llvm-project/clang/lib/CodeGen/ABIInfo.h

117 lines
3.8 KiB
C
Raw Normal View History

//===----- ABIInfo.h - ABI information access & encapsulation ---*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_CLANG_LIB_CODEGEN_ABIINFO_H
#define LLVM_CLANG_LIB_CODEGEN_ABIINFO_H
#include "clang/AST/Type.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/Type.h"
namespace llvm {
class Value;
class LLVMContext;
class DataLayout;
}
namespace clang {
class ASTContext;
class TargetInfo;
namespace CodeGen {
Compute and preserve alignment more faithfully in IR-generation. Introduce an Address type to bundle a pointer value with an alignment. Introduce APIs on CGBuilderTy to work with Address values. Change core APIs on CGF/CGM to traffic in Address where appropriate. Require alignments to be non-zero. Update a ton of code to compute and propagate alignment information. As part of this, I've promoted CGBuiltin's EmitPointerWithAlignment helper function to CGF and made use of it in a number of places in the expression emitter. The end result is that we should now be significantly more correct when performing operations on objects that are locally known to be under-aligned. Since alignment is not reliably tracked in the type system, there are inherent limits to this, but at least we are no longer confused by standard operations like derived-to-base conversions and array-to-pointer decay. I've also fixed a large number of bugs where we were applying the complete-object alignment to a pointer instead of the non-virtual alignment, although most of these were hidden by the very conservative approach we took with member alignment. Also, because IRGen now reliably asserts on zero alignments, we should no longer be subject to an absurd but frustrating recurring bug where an incomplete type would report a zero alignment and then we'd naively do a alignmentAtOffset on it and emit code using an alignment equal to the largest power-of-two factor of the offset. We should also now be emitting much more aggressive alignment attributes in the presence of over-alignment. In particular, field access now uses alignmentAtOffset instead of min. Several times in this patch, I had to change the existing code-generation pattern in order to more effectively use the Address APIs. For the most part, this seems to be a strict improvement, like doing pointer arithmetic with GEPs instead of ptrtoint. That said, I've tried very hard to not change semantics, but it is likely that I've failed in a few places, for which I apologize. ABIArgInfo now always carries the assumed alignment of indirect and indirect byval arguments. In order to cut down on what was already a dauntingly large patch, I changed the code to never set align attributes in the IR on non-byval indirect arguments. That is, we still generate code which assumes that indirect arguments have the given alignment, but we don't express this information to the backend except where it's semantically required (i.e. on byvals). This is likely a minor regression for those targets that did provide this information, but it'll be trivial to add it back in a later patch. I partially punted on applying this work to CGBuiltin. Please do not add more uses of the CreateDefaultAligned{Load,Store} APIs; they will be going away eventually. llvm-svn: 246985
2015-09-08 16:05:57 +08:00
class ABIArgInfo;
class Address;
class CGCXXABI;
class CGFunctionInfo;
class CodeGenFunction;
class CodeGenTypes;
}
2010-07-29 07:46:15 +08:00
// FIXME: All of this stuff should be part of the target interface
// somehow. It is currently here because it is not clear how to factor
// the targets to support this, since the Targets currently live in a
// layer below types n'stuff.
/// ABIInfo - Target specific hooks for defining how a type should be
/// passed or returned from functions.
class ABIInfo {
public:
CodeGen::CodeGenTypes &CGT;
protected:
llvm::CallingConv::ID RuntimeCC;
llvm::CallingConv::ID BuiltinCC;
public:
ABIInfo(CodeGen::CodeGenTypes &cgt)
: CGT(cgt),
RuntimeCC(llvm::CallingConv::C),
BuiltinCC(llvm::CallingConv::C) {}
2010-10-19 14:39:39 +08:00
virtual ~ABIInfo();
2010-10-19 14:39:39 +08:00
CodeGen::CGCXXABI &getCXXABI() const;
ASTContext &getContext() const;
llvm::LLVMContext &getVMContext() const;
const llvm::DataLayout &getDataLayout() const;
const TargetInfo &getTarget() const;
/// Return the calling convention to use for system runtime
/// functions.
llvm::CallingConv::ID getRuntimeCC() const {
return RuntimeCC;
}
/// Return the calling convention to use for compiler builtins
llvm::CallingConv::ID getBuiltinCC() const {
return BuiltinCC;
}
virtual void computeInfo(CodeGen::CGFunctionInfo &FI) const = 0;
/// EmitVAArg - Emit the target dependent code to load a value of
/// \arg Ty from the va_list pointed to by \arg VAListAddr.
// FIXME: This is a gaping layering violation if we wanted to drop
// the ABI information any lower than CodeGen. Of course, for
// VAArg handling it has to be at this level; there is no way to
// abstract this out.
Compute and preserve alignment more faithfully in IR-generation. Introduce an Address type to bundle a pointer value with an alignment. Introduce APIs on CGBuilderTy to work with Address values. Change core APIs on CGF/CGM to traffic in Address where appropriate. Require alignments to be non-zero. Update a ton of code to compute and propagate alignment information. As part of this, I've promoted CGBuiltin's EmitPointerWithAlignment helper function to CGF and made use of it in a number of places in the expression emitter. The end result is that we should now be significantly more correct when performing operations on objects that are locally known to be under-aligned. Since alignment is not reliably tracked in the type system, there are inherent limits to this, but at least we are no longer confused by standard operations like derived-to-base conversions and array-to-pointer decay. I've also fixed a large number of bugs where we were applying the complete-object alignment to a pointer instead of the non-virtual alignment, although most of these were hidden by the very conservative approach we took with member alignment. Also, because IRGen now reliably asserts on zero alignments, we should no longer be subject to an absurd but frustrating recurring bug where an incomplete type would report a zero alignment and then we'd naively do a alignmentAtOffset on it and emit code using an alignment equal to the largest power-of-two factor of the offset. We should also now be emitting much more aggressive alignment attributes in the presence of over-alignment. In particular, field access now uses alignmentAtOffset instead of min. Several times in this patch, I had to change the existing code-generation pattern in order to more effectively use the Address APIs. For the most part, this seems to be a strict improvement, like doing pointer arithmetic with GEPs instead of ptrtoint. That said, I've tried very hard to not change semantics, but it is likely that I've failed in a few places, for which I apologize. ABIArgInfo now always carries the assumed alignment of indirect and indirect byval arguments. In order to cut down on what was already a dauntingly large patch, I changed the code to never set align attributes in the IR on non-byval indirect arguments. That is, we still generate code which assumes that indirect arguments have the given alignment, but we don't express this information to the backend except where it's semantically required (i.e. on byvals). This is likely a minor regression for those targets that did provide this information, but it'll be trivial to add it back in a later patch. I partially punted on applying this work to CGBuiltin. Please do not add more uses of the CreateDefaultAligned{Load,Store} APIs; they will be going away eventually. llvm-svn: 246985
2015-09-08 16:05:57 +08:00
virtual CodeGen::Address EmitVAArg(CodeGen::CodeGenFunction &CGF,
CodeGen::Address VAListAddr,
QualType Ty) const = 0;
/// Emit the target dependent code to load a value of
/// \arg Ty from the \c __builtin_ms_va_list pointed to by \arg VAListAddr.
virtual CodeGen::Address EmitMSVAArg(CodeGen::CodeGenFunction &CGF,
CodeGen::Address VAListAddr,
QualType Ty) const;
virtual bool isHomogeneousAggregateBaseType(QualType Ty) const;
virtual bool isHomogeneousAggregateSmallEnough(const Type *Base,
uint64_t Members) const;
virtual bool shouldSignExtUnsignedType(QualType Ty) const;
bool isHomogeneousAggregate(QualType Ty, const Type *&Base,
uint64_t &Members) const;
Compute and preserve alignment more faithfully in IR-generation. Introduce an Address type to bundle a pointer value with an alignment. Introduce APIs on CGBuilderTy to work with Address values. Change core APIs on CGF/CGM to traffic in Address where appropriate. Require alignments to be non-zero. Update a ton of code to compute and propagate alignment information. As part of this, I've promoted CGBuiltin's EmitPointerWithAlignment helper function to CGF and made use of it in a number of places in the expression emitter. The end result is that we should now be significantly more correct when performing operations on objects that are locally known to be under-aligned. Since alignment is not reliably tracked in the type system, there are inherent limits to this, but at least we are no longer confused by standard operations like derived-to-base conversions and array-to-pointer decay. I've also fixed a large number of bugs where we were applying the complete-object alignment to a pointer instead of the non-virtual alignment, although most of these were hidden by the very conservative approach we took with member alignment. Also, because IRGen now reliably asserts on zero alignments, we should no longer be subject to an absurd but frustrating recurring bug where an incomplete type would report a zero alignment and then we'd naively do a alignmentAtOffset on it and emit code using an alignment equal to the largest power-of-two factor of the offset. We should also now be emitting much more aggressive alignment attributes in the presence of over-alignment. In particular, field access now uses alignmentAtOffset instead of min. Several times in this patch, I had to change the existing code-generation pattern in order to more effectively use the Address APIs. For the most part, this seems to be a strict improvement, like doing pointer arithmetic with GEPs instead of ptrtoint. That said, I've tried very hard to not change semantics, but it is likely that I've failed in a few places, for which I apologize. ABIArgInfo now always carries the assumed alignment of indirect and indirect byval arguments. In order to cut down on what was already a dauntingly large patch, I changed the code to never set align attributes in the IR on non-byval indirect arguments. That is, we still generate code which assumes that indirect arguments have the given alignment, but we don't express this information to the backend except where it's semantically required (i.e. on byvals). This is likely a minor regression for those targets that did provide this information, but it'll be trivial to add it back in a later patch. I partially punted on applying this work to CGBuiltin. Please do not add more uses of the CreateDefaultAligned{Load,Store} APIs; they will be going away eventually. llvm-svn: 246985
2015-09-08 16:05:57 +08:00
/// A convenience method to return an indirect ABIArgInfo with an
/// expected alignment equal to the ABI alignment of the given type.
CodeGen::ABIArgInfo
getNaturalAlignIndirect(QualType Ty, bool ByRef = true,
bool Realign = false,
llvm::Type *Padding = nullptr) const;
CodeGen::ABIArgInfo
getNaturalAlignIndirectInReg(QualType Ty, bool Realign = false) const;
};
} // end namespace clang
#endif