a range of T*s, but that *can* be separated into another patch, and it isn't yet 100% clear whether this is the right move. However, this change gets us most of the benefit and cleans up a substantial amount of code around Use and User. =] llvm-svn: 203364
2014-03-09 11:16:01 +08:00
return false;
}
static ISD::NodeType getPreferredExtendForValue(const Value *V) {
// For the users of the source value being used for compare instruction, if
// the number of signed predicate is greater than unsigned predicate, we
// prefer to use SIGN_EXTEND.
//
// With this optimization, we would be able to reduce some redundant sign or
// zero extension instruction, and eventually more machine CSE opportunities
// can be exposed.
ISD::NodeType ExtendKind = ISD::ANY_EXTEND;
unsigned NumOfSigned = 0, NumOfUnsigned = 0;
for (const User *U : V->users()) {
if (const auto *CI = dyn_cast<CmpInst>(U)) {
NumOfSigned += CI->isSigned();
NumOfUnsigned += CI->isUnsigned();
}
}
if (NumOfSigned > NumOfUnsigned)
ExtendKind = ISD::SIGN_EXTEND;
return ExtendKind;
}
void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf,
SelectionDAG *DAG) {
Fn = &fn;
MF = &mf;
TLI = MF->getSubtarget().getTargetLowering();
RegInfo = &MF->getRegInfo();
const TargetFrameLowering *TFI = MF->getSubtarget().getFrameLowering();
unsigned StackAlign = TFI->getStackAlignment();
// Check whether the function can return without sret-demotion.
SmallVector<ISD::OutputArg, 4> Outs;
CallingConv::ID CC = Fn->getCallingConv();
GetReturnInfo(CC, Fn->getReturnType(), Fn->getAttributes(), Outs, *TLI,
mf.getDataLayout());
CanLowerReturn =
TLI->CanLowerReturn(CC, *MF, Fn->isVarArg(), Outs, Fn->getContext());
// If this personality uses funclets, we need to do a bit more work.
DenseMap<const AllocaInst *, TinyPtrVector<int *>> CatchObjects;
EHPersonality Personality = classifyEHPersonality(
Fn->hasPersonalityFn() ? Fn->getPersonalityFn() : nullptr);
if (isFuncletEHPersonality(Personality)) {
// Calculate state numbers if we haven't already.
WinEHFuncInfo &EHInfo = *MF->getWinEHFuncInfo();
if (Personality == EHPersonality::MSVC_CXX)
calculateWinCXXEHStateNumbers(&fn, EHInfo);
else if (isAsynchronousEHPersonality(Personality))
calculateSEHStateNumbers(&fn, EHInfo);
else if (Personality == EHPersonality::CoreCLR)
calculateClrEHStateNumbers(&fn, EHInfo);
// Map all BB references in the WinEH data to MBBs.
for (WinEHTryBlockMapEntry &TBME : EHInfo.TryBlockMap) {
for (WinEHHandlerType &H : TBME.HandlerArray) {
if (const AllocaInst *AI = H.CatchObj.Alloca)
CatchObjects.insert({AI, {}}).first->second.push_back(
&H.CatchObj.FrameIndex);
else
H.CatchObj.FrameIndex = INT_MAX;
}
}
}
if (Personality == EHPersonality::Wasm_CXX) {
WasmEHFuncInfo &EHInfo = *MF->getWasmEHFuncInfo();
calculateWasmEHInfo(&fn, EHInfo);
}
// Initialize the mapping of values to registers. This is only set up for
// instruction values that are used outside of the block that defines
// them.
for (const BasicBlock &BB : *Fn) {
for (const Instruction &I : BB) {
if (const AllocaInst *AI = dyn_cast<AllocaInst>(&I)) {
Type *Ty = AI->getAllocatedType();
unsigned Align =
std::max((unsigned)MF->getDataLayout().getPrefTypeAlignment(Ty),
AI->getAlignment());
// Static allocas can be folded into the initial stack frame
// adjustment. For targets that don't realign the stack, don't
// do this if there is an extra alignment requirement.
if (AI->isStaticAlloca() &&
(TFI->isStackRealignable() || (Align <= StackAlign))) {
const ConstantInt *CUI = cast<ConstantInt>(AI->getArraySize());
uint64_t TySize = MF->getDataLayout().getTypeAllocSize(Ty);
TySize *= CUI->getZExtValue(); // Get total allocated size.
if (TySize == 0) TySize = 1; // Don't create zero-sized stack objects.
int FrameIndex = INT_MAX;
auto Iter = CatchObjects.find(AI);
if (Iter != CatchObjects.end() && TLI->needsFixedCatchObjects()) {
FrameIndex = MF->getFrameInfo().CreateFixedObject(
TySize, 0, /*Immutable=*/false, /*isAliased=*/true);
MF->getFrameInfo().setObjectAlignment(FrameIndex, Align);
} else {
FrameIndex =
MF->getFrameInfo().CreateStackObject(TySize, Align, false, AI);
}
StaticAllocaMap[AI] = FrameIndex;
// Update the catch handler information.
if (Iter != CatchObjects.end()) {
for (int *CatchObjPtr : Iter->second)
*CatchObjPtr = FrameIndex;
}
} else {
// FIXME: Overaligned static allocas should be grouped into
// a single dynamic allocation instead of using a separate
// stack allocation for each one.
if (Align <= StackAlign)
Align = 0;
// Inform the Frame Information that we have variable-sized objects.
MF->getFrameInfo().CreateVariableSizedObject(Align ? Align : 1, AI);
}
}
// Look for inline asm that clobbers the SP register.
if (isa<CallInst>(I) || isa<InvokeInst>(I)) {
ImmutableCallSite CS(&I);
if (isa<InlineAsm>(CS.getCalledValue())) {
unsigned SP = TLI->getStackPointerRegisterToSaveRestore();
const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo();
std::vector<TargetLowering::AsmOperandInfo> Ops =
TLI->ParseConstraints(Fn->getParent()->getDataLayout(), TRI, CS);
for (TargetLowering::AsmOperandInfo &Op : Ops) {
if (Op.Type == InlineAsm::isClobber) {
// Clobbers don't have SDValue operands, hence SDValue().
TLI->ComputeConstraintToUse(Op, SDValue(), DAG);
std::pair<unsigned, const TargetRegisterClass *> PhysReg =
TLI->getRegForInlineAsmConstraint(TRI, Op.ConstraintCode,
Op.ConstraintVT);
if (PhysReg.first == SP)
MF->getFrameInfo().setHasOpaqueSPAdjustment(true);
}
}
}
}
// Look for calls to the @llvm.va_start intrinsic. We can omit some
// prologue boilerplate for variadic functions that don't examine their
// arguments.
if (const auto *II = dyn_cast<IntrinsicInst>(&I)) {
if (II->getIntrinsicID() == Intrinsic::vastart)
MF->getFrameInfo().setHasVAStart(true);
}
// If we have a musttail call in a variadic function, we need to ensure we
// forward implicit register parameters.
if (const auto *CI = dyn_cast<CallInst>(&I)) {
if (CI->isMustTailCall() && Fn->isVarArg())
MF->getFrameInfo().setHasMustTailInVarArgFunc(true);
}
// Mark values used outside their block as exported, by allocating
// a virtual register for them.
if (isUsedOutsideOfDefiningBlock(&I))
if (!isa<AllocaInst>(I) || !StaticAllocaMap.count(cast<AllocaInst>(&I)))
InitializeRegForValue(&I);
// Decide the preferred extend type for a value.
PreferredExtendType[&I] = getPreferredExtendForValue(&I);
}
}
// Create an initial MachineBasicBlock for each LLVM BasicBlock in F. This
// also creates the initial PHI MachineInstrs, though none of the input
// operands are populated.
for (const BasicBlock &BB : *Fn) {
// Don't create MachineBasicBlocks for imaginary EH pad blocks. These blocks
// are really data, and no instructions can live here.
if (BB.isEHPad()) {
const Instruction *PadInst = BB.getFirstNonPHI();
// If this is a non-landingpad EH pad, mark this function as using
// funclets.
// FIXME: SEH catchpads do not create EH scope/funclets, so we could avoid
// setting this in such cases in order to improve frame layout.
if (!isa<LandingPadInst>(PadInst)) {
MF->setHasEHScopes(true);
MF->setHasEHFunclets(true);
MF->getFrameInfo().setHasOpaqueSPAdjustment(true);
}
if (isa<CatchSwitchInst>(PadInst)) {
assert(&*BB.begin() == PadInst &&
"WinEHPrepare failed to remove PHIs from imaginary BBs");
continue;
}
if (isa<FuncletPadInst>(PadInst))
assert(&*BB.begin() == PadInst && "WinEHPrepare failed to demote PHIs");
}
MachineBasicBlock *MBB = mf.CreateMachineBasicBlock(&BB);
MBBMap[&BB] = MBB;
MF->push_back(MBB);
// Transfer the address-taken flag. This is necessary because there could
// be multiple MachineBasicBlocks corresponding to one BasicBlock, and only
// the first one should be marked.
if (BB.hasAddressTaken())
MBB->setHasAddressTaken();
// Mark landing pad blocks.
if (BB.isEHPad())
MBB->setIsEHPad();
// Create Machine PHI nodes for LLVM PHI nodes, lowering them as
// appropriate.
for (const PHINode &PN : BB.phis()) {
if (PN.use_empty())
continue;
// Skip empty types
if (PN.getType()->isEmptyTy())
continue;
DebugLoc DL = PN.getDebugLoc();
unsigned PHIReg = ValueMap[&PN];
assert(PHIReg && "PHI node does not have an assigned virtual register!");
SmallVector<EVT, 4> ValueVTs;
ComputeValueVTs(*TLI, MF->getDataLayout(), PN.getType(), ValueVTs);
for (EVT VT : ValueVTs) {
unsigned NumRegisters = TLI->getNumRegisters(Fn->getContext(), VT);
const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();
for (unsigned i = 0; i != NumRegisters; ++i)
BuildMI(MBB, DL, TII->get(TargetOpcode::PHI), PHIReg + i);
PHIReg += NumRegisters;
}
}
}
if (isFuncletEHPersonality(Personality)) {
WinEHFuncInfo &EHInfo = *MF->getWinEHFuncInfo();
// Map all BB references in the WinEH data to MBBs.
for (WinEHTryBlockMapEntry &TBME : EHInfo.TryBlockMap) {
for (WinEHHandlerType &H : TBME.HandlerArray) {
if (H.Handler)
H.Handler = MBBMap[H.Handler.get<const BasicBlock *>()];
}
}
for (CxxUnwindMapEntry &UME : EHInfo.CxxUnwindMap)
if (UME.Cleanup)
UME.Cleanup = MBBMap[UME.Cleanup.get<const BasicBlock *>()];
for (SEHUnwindMapEntry &UME : EHInfo.SEHUnwindMap) {
const auto *BB = UME.Handler.get<const BasicBlock *>();
UME.Handler = MBBMap[BB];
}
for (ClrEHUnwindMapEntry &CME : EHInfo.ClrEHUnwindMap) {
const auto *BB = CME.Handler.get<const BasicBlock *>();
CME.Handler = MBBMap[BB];
}
}
else if (Personality == EHPersonality::Wasm_CXX) {
WasmEHFuncInfo &EHInfo = *MF->getWasmEHFuncInfo();
// Map all BB references in the WinEH data to MBBs.
DenseMap<BBOrMBB, BBOrMBB> NewMap;
for (auto &KV : EHInfo.EHPadUnwindMap) {
const auto *Src = KV.first.get<const BasicBlock *>();
const auto *Dst = KV.second.get<const BasicBlock *>();
NewMap[MBBMap[Src]] = MBBMap[Dst];
}
EHInfo.EHPadUnwindMap = std::move(NewMap);
NewMap.clear();
for (auto &KV : EHInfo.ThrowUnwindMap) {
const auto *Src = KV.first.get<const BasicBlock *>();
const auto *Dst = KV.second.get<const BasicBlock *>();
NewMap[MBBMap[Src]] = MBBMap[Dst];
}
EHInfo.ThrowUnwindMap = std::move(NewMap);
}
}
/// clear - Clear out all the function-specific state. This returns this
/// FunctionLoweringInfo to an empty state, ready to be used for a
/// different function.
void FunctionLoweringInfo::clear() {
MBBMap.clear();
ValueMap.clear();
VirtReg2Value.clear();
StaticAllocaMap.clear();
LiveOutRegInfo.clear();
VisitedBBs.clear();
ArgDbgValues.clear();
ByValArgFrameIndexMap.clear();
RegFixups.clear();
[FastISel] Sink local value materializations to first use Summary: Local values are constants, global addresses, and stack addresses that can't be folded into the instruction that uses them. For example, when storing the address of a global variable into memory, we need to materialize that address into a register. FastISel doesn't want to materialize any given local value more than once, so it generates all local value materialization code at EmitStartPt, which always dominates the current insertion point. This allows it to maintain a map of local value registers, and it knows that the local value area will always dominate the current insertion point. The downside is that local value instructions are always emitted without a source location. This is done to prevent jumpy line tables, but it means that the local value area will be considered part of the previous statement. Consider this C code: call1(); // line 1 ++global; // line 2 ++global; // line 3 call2(&global, &local); // line 4 Today we end up with assembly and line tables like this: .loc 1 1 callq call1 leaq global(%rip), %rdi leaq local(%rsp), %rsi .loc 1 2 addq $1, global(%rip) .loc 1 3 addq $1, global(%rip) .loc 1 4 callq call2 The LEA instructions in the local value area have no source location and are treated as being on line 1. Stepping through the code in a debugger and correlating it with the assembly won't make much sense, because these materializations are only required for line 4. This is actually problematic for the VS debugger "set next statement" feature, which effectively assumes that there are no registers live across statement boundaries. By sinking the local value code into the statement and fixing up the source location, we can make that feature work. This was filed as https://bugs.llvm.org/show_bug.cgi?id=35975 and https://crbug.com/793819. This change is obviously not enough to make this feature work reliably in all cases, but I felt that it was worth doing anyway because it usually generates smaller, more comprehensible -O0 code. I measured a 0.12% regression in code generation time with LLC on the sqlite3 amalgamation, so I think this is worth doing. There are some special cases worth calling out in the commit message: 1. local values materialized for phis 2. local values used by no-op casts 3. dead local value code Local values can be materialized for phis, and this does not show up as a vreg use in MachineRegisterInfo. In this case, if there are no other uses, this patch sinks the value to the first terminator, EH label, or the end of the BB if nothing else exists. Local values may also be used by no-op casts, which adds the register to the RegFixups table. Without reversing the RegFixups map direction, we don't have enough information to sink these instructions. Lastly, if the local value register has no other uses, we can delete it. This comes up when fastisel tries two instruction selection approaches and the first materializes the value but fails and the second succeeds without using the local value. Reviewers: aprantl, dblaikie, qcolombet, MatzeB, vsk, echristo Subscribers: dotdash, chandlerc, hans, sdardis, amccarth, javed.absar, zturner, llvm-commits, hiraditya Differential Revision: https://reviews.llvm.org/D43093 llvm-svn: 327581
2018-03-15 05:54:21 +08:00
RegsWithFixups.clear();
[Statepoints 3/4] Statepoint infrastructure for garbage collection: SelectionDAGBuilder This is the third patch in a small series. It contains the CodeGen support for lowering the gc.statepoint intrinsic sequences (223078) to the STATEPOINT pseudo machine instruction (223085). The change also includes the set of helper routines and classes for working with gc.statepoints, gc.relocates, and gc.results since the lowering code uses them. With this change, gc.statepoints should be functionally complete. The documentation will follow in the fourth change, and there will likely be some cleanup changes, but interested parties can start experimenting now. I'm not particularly happy with the amount of code or complexity involved with the lowering step, but at least it's fairly well isolated. The statepoint lowering code is split into it's own files and anyone not working on the statepoint support itself should be able to ignore it. During the lowering process, we currently spill aggressively to stack. This is not entirely ideal (and we have plans to do better), but it's functional, relatively straight forward, and matches closely the implementations of the patchpoint intrinsics. Most of the complexity comes from trying to keep relocated copies of values in the same stack slots across statepoints. Doing so avoids the insertion of pointless load and store instructions to reshuffle the stack. The current implementation isn't as effective as I'd like, but it is functional and 'good enough' for many common use cases. In the long term, I'd like to figure out how to integrate the statepoint lowering with the register allocator. In principal, we shouldn't need to eagerly spill at all. The register allocator should do any spilling required and the statepoint should simply record that fact. Depending on how challenging that turns out to be, we may invest in a smarter global stack slot assignment mechanism as a stop gap measure. Reviewed by: atrick, ributzka llvm-svn: 223137
2014-12-03 02:50:36 +08:00
StatepointStackSlots.clear();
StatepointSpillMaps.clear();
PreferredExtendType.clear();
}
/// CreateReg - Allocate a single virtual register for the given type.
unsigned FunctionLoweringInfo::CreateReg(MVT VT) {
return RegInfo->createVirtualRegister(
MF->getSubtarget().getTargetLowering()->getRegClassFor(VT));
}
/// CreateRegs - Allocate the appropriate number of virtual registers of
/// the correctly promoted or expanded types. Assign these registers
/// consecutive vreg numbers and return the first assigned number.
///
/// In the case that the given value has struct or array type, this function
/// will assign registers for each member or element.
///
unsigned FunctionLoweringInfo::CreateRegs(Type *Ty) {
const TargetLowering *TLI = MF->getSubtarget().getTargetLowering();
SmallVector<EVT, 4> ValueVTs;
ComputeValueVTs(*TLI, MF->getDataLayout(), Ty, ValueVTs);
unsigned FirstReg = 0;
for (unsigned Value = 0, e = ValueVTs.size(); Value != e; ++Value) {
EVT ValueVT = ValueVTs[Value];
MVT RegisterVT = TLI->getRegisterType(Ty->getContext(), ValueVT);
unsigned NumRegs = TLI->getNumRegisters(Ty->getContext(), ValueVT);
for (unsigned i = 0; i != NumRegs; ++i) {
unsigned R = CreateReg(RegisterVT);
if (!FirstReg) FirstReg = R;
}
}
return FirstReg;
}
/// GetLiveOutRegInfo - Gets LiveOutInfo for a register, returning NULL if the
/// register is a PHI destination and the PHI's LiveOutInfo is not valid. If
/// the register's LiveOutInfo is for a smaller bit width, it is extended to
/// the larger bit width by zero extension. The bit width must be no smaller
/// than the LiveOutInfo's existing bit width.
const FunctionLoweringInfo::LiveOutInfo *
FunctionLoweringInfo::GetLiveOutRegInfo(unsigned Reg, unsigned BitWidth) {
if (!LiveOutRegInfo.inBounds(Reg))
return nullptr;
LiveOutInfo *LOI = &LiveOutRegInfo[Reg];
if (!LOI->IsValid)
return nullptr;
if (BitWidth > LOI->Known.getBitWidth()) {
LOI->NumSignBits = 1;
LOI->Known = LOI->Known.zextOrTrunc(BitWidth);
}
return LOI;
}
/// ComputePHILiveOutRegInfo - Compute LiveOutInfo for a PHI's destination
/// register based on the LiveOutInfo of its operands.
void FunctionLoweringInfo::ComputePHILiveOutRegInfo(const PHINode *PN) {
Type *Ty = PN->getType();
if (!Ty->isIntegerTy() || Ty->isVectorTy())
return;
SmallVector<EVT, 1> ValueVTs;
ComputeValueVTs(*TLI, MF->getDataLayout(), Ty, ValueVTs);
assert(ValueVTs.size() == 1 &&
"PHIs with non-vector integer types should have a single VT.");
EVT IntVT = ValueVTs[0];
if (TLI->getNumRegisters(PN->getContext(), IntVT) != 1)
return;
IntVT = TLI->getTypeToTransformTo(PN->getContext(), IntVT);
unsigned BitWidth = IntVT.getSizeInBits();
unsigned DestReg = ValueMap[PN];
if (!TargetRegisterInfo::isVirtualRegister(DestReg))
return;
LiveOutRegInfo.grow(DestReg);
LiveOutInfo &DestLOI = LiveOutRegInfo[DestReg];
Value *V = PN->getIncomingValue(0);
if (isa<UndefValue>(V) || isa<ConstantExpr>(V)) {
DestLOI.NumSignBits = 1;
DestLOI.Known = KnownBits(BitWidth);
return;
}
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
APInt Val = CI->getValue().zextOrTrunc(BitWidth);
DestLOI.NumSignBits = Val.getNumSignBits();
DestLOI.Known.Zero = ~Val;
DestLOI.Known.One = Val;
} else {
assert(ValueMap.count(V) && "V should have been placed in ValueMap when its"
"CopyToReg node was created.");
unsigned SrcReg = ValueMap[V];
if (!TargetRegisterInfo::isVirtualRegister(SrcReg)) {
DestLOI.IsValid = false;
return;
}
const LiveOutInfo *SrcLOI = GetLiveOutRegInfo(SrcReg, BitWidth);
if (!SrcLOI) {
DestLOI.IsValid = false;
return;
}
DestLOI = *SrcLOI;
}
assert(DestLOI.Known.Zero.getBitWidth() == BitWidth &&
DestLOI.Known.One.getBitWidth() == BitWidth &&
"Masks should have the same bit width as the type.");
for (unsigned i = 1, e = PN->getNumIncomingValues(); i != e; ++i) {
Value *V = PN->getIncomingValue(i);
if (isa<UndefValue>(V) || isa<ConstantExpr>(V)) {
DestLOI.NumSignBits = 1;
DestLOI.Known = KnownBits(BitWidth);
return;
}
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
APInt Val = CI->getValue().zextOrTrunc(BitWidth);
DestLOI.NumSignBits = std::min(DestLOI.NumSignBits, Val.getNumSignBits());
DestLOI.Known.Zero &= ~Val;
DestLOI.Known.One &= Val;
continue;
}
assert(ValueMap.count(V) && "V should have been placed in ValueMap when "
"its CopyToReg node was created.");
unsigned SrcReg = ValueMap[V];
if (!TargetRegisterInfo::isVirtualRegister(SrcReg)) {
DestLOI.IsValid = false;
return;
}
const LiveOutInfo *SrcLOI = GetLiveOutRegInfo(SrcReg, BitWidth);
if (!SrcLOI) {
DestLOI.IsValid = false;
return;
}
DestLOI.NumSignBits = std::min(DestLOI.NumSignBits, SrcLOI->NumSignBits);
DestLOI.Known.Zero &= SrcLOI->Known.Zero;
DestLOI.Known.One &= SrcLOI->Known.One;
}
}
/// setArgumentFrameIndex - Record frame index for the byval
/// argument. This overrides previous frame index entry for this argument,
/// if any.
void FunctionLoweringInfo::setArgumentFrameIndex(const Argument *A,
int FI) {
ByValArgFrameIndexMap[A] = FI;
}
/// getArgumentFrameIndex - Get frame index for the byval argument.
/// If the argument does not have any assigned frame index then 0 is
/// returned.
int FunctionLoweringInfo::getArgumentFrameIndex(const Argument *A) {
auto I = ByValArgFrameIndexMap.find(A);
if (I != ByValArgFrameIndexMap.end())
return I->second;
LLVM_DEBUG(dbgs() << "Argument does not have assigned frame index!\n");
return INT_MAX;
}
unsigned FunctionLoweringInfo::getCatchPadExceptionPointerVReg(
const Value *CPI, const TargetRegisterClass *RC) {
MachineRegisterInfo &MRI = MF->getRegInfo();
auto I = CatchPadExceptionPointers.insert({CPI, 0});
unsigned &VReg = I.first->second;
if (I.second)
VReg = MRI.createVirtualRegister(RC);
assert(VReg && "null vreg in exception pointer table!");
return VReg;
}
unsigned
FunctionLoweringInfo::getOrCreateSwiftErrorVReg(const MachineBasicBlock *MBB,
const Value *Val) {
auto Key = std::make_pair(MBB, Val);
auto It = SwiftErrorVRegDefMap.find(Key);
// If this is the first use of this swifterror value in this basic block,
// create a new virtual register.
// After we processed all basic blocks we will satisfy this "upwards exposed
// use" by inserting a copy or phi at the beginning of this block.
if (It == SwiftErrorVRegDefMap.end()) {
auto &DL = MF->getDataLayout();
const TargetRegisterClass *RC = TLI->getRegClassFor(TLI->getPointerTy(DL));
auto VReg = MF->getRegInfo().createVirtualRegister(RC);
SwiftErrorVRegDefMap[Key] = VReg;
SwiftErrorVRegUpwardsUse[Key] = VReg;
return VReg;
} else return It->second;
Swift Calling Convention: swifterror target-independent change. At IR level, the swifterror argument is an input argument with type ErrorObject**. For targets that support swifterror, we want to optimize it to behave as an inout value with type ErrorObject*; it will be passed in a fixed physical register. The main idea is to track the virtual registers for each swifterror value. We define swifterror values as AllocaInsts with swifterror attribute or a function argument with swifterror attribute. In SelectionDAGISel.cpp, we set up swifterror values (SwiftErrorVals) before handling the basic blocks. When iterating over all basic blocks in RPO, before actually visiting the basic block, we call mergeIncomingSwiftErrors to merge incoming swifterror values when there are multiple predecessors or to simply propagate them. There, we create a virtual register for each swifterror value in the entry block. For predecessors that are not yet visited, we create virtual registers to hold the swifterror values at the end of the predecessor. The assignments are saved in SwiftErrorWorklist and will be materialized at the end of visiting the basic block. When visiting a load from a swifterror value, we copy from the current virtual register assignment. When visiting a store to a swifterror value, we create a virtual register to hold the swifterror value and update SwiftErrorMap to track the current virtual register assignment. Differential Revision: http://reviews.llvm.org/D18108 llvm-svn: 265433
2016-04-06 02:13:16 +08:00
}
void FunctionLoweringInfo::setCurrentSwiftErrorVReg(
const MachineBasicBlock *MBB, const Value *Val, unsigned VReg) {
SwiftErrorVRegDefMap[std::make_pair(MBB, Val)] = VReg;
Swift Calling Convention: swifterror target-independent change. At IR level, the swifterror argument is an input argument with type ErrorObject**. For targets that support swifterror, we want to optimize it to behave as an inout value with type ErrorObject*; it will be passed in a fixed physical register. The main idea is to track the virtual registers for each swifterror value. We define swifterror values as AllocaInsts with swifterror attribute or a function argument with swifterror attribute. In SelectionDAGISel.cpp, we set up swifterror values (SwiftErrorVals) before handling the basic blocks. When iterating over all basic blocks in RPO, before actually visiting the basic block, we call mergeIncomingSwiftErrors to merge incoming swifterror values when there are multiple predecessors or to simply propagate them. There, we create a virtual register for each swifterror value in the entry block. For predecessors that are not yet visited, we create virtual registers to hold the swifterror values at the end of the predecessor. The assignments are saved in SwiftErrorWorklist and will be materialized at the end of visiting the basic block. When visiting a load from a swifterror value, we copy from the current virtual register assignment. When visiting a store to a swifterror value, we create a virtual register to hold the swifterror value and update SwiftErrorMap to track the current virtual register assignment. Differential Revision: http://reviews.llvm.org/D18108 llvm-svn: 265433
2016-04-06 02:13:16 +08:00
}
std::pair<unsigned, bool>
FunctionLoweringInfo::getOrCreateSwiftErrorVRegDefAt(const Instruction *I) {
auto Key = PointerIntPair<const Instruction *, 1, bool>(I, true);
auto It = SwiftErrorVRegDefUses.find(Key);
if (It == SwiftErrorVRegDefUses.end()) {
auto &DL = MF->getDataLayout();
const TargetRegisterClass *RC = TLI->getRegClassFor(TLI->getPointerTy(DL));
unsigned VReg = MF->getRegInfo().createVirtualRegister(RC);
SwiftErrorVRegDefUses[Key] = VReg;
return std::make_pair(VReg, true);
}
return std::make_pair(It->second, false);
}
std::pair<unsigned, bool>
FunctionLoweringInfo::getOrCreateSwiftErrorVRegUseAt(const Instruction *I, const MachineBasicBlock *MBB, const Value *Val) {
auto Key = PointerIntPair<const Instruction *, 1, bool>(I, false);
auto It = SwiftErrorVRegDefUses.find(Key);
if (It == SwiftErrorVRegDefUses.end()) {
unsigned VReg = getOrCreateSwiftErrorVReg(MBB, Val);
SwiftErrorVRegDefUses[Key] = VReg;
return std::make_pair(VReg, true);
}
return std::make_pair(It->second, false);
}
const Value *
FunctionLoweringInfo::getValueFromVirtualReg(unsigned Vreg) {
if (VirtReg2Value.empty()) {
AMDGPU: Fix various issues around the VirtReg2Value mapping Summary: The VirtReg2Value mapping is crucial for getting consistently reliable divergence information into the SelectionDAG. This patch fixes a bunch of issues that lead to incorrect divergence info and introduces tight assertions to ensure we don't regress: 1. VirtReg2Value is generated lazily; there were some cases where a lookup was performed before all relevant virtual registers were created, leading to an out-of-sync mapping. Those cases were: - Complex code to lower formal arguments that generated CopyFromReg nodes from live-in registers (fixed by never querying the mapping for live-in registers). - Code that generates CopyToReg for formal arguments that are used outside the entry basic block (fixed by never querying the mapping for Register nodes, which don't need the divergence info anyway). 2. For complex values that are lowered to a sequence of registers, all registers must be reflected in the VirtReg2Value mapping. I am not adding any new tests, since I'm not actually aware of any bugs that these problems are causing with trunk as-is. However, I recently added a test case (in r346423) which fails when D53283 is applied without this change. Also, the new assertions should provide most of the effective test coverage. There is one test change in sdwa-peephole.ll. The underlying issue is that since the divergence info is now correct, the DAGISel will select V_OR_B32 directly instead of S_OR_B32. This leads to an extra COPY which affects the behavior of MachineLICM in a way that ends up with the S_MOV_B32 with the constant in a different basic block than the V_OR_B32, which is presumably what defeats the peephole. Reviewers: alex-t, arsenm, rampitec Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D54340 llvm-svn: 348049
2018-12-01 06:55:29 +08:00
SmallVector<EVT, 4> ValueVTs;
for (auto &P : ValueMap) {
AMDGPU: Fix various issues around the VirtReg2Value mapping Summary: The VirtReg2Value mapping is crucial for getting consistently reliable divergence information into the SelectionDAG. This patch fixes a bunch of issues that lead to incorrect divergence info and introduces tight assertions to ensure we don't regress: 1. VirtReg2Value is generated lazily; there were some cases where a lookup was performed before all relevant virtual registers were created, leading to an out-of-sync mapping. Those cases were: - Complex code to lower formal arguments that generated CopyFromReg nodes from live-in registers (fixed by never querying the mapping for live-in registers). - Code that generates CopyToReg for formal arguments that are used outside the entry basic block (fixed by never querying the mapping for Register nodes, which don't need the divergence info anyway). 2. For complex values that are lowered to a sequence of registers, all registers must be reflected in the VirtReg2Value mapping. I am not adding any new tests, since I'm not actually aware of any bugs that these problems are causing with trunk as-is. However, I recently added a test case (in r346423) which fails when D53283 is applied without this change. Also, the new assertions should provide most of the effective test coverage. There is one test change in sdwa-peephole.ll. The underlying issue is that since the divergence info is now correct, the DAGISel will select V_OR_B32 directly instead of S_OR_B32. This leads to an extra COPY which affects the behavior of MachineLICM in a way that ends up with the S_MOV_B32 with the constant in a different basic block than the V_OR_B32, which is presumably what defeats the peephole. Reviewers: alex-t, arsenm, rampitec Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D54340 llvm-svn: 348049
2018-12-01 06:55:29 +08:00
ValueVTs.clear();
ComputeValueVTs(*TLI, Fn->getParent()->getDataLayout(),
P.first->getType(), ValueVTs);
unsigned Reg = P.second;
for (EVT VT : ValueVTs) {
unsigned NumRegisters = TLI->getNumRegisters(Fn->getContext(), VT);
for (unsigned i = 0, e = NumRegisters; i != e; ++i)
VirtReg2Value[Reg++] = P.first;
}
}
}
AMDGPU: Fix various issues around the VirtReg2Value mapping Summary: The VirtReg2Value mapping is crucial for getting consistently reliable divergence information into the SelectionDAG. This patch fixes a bunch of issues that lead to incorrect divergence info and introduces tight assertions to ensure we don't regress: 1. VirtReg2Value is generated lazily; there were some cases where a lookup was performed before all relevant virtual registers were created, leading to an out-of-sync mapping. Those cases were: - Complex code to lower formal arguments that generated CopyFromReg nodes from live-in registers (fixed by never querying the mapping for live-in registers). - Code that generates CopyToReg for formal arguments that are used outside the entry basic block (fixed by never querying the mapping for Register nodes, which don't need the divergence info anyway). 2. For complex values that are lowered to a sequence of registers, all registers must be reflected in the VirtReg2Value mapping. I am not adding any new tests, since I'm not actually aware of any bugs that these problems are causing with trunk as-is. However, I recently added a test case (in r346423) which fails when D53283 is applied without this change. Also, the new assertions should provide most of the effective test coverage. There is one test change in sdwa-peephole.ll. The underlying issue is that since the divergence info is now correct, the DAGISel will select V_OR_B32 directly instead of S_OR_B32. This leads to an extra COPY which affects the behavior of MachineLICM in a way that ends up with the S_MOV_B32 with the constant in a different basic block than the V_OR_B32, which is presumably what defeats the peephole. Reviewers: alex-t, arsenm, rampitec Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D54340 llvm-svn: 348049
2018-12-01 06:55:29 +08:00
return VirtReg2Value.lookup(Vreg);
}