hanchenye-llvm-project/clang/lib/CodeGen/CGValue.h

453 lines
13 KiB
C
Raw Normal View History

//===-- CGValue.h - LLVM CodeGen wrappers for llvm::Value* ------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// These classes implement wrappers around llvm::Value in order to
// fully represent the range of values for C L- and R- values.
//
//===----------------------------------------------------------------------===//
#ifndef CLANG_CODEGEN_CGVALUE_H
#define CLANG_CODEGEN_CGVALUE_H
#include "clang/AST/ASTContext.h"
#include "clang/AST/Type.h"
namespace llvm {
class Constant;
class Value;
}
namespace clang {
class ObjCPropertyRefExpr;
namespace CodeGen {
class CGBitFieldInfo;
/// RValue - This trivial value class is used to represent the result of an
/// expression that is evaluated. It can be one of three things: either a
/// simple LLVM SSA value, a pair of SSA values for complex numbers, or the
/// address of an aggregate value in memory.
class RValue {
enum Flavor { Scalar, Complex, Aggregate };
// Stores first value and flavor.
llvm::PointerIntPair<llvm::Value *, 2, Flavor> V1;
// Stores second value and volatility.
llvm::PointerIntPair<llvm::Value *, 1, bool> V2;
public:
bool isScalar() const { return V1.getInt() == Scalar; }
bool isComplex() const { return V1.getInt() == Complex; }
bool isAggregate() const { return V1.getInt() == Aggregate; }
bool isVolatileQualified() const { return V2.getInt(); }
2009-11-04 00:11:57 +08:00
/// getScalarVal() - Return the Value* of this scalar value.
llvm::Value *getScalarVal() const {
assert(isScalar() && "Not a scalar!");
return V1.getPointer();
}
/// getComplexVal - Return the real/imag components of this complex value.
///
std::pair<llvm::Value *, llvm::Value *> getComplexVal() const {
return std::make_pair(V1.getPointer(), V2.getPointer());
}
/// getAggregateAddr() - Return the Value* of the address of the aggregate.
llvm::Value *getAggregateAddr() const {
assert(isAggregate() && "Not an aggregate!");
return V1.getPointer();
}
static RValue get(llvm::Value *V) {
RValue ER;
ER.V1.setPointer(V);
ER.V1.setInt(Scalar);
ER.V2.setInt(false);
return ER;
}
static RValue getComplex(llvm::Value *V1, llvm::Value *V2) {
RValue ER;
ER.V1.setPointer(V1);
ER.V2.setPointer(V2);
ER.V1.setInt(Complex);
ER.V2.setInt(false);
return ER;
}
static RValue getComplex(const std::pair<llvm::Value *, llvm::Value *> &C) {
return getComplex(C.first, C.second);
}
// FIXME: Aggregate rvalues need to retain information about whether they are
// volatile or not. Remove default to find all places that probably get this
// wrong.
static RValue getAggregate(llvm::Value *V, bool Volatile = false) {
RValue ER;
ER.V1.setPointer(V);
ER.V1.setInt(Aggregate);
ER.V2.setInt(Volatile);
return ER;
}
};
/// LValue - This represents an lvalue references. Because C/C++ allow
/// bitfields, this is not a simple LLVM pointer, it may be a pointer plus a
/// bitrange.
class LValue {
enum {
Simple, // This is a normal l-value, use getAddress().
VectorElt, // This is a vector element l-value (V[i]), use getVector*
BitField, // This is a bitfield l-value, use getBitfield*.
ExtVectorElt, // This is an extended vector subset, use getExtVectorComp
PropertyRef // This is an Objective-C property reference, use
// getPropertyRefExpr
} LVType;
llvm::Value *V;
union {
// Index into a vector subscript: V[i]
llvm::Value *VectorIdx;
// ExtVector element subset: V.xyx
llvm::Constant *VectorElts;
// BitField start bit and size
const CGBitFieldInfo *BitFieldInfo;
// Obj-C property reference expression
const ObjCPropertyRefExpr *PropertyRefExpr;
};
QualType Type;
// 'const' is unused here
Qualifiers Quals;
/// The alignment to use when accessing this lvalue.
unsigned short Alignment;
// objective-c's ivar
bool Ivar:1;
// objective-c's ivar is an array
bool ObjIsArray:1;
// LValue is non-gc'able for any reason, including being a parameter or local
// variable.
bool NonGC: 1;
// Lvalue is a global reference of an objective-c object
bool GlobalObjCRef : 1;
// Lvalue is a thread local reference
bool ThreadLocalRef : 1;
Expr *BaseIvarExp;
/// TBAAInfo - TBAA information to attach to dereferences of this LValue.
llvm::MDNode *TBAAInfo;
private:
void Initialize(QualType Type, Qualifiers Quals, unsigned Alignment = 0,
llvm::MDNode *TBAAInfo = 0) {
this->Type = Type;
this->Quals = Quals;
this->Alignment = Alignment;
assert(this->Alignment == Alignment && "Alignment exceeds allowed max!");
// Initialize Objective-C flags.
this->Ivar = this->ObjIsArray = this->NonGC = this->GlobalObjCRef = false;
this->ThreadLocalRef = false;
this->BaseIvarExp = 0;
this->TBAAInfo = TBAAInfo;
}
public:
bool isSimple() const { return LVType == Simple; }
bool isVectorElt() const { return LVType == VectorElt; }
bool isBitField() const { return LVType == BitField; }
bool isExtVectorElt() const { return LVType == ExtVectorElt; }
bool isPropertyRef() const { return LVType == PropertyRef; }
bool isVolatileQualified() const { return Quals.hasVolatile(); }
bool isRestrictQualified() const { return Quals.hasRestrict(); }
unsigned getVRQualifiers() const {
return Quals.getCVRQualifiers() & ~Qualifiers::Const;
}
QualType getType() const { return Type; }
Qualifiers::ObjCLifetime getObjCLifetime() const {
return Quals.getObjCLifetime();
}
bool isObjCIvar() const { return Ivar; }
void setObjCIvar(bool Value) { Ivar = Value; }
bool isObjCArray() const { return ObjIsArray; }
void setObjCArray(bool Value) { ObjIsArray = Value; }
bool isNonGC () const { return NonGC; }
void setNonGC(bool Value) { NonGC = Value; }
bool isGlobalObjCRef() const { return GlobalObjCRef; }
void setGlobalObjCRef(bool Value) { GlobalObjCRef = Value; }
bool isThreadLocalRef() const { return ThreadLocalRef; }
void setThreadLocalRef(bool Value) { ThreadLocalRef = Value;}
bool isObjCWeak() const {
return Quals.getObjCGCAttr() == Qualifiers::Weak;
}
bool isObjCStrong() const {
return Quals.getObjCGCAttr() == Qualifiers::Strong;
}
bool isVolatile() const {
return Quals.hasVolatile();
}
Expr *getBaseIvarExp() const { return BaseIvarExp; }
void setBaseIvarExp(Expr *V) { BaseIvarExp = V; }
llvm::MDNode *getTBAAInfo() const { return TBAAInfo; }
void setTBAAInfo(llvm::MDNode *N) { TBAAInfo = N; }
const Qualifiers &getQuals() const { return Quals; }
Qualifiers &getQuals() { return Quals; }
unsigned getAddressSpace() const { return Quals.getAddressSpace(); }
unsigned getAlignment() const { return Alignment; }
// simple lvalue
llvm::Value *getAddress() const { assert(isSimple()); return V; }
void setAddress(llvm::Value *address) {
assert(isSimple());
V = address;
}
// vector elt lvalue
llvm::Value *getVectorAddr() const { assert(isVectorElt()); return V; }
llvm::Value *getVectorIdx() const { assert(isVectorElt()); return VectorIdx; }
// extended vector elements.
llvm::Value *getExtVectorAddr() const { assert(isExtVectorElt()); return V; }
llvm::Constant *getExtVectorElts() const {
assert(isExtVectorElt());
return VectorElts;
}
// bitfield lvalue
llvm::Value *getBitFieldBaseAddr() const {
assert(isBitField());
return V;
}
const CGBitFieldInfo &getBitFieldInfo() const {
assert(isBitField());
return *BitFieldInfo;
}
// property ref lvalue
llvm::Value *getPropertyRefBaseAddr() const {
assert(isPropertyRef());
return V;
}
const ObjCPropertyRefExpr *getPropertyRefExpr() const {
assert(isPropertyRef());
return PropertyRefExpr;
}
static LValue MakeAddr(llvm::Value *address, QualType type,
unsigned alignment, ASTContext &Context,
llvm::MDNode *TBAAInfo = 0) {
Qualifiers qs = type.getQualifiers();
qs.setObjCGCAttr(Context.getObjCGCAttrKind(type));
LValue R;
R.LVType = Simple;
R.V = address;
R.Initialize(type, qs, alignment, TBAAInfo);
return R;
}
static LValue MakeVectorElt(llvm::Value *Vec, llvm::Value *Idx,
QualType type) {
LValue R;
R.LVType = VectorElt;
R.V = Vec;
R.VectorIdx = Idx;
R.Initialize(type, type.getQualifiers());
return R;
}
static LValue MakeExtVectorElt(llvm::Value *Vec, llvm::Constant *Elts,
QualType type) {
LValue R;
R.LVType = ExtVectorElt;
R.V = Vec;
R.VectorElts = Elts;
R.Initialize(type, type.getQualifiers());
return R;
}
/// \brief Create a new object to represent a bit-field access.
///
/// \param BaseValue - The base address of the structure containing the
/// bit-field.
/// \param Info - The information describing how to perform the bit-field
/// access.
static LValue MakeBitfield(llvm::Value *BaseValue,
const CGBitFieldInfo &Info,
QualType type) {
LValue R;
R.LVType = BitField;
R.V = BaseValue;
R.BitFieldInfo = &Info;
R.Initialize(type, type.getQualifiers());
return R;
}
2009-05-16 15:57:57 +08:00
// FIXME: It is probably bad that we aren't emitting the target when we build
// the lvalue. However, this complicates the code a bit, and I haven't figured
// out how to make it go wrong yet.
static LValue MakePropertyRef(const ObjCPropertyRefExpr *E,
llvm::Value *Base) {
LValue R;
R.LVType = PropertyRef;
R.V = Base;
R.PropertyRefExpr = E;
R.Initialize(QualType(), Qualifiers());
return R;
}
};
/// An aggregate value slot.
class AggValueSlot {
/// The address.
llvm::Value *Addr;
// Qualifiers
Qualifiers Quals;
// Associated flags.
bool LifetimeFlag : 1;
bool RequiresGCollection : 1;
Improve codegen for initializer lists to use memset more aggressively when an initializer is variable (I handled the constant case in a previous patch). This has three pieces: 1. Enhance AggValueSlot to have a 'isZeroed' bit to tell CGExprAgg that the memory being stored into has previously been memset to zero. 2. Teach CGExprAgg to not emit stores of zero to isZeroed memory. 3. Teach CodeGenFunction::EmitAggExpr to scan initializers to determine whether they are profitable to emit a memset + inividual stores vs stores for everything. The heuristic used is that a global has to be more than 16 bytes and has to be 3/4 zero to be candidate for this xform. The two testcases are illustrative of the scenarios this catches. We now codegen test9 into: call void @llvm.memset.p0i8.i64(i8* %0, i8 0, i64 400, i32 4, i1 false) %.array = getelementptr inbounds [100 x i32]* %Arr, i32 0, i32 0 %tmp = load i32* %X.addr, align 4 store i32 %tmp, i32* %.array and test10 into: call void @llvm.memset.p0i8.i64(i8* %0, i8 0, i64 392, i32 8, i1 false) %tmp = getelementptr inbounds %struct.b* %S, i32 0, i32 0 %tmp1 = getelementptr inbounds %struct.a* %tmp, i32 0, i32 0 %tmp2 = load i32* %X.addr, align 4 store i32 %tmp2, i32* %tmp1, align 4 %tmp5 = getelementptr inbounds %struct.b* %S, i32 0, i32 3 %tmp10 = getelementptr inbounds %struct.a* %tmp5, i32 0, i32 4 %tmp11 = load i32* %X.addr, align 4 store i32 %tmp11, i32* %tmp10, align 4 Previously we produced 99 stores of zero for test9 and also tons for test10. This xforms should substantially speed up -O0 builds when it kicks in as well as reducing code size and optimizer heartburn on insane cases. This resolves PR279. llvm-svn: 120692
2010-12-02 15:07:26 +08:00
/// ZeroedFlag - This is set to true if the destination is known to be zero
Improve codegen for initializer lists to use memset more aggressively when an initializer is variable (I handled the constant case in a previous patch). This has three pieces: 1. Enhance AggValueSlot to have a 'isZeroed' bit to tell CGExprAgg that the memory being stored into has previously been memset to zero. 2. Teach CGExprAgg to not emit stores of zero to isZeroed memory. 3. Teach CodeGenFunction::EmitAggExpr to scan initializers to determine whether they are profitable to emit a memset + inividual stores vs stores for everything. The heuristic used is that a global has to be more than 16 bytes and has to be 3/4 zero to be candidate for this xform. The two testcases are illustrative of the scenarios this catches. We now codegen test9 into: call void @llvm.memset.p0i8.i64(i8* %0, i8 0, i64 400, i32 4, i1 false) %.array = getelementptr inbounds [100 x i32]* %Arr, i32 0, i32 0 %tmp = load i32* %X.addr, align 4 store i32 %tmp, i32* %.array and test10 into: call void @llvm.memset.p0i8.i64(i8* %0, i8 0, i64 392, i32 8, i1 false) %tmp = getelementptr inbounds %struct.b* %S, i32 0, i32 0 %tmp1 = getelementptr inbounds %struct.a* %tmp, i32 0, i32 0 %tmp2 = load i32* %X.addr, align 4 store i32 %tmp2, i32* %tmp1, align 4 %tmp5 = getelementptr inbounds %struct.b* %S, i32 0, i32 3 %tmp10 = getelementptr inbounds %struct.a* %tmp5, i32 0, i32 4 %tmp11 = load i32* %X.addr, align 4 store i32 %tmp11, i32* %tmp10, align 4 Previously we produced 99 stores of zero for test9 and also tons for test10. This xforms should substantially speed up -O0 builds when it kicks in as well as reducing code size and optimizer heartburn on insane cases. This resolves PR279. llvm-svn: 120692
2010-12-02 15:07:26 +08:00
/// before the assignment into it. This means that zero fields don't need to
/// be set.
bool ZeroedFlag : 1;
/// AliasedFlag - This generally defaults to false, but can be true
/// if the memory is known not to be aliased.
bool AliasedFlag : 1;
public:
enum IsAliased_t { IsNotAliased, IsAliased };
enum IsDestructed_t { IsNotDestructed, IsDestructed };
enum IsZeroed_t { IsNotZeroed, IsZeroed };
enum NeedsGCBarriers_t { DoesNotNeedGCBarriers, NeedsGCBarriers };
/// ignored - Returns an aggregate value slot indicating that the
/// aggregate value is being ignored.
static AggValueSlot ignored() {
AggValueSlot AV;
AV.Addr = 0;
AV.Quals = Qualifiers();
AV.LifetimeFlag = AV.RequiresGCollection = AV.ZeroedFlag = false;
// If there's ever an address here, it will be a temporary.
AV.AliasedFlag = false;
return AV;
}
/// forAddr - Make a slot for an aggregate value.
///
/// \param quals - The qualifiers that dictate how the slot should
/// be initialied. Only 'volatile' and the Objective-C lifetime
/// qualifiers matter.
///
/// \param isDestructed - true if something else is responsible
/// for calling destructors on this object
/// \param needsGC - true if the slot is potentially located
/// somewhere that ObjC GC calls should be emitted for
static AggValueSlot forAddr(llvm::Value *addr, Qualifiers quals,
IsDestructed_t isDestructed,
NeedsGCBarriers_t needsGC,
IsAliased_t isAliased = IsAliased,
IsZeroed_t isZeroed = IsNotZeroed) {
AggValueSlot AV;
AV.Addr = addr;
AV.Quals = quals;
AV.LifetimeFlag = isDestructed;
AV.RequiresGCollection = needsGC;
AV.ZeroedFlag = isZeroed;
AV.AliasedFlag = isAliased;
return AV;
}
static AggValueSlot forLValue(LValue LV, IsDestructed_t isDestructed,
NeedsGCBarriers_t needsGC,
IsAliased_t isAliased = IsAliased,
IsZeroed_t isZeroed = IsNotZeroed) {
return forAddr(LV.getAddress(), LV.getQuals(),
isDestructed, needsGC, isAliased, isZeroed);
}
IsDestructed_t isLifetimeExternallyManaged() const {
return IsDestructed_t(LifetimeFlag);
}
void setLifetimeExternallyManaged(bool Managed = true) {
LifetimeFlag = Managed;
}
Qualifiers getQualifiers() const { return Quals; }
bool isVolatile() const {
return Quals.hasVolatile();
}
Qualifiers::ObjCLifetime getObjCLifetime() const {
return Quals.getObjCLifetime();
}
NeedsGCBarriers_t requiresGCollection() const {
return NeedsGCBarriers_t(RequiresGCollection);
}
llvm::Value *getAddr() const {
return Addr;
}
bool isIgnored() const {
return Addr == 0;
}
IsAliased_t isPotentiallyAliased() const {
return IsAliased_t(AliasedFlag);
}
RValue asRValue() const {
return RValue::getAggregate(getAddr(), isVolatile());
}
void setZeroed(bool V = true) { ZeroedFlag = V; }
IsZeroed_t isZeroed() const {
return IsZeroed_t(ZeroedFlag);
Improve codegen for initializer lists to use memset more aggressively when an initializer is variable (I handled the constant case in a previous patch). This has three pieces: 1. Enhance AggValueSlot to have a 'isZeroed' bit to tell CGExprAgg that the memory being stored into has previously been memset to zero. 2. Teach CGExprAgg to not emit stores of zero to isZeroed memory. 3. Teach CodeGenFunction::EmitAggExpr to scan initializers to determine whether they are profitable to emit a memset + inividual stores vs stores for everything. The heuristic used is that a global has to be more than 16 bytes and has to be 3/4 zero to be candidate for this xform. The two testcases are illustrative of the scenarios this catches. We now codegen test9 into: call void @llvm.memset.p0i8.i64(i8* %0, i8 0, i64 400, i32 4, i1 false) %.array = getelementptr inbounds [100 x i32]* %Arr, i32 0, i32 0 %tmp = load i32* %X.addr, align 4 store i32 %tmp, i32* %.array and test10 into: call void @llvm.memset.p0i8.i64(i8* %0, i8 0, i64 392, i32 8, i1 false) %tmp = getelementptr inbounds %struct.b* %S, i32 0, i32 0 %tmp1 = getelementptr inbounds %struct.a* %tmp, i32 0, i32 0 %tmp2 = load i32* %X.addr, align 4 store i32 %tmp2, i32* %tmp1, align 4 %tmp5 = getelementptr inbounds %struct.b* %S, i32 0, i32 3 %tmp10 = getelementptr inbounds %struct.a* %tmp5, i32 0, i32 4 %tmp11 = load i32* %X.addr, align 4 store i32 %tmp11, i32* %tmp10, align 4 Previously we produced 99 stores of zero for test9 and also tons for test10. This xforms should substantially speed up -O0 builds when it kicks in as well as reducing code size and optimizer heartburn on insane cases. This resolves PR279. llvm-svn: 120692
2010-12-02 15:07:26 +08:00
}
};
} // end namespace CodeGen
} // end namespace clang
#endif