llvm-6502/include/llvm/IR/DIBuilder.h

724 lines
36 KiB
C
Raw Normal View History

//===- DIBuilder.h - Debug Information Builder ------------------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file defines a DIBuilder that is useful for creating debugging
// information entries in LLVM IR form.
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_IR_DIBUILDER_H
#define LLVM_IR_DIBUILDER_H
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/IR/DebugInfo.h"
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
#include "llvm/IR/TrackingMDRef.h"
#include "llvm/IR/ValueHandle.h"
#include "llvm/Support/DataTypes.h"
namespace llvm {
class BasicBlock;
class Instruction;
class Function;
class Module;
class Value;
class Constant;
class LLVMContext;
class MDNode;
class StringRef;
class DIBasicType;
class DICompileUnit;
class DICompositeType;
class DIDerivedType;
class DIDescriptor;
class DIFile;
class DIEnumerator;
class DIType;
class DIGlobalVariable;
class DIImportedEntity;
class DINameSpace;
class DIVariable;
class DISubrange;
class DILexicalBlockFile;
class DILexicalBlock;
class DIScope;
class DISubprogram;
class DITemplateTypeParameter;
class DITemplateValueParameter;
class DIObjCProperty;
class DIBuilder {
Module &M;
LLVMContext &VMContext;
MDNode *TempEnumTypes;
MDNode *TempRetainTypes;
MDNode *TempSubprograms;
MDNode *TempGVs;
MDNode *TempImportedModules;
Function *DeclareFn; // llvm.dbg.declare
Function *ValueFn; // llvm.dbg.value
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
SmallVector<Metadata *, 4> AllEnumTypes;
/// Track the RetainTypes, since they can be updated later on.
SmallVector<TrackingMDNodeRef, 4> AllRetainTypes;
SmallVector<Metadata *, 4> AllSubprograms;
SmallVector<Metadata *, 4> AllGVs;
SmallVector<TrackingMDNodeRef, 4> AllImportedModules;
/// \brief Track nodes that may be unresolved.
SmallVector<TrackingMDNodeRef, 4> UnresolvedNodes;
bool AllowUnresolvedNodes;
/// Each subprogram's preserved local variables.
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
DenseMap<MDNode *, std::vector<TrackingMDNodeRef>> PreservedVariables;
DIBuilder(const DIBuilder &) LLVM_DELETED_FUNCTION;
void operator=(const DIBuilder &) LLVM_DELETED_FUNCTION;
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
/// \brief Create a temporary.
///
/// Create an \a temporary node and track it in \a UnresolvedNodes.
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
void trackIfUnresolved(MDNode *N);
public:
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
/// \brief Construct a builder for a module.
///
/// If \c AllowUnresolved, collect unresolved nodes attached to the module
/// in order to resolve cycles during \a finalize().
explicit DIBuilder(Module &M, bool AllowUnresolved = true);
enum DebugEmissionKind { FullDebug=1, LineTablesOnly };
/// finalize - Construct any deferred debug info descriptors.
void finalize();
/// createCompileUnit - A CompileUnit provides an anchor for all debugging
/// information generated during this instance of compilation.
/// @param Lang Source programming language, eg. dwarf::DW_LANG_C99
/// @param File File name
/// @param Dir Directory
/// @param Producer Identify the producer of debugging information and code.
/// Usually this is a compiler version string.
/// @param isOptimized A boolean flag which indicates whether optimization
/// is ON or not.
/// @param Flags This string lists command line options. This string is
/// directly embedded in debug info output which may be used
/// by a tool analyzing generated debugging information.
/// @param RV This indicates runtime version for languages like
/// Objective-C.
/// @param SplitName The name of the file that we'll split debug info out
/// into.
/// @param Kind The kind of debug information to generate.
/// @param EmitDebugInfo A boolean flag which indicates whether debug
/// information should be written to the final
/// output or not. When this is false, debug
/// information annotations will be present in
/// the IL but they are not written to the final
/// assembly or object file. This supports tracking
/// source location information in the back end
/// without actually changing the output (e.g.,
/// when using optimization remarks).
DICompileUnit createCompileUnit(unsigned Lang, StringRef File,
StringRef Dir, StringRef Producer,
bool isOptimized, StringRef Flags,
unsigned RV,
StringRef SplitName = StringRef(),
DebugEmissionKind Kind = FullDebug,
bool EmitDebugInfo = true);
/// createFile - Create a file descriptor to hold debugging information
/// for a file.
DIFile createFile(StringRef Filename, StringRef Directory);
/// createEnumerator - Create a single enumerator value.
DIEnumerator createEnumerator(StringRef Name, int64_t Val);
/// \brief Create a DWARF unspecified type.
DIBasicType createUnspecifiedType(StringRef Name);
/// \brief Create C++11 nullptr type.
DIBasicType createNullPtrType();
/// createBasicType - Create debugging information entry for a basic
/// type.
/// @param Name Type name.
/// @param SizeInBits Size of the type.
/// @param AlignInBits Type alignment.
/// @param Encoding DWARF encoding code, e.g. dwarf::DW_ATE_float.
DIBasicType createBasicType(StringRef Name, uint64_t SizeInBits,
uint64_t AlignInBits, unsigned Encoding);
/// createQualifiedType - Create debugging information entry for a qualified
/// type, e.g. 'const int'.
/// @param Tag Tag identifing type, e.g. dwarf::TAG_volatile_type
/// @param FromTy Base Type.
DIDerivedType createQualifiedType(unsigned Tag, DIType FromTy);
/// createPointerType - Create debugging information entry for a pointer.
/// @param PointeeTy Type pointed by this pointer.
/// @param SizeInBits Size.
/// @param AlignInBits Alignment. (optional)
/// @param Name Pointer type name. (optional)
DIDerivedType
createPointerType(DIType PointeeTy, uint64_t SizeInBits,
uint64_t AlignInBits = 0, StringRef Name = StringRef());
/// \brief Create debugging information entry for a pointer to member.
/// @param PointeeTy Type pointed to by this pointer.
/// @param SizeInBits Size.
/// @param AlignInBits Alignment. (optional)
/// @param Class Type for which this pointer points to members of.
DIDerivedType createMemberPointerType(DIType PointeeTy, DIType Class,
uint64_t SizeInBits,
uint64_t AlignInBits = 0);
/// createReferenceType - Create debugging information entry for a c++
/// style reference or rvalue reference type.
DIDerivedType createReferenceType(unsigned Tag, DIType RTy);
/// createTypedef - Create debugging information entry for a typedef.
/// @param Ty Original type.
/// @param Name Typedef name.
/// @param File File where this type is defined.
/// @param LineNo Line number.
/// @param Context The surrounding context for the typedef.
DIDerivedType createTypedef(DIType Ty, StringRef Name, DIFile File,
unsigned LineNo, DIDescriptor Context);
/// createFriend - Create debugging information entry for a 'friend'.
DIDerivedType createFriend(DIType Ty, DIType FriendTy);
/// createInheritance - Create debugging information entry to establish
/// inheritance relationship between two types.
/// @param Ty Original type.
/// @param BaseTy Base type. Ty is inherits from base.
/// @param BaseOffset Base offset.
/// @param Flags Flags to describe inheritance attribute,
/// e.g. private
DIDerivedType createInheritance(DIType Ty, DIType BaseTy,
uint64_t BaseOffset, unsigned Flags);
/// createMemberType - Create debugging information entry for a member.
/// @param Scope Member scope.
/// @param Name Member name.
/// @param File File where this member is defined.
/// @param LineNo Line number.
/// @param SizeInBits Member size.
/// @param AlignInBits Member alignment.
/// @param OffsetInBits Member offset.
/// @param Flags Flags to encode member attribute, e.g. private
/// @param Ty Parent type.
DIDerivedType
createMemberType(DIDescriptor Scope, StringRef Name, DIFile File,
unsigned LineNo, uint64_t SizeInBits, uint64_t AlignInBits,
uint64_t OffsetInBits, unsigned Flags, DIType Ty);
/// createStaticMemberType - Create debugging information entry for a
/// C++ static data member.
/// @param Scope Member scope.
/// @param Name Member name.
/// @param File File where this member is declared.
/// @param LineNo Line number.
/// @param Ty Type of the static member.
/// @param Flags Flags to encode member attribute, e.g. private.
/// @param Val Const initializer of the member.
DIDerivedType createStaticMemberType(DIDescriptor Scope, StringRef Name,
DIFile File, unsigned LineNo,
DIType Ty, unsigned Flags,
llvm::Constant *Val);
/// createObjCIVar - Create debugging information entry for Objective-C
/// instance variable.
/// @param Name Member name.
/// @param File File where this member is defined.
/// @param LineNo Line number.
/// @param SizeInBits Member size.
/// @param AlignInBits Member alignment.
/// @param OffsetInBits Member offset.
/// @param Flags Flags to encode member attribute, e.g. private
/// @param Ty Parent type.
/// @param PropertyNode Property associated with this ivar.
DIDerivedType createObjCIVar(StringRef Name, DIFile File,
unsigned LineNo, uint64_t SizeInBits,
uint64_t AlignInBits, uint64_t OffsetInBits,
unsigned Flags, DIType Ty,
MDNode *PropertyNode);
/// createObjCProperty - Create debugging information entry for Objective-C
/// property.
/// @param Name Property name.
/// @param File File where this property is defined.
/// @param LineNumber Line number.
/// @param GetterName Name of the Objective C property getter selector.
/// @param SetterName Name of the Objective C property setter selector.
/// @param PropertyAttributes Objective C property attributes.
/// @param Ty Type.
DIObjCProperty createObjCProperty(StringRef Name,
DIFile File, unsigned LineNumber,
StringRef GetterName,
StringRef SetterName,
unsigned PropertyAttributes,
DIType Ty);
/// createClassType - Create debugging information entry for a class.
/// @param Scope Scope in which this class is defined.
/// @param Name class name.
/// @param File File where this member is defined.
/// @param LineNumber Line number.
/// @param SizeInBits Member size.
/// @param AlignInBits Member alignment.
/// @param OffsetInBits Member offset.
/// @param Flags Flags to encode member attribute, e.g. private
/// @param Elements class members.
/// @param VTableHolder Debug info of the base class that contains vtable
/// for this type. This is used in
/// DW_AT_containing_type. See DWARF documentation
/// for more info.
/// @param TemplateParms Template type parameters.
/// @param UniqueIdentifier A unique identifier for the class.
DICompositeType createClassType(DIDescriptor Scope, StringRef Name,
DIFile File, unsigned LineNumber,
uint64_t SizeInBits, uint64_t AlignInBits,
uint64_t OffsetInBits, unsigned Flags,
DIType DerivedFrom, DIArray Elements,
DIType VTableHolder = DIType(),
MDNode *TemplateParms = nullptr,
StringRef UniqueIdentifier = StringRef());
/// createStructType - Create debugging information entry for a struct.
/// @param Scope Scope in which this struct is defined.
/// @param Name Struct name.
/// @param File File where this member is defined.
/// @param LineNumber Line number.
/// @param SizeInBits Member size.
/// @param AlignInBits Member alignment.
/// @param Flags Flags to encode member attribute, e.g. private
/// @param Elements Struct elements.
/// @param RunTimeLang Optional parameter, Objective-C runtime version.
/// @param UniqueIdentifier A unique identifier for the struct.
DICompositeType createStructType(DIDescriptor Scope, StringRef Name,
DIFile File, unsigned LineNumber,
uint64_t SizeInBits, uint64_t AlignInBits,
unsigned Flags, DIType DerivedFrom,
DIArray Elements, unsigned RunTimeLang = 0,
DIType VTableHolder = DIType(),
StringRef UniqueIdentifier = StringRef());
/// createUnionType - Create debugging information entry for an union.
/// @param Scope Scope in which this union is defined.
/// @param Name Union name.
/// @param File File where this member is defined.
/// @param LineNumber Line number.
/// @param SizeInBits Member size.
/// @param AlignInBits Member alignment.
/// @param Flags Flags to encode member attribute, e.g. private
/// @param Elements Union elements.
/// @param RunTimeLang Optional parameter, Objective-C runtime version.
/// @param UniqueIdentifier A unique identifier for the union.
DICompositeType createUnionType(
DIDescriptor Scope, StringRef Name, DIFile File, unsigned LineNumber,
uint64_t SizeInBits, uint64_t AlignInBits, unsigned Flags,
DIArray Elements, unsigned RunTimeLang = 0,
StringRef UniqueIdentifier = StringRef());
/// createTemplateTypeParameter - Create debugging information for template
/// type parameter.
/// @param Scope Scope in which this type is defined.
/// @param Name Type parameter name.
/// @param Ty Parameter type.
/// @param File File where this type parameter is defined.
/// @param LineNo Line number.
/// @param ColumnNo Column Number.
DITemplateTypeParameter
createTemplateTypeParameter(DIDescriptor Scope, StringRef Name, DIType Ty,
MDNode *File = nullptr, unsigned LineNo = 0,
unsigned ColumnNo = 0);
/// createTemplateValueParameter - Create debugging information for template
/// value parameter.
/// @param Scope Scope in which this type is defined.
/// @param Name Value parameter name.
/// @param Ty Parameter type.
/// @param Val Constant parameter value.
/// @param File File where this type parameter is defined.
/// @param LineNo Line number.
/// @param ColumnNo Column Number.
DITemplateValueParameter
createTemplateValueParameter(DIDescriptor Scope, StringRef Name, DIType Ty,
Constant *Val, MDNode *File = nullptr,
unsigned LineNo = 0, unsigned ColumnNo = 0);
/// \brief Create debugging information for a template template parameter.
/// @param Scope Scope in which this type is defined.
/// @param Name Value parameter name.
/// @param Ty Parameter type.
/// @param Val The fully qualified name of the template.
/// @param File File where this type parameter is defined.
/// @param LineNo Line number.
/// @param ColumnNo Column Number.
DITemplateValueParameter
createTemplateTemplateParameter(DIDescriptor Scope, StringRef Name,
DIType Ty, StringRef Val,
MDNode *File = nullptr, unsigned LineNo = 0,
unsigned ColumnNo = 0);
/// \brief Create debugging information for a template parameter pack.
/// @param Scope Scope in which this type is defined.
/// @param Name Value parameter name.
/// @param Ty Parameter type.
/// @param Val An array of types in the pack.
/// @param File File where this type parameter is defined.
/// @param LineNo Line number.
/// @param ColumnNo Column Number.
DITemplateValueParameter
createTemplateParameterPack(DIDescriptor Scope, StringRef Name,
DIType Ty, DIArray Val, MDNode *File = nullptr,
unsigned LineNo = 0, unsigned ColumnNo = 0);
/// createArrayType - Create debugging information entry for an array.
/// @param Size Array size.
/// @param AlignInBits Alignment.
/// @param Ty Element type.
/// @param Subscripts Subscripts.
DICompositeType createArrayType(uint64_t Size, uint64_t AlignInBits,
DIType Ty, DIArray Subscripts);
/// createVectorType - Create debugging information entry for a vector type.
/// @param Size Array size.
/// @param AlignInBits Alignment.
/// @param Ty Element type.
/// @param Subscripts Subscripts.
DICompositeType createVectorType(uint64_t Size, uint64_t AlignInBits,
DIType Ty, DIArray Subscripts);
/// createEnumerationType - Create debugging information entry for an
/// enumeration.
/// @param Scope Scope in which this enumeration is defined.
/// @param Name Union name.
/// @param File File where this member is defined.
/// @param LineNumber Line number.
/// @param SizeInBits Member size.
/// @param AlignInBits Member alignment.
/// @param Elements Enumeration elements.
/// @param UnderlyingType Underlying type of a C++11/ObjC fixed enum.
/// @param UniqueIdentifier A unique identifier for the enum.
DICompositeType createEnumerationType(DIDescriptor Scope, StringRef Name,
DIFile File, unsigned LineNumber, uint64_t SizeInBits,
uint64_t AlignInBits, DIArray Elements, DIType UnderlyingType,
StringRef UniqueIdentifier = StringRef());
/// createSubroutineType - Create subroutine type.
/// @param File File in which this subroutine is defined.
/// @param ParameterTypes An array of subroutine parameter types. This
/// includes return type at 0th index.
/// @param Flags E.g.: LValueReference.
/// These flags are used to emit dwarf attributes.
DISubroutineType createSubroutineType(DIFile File,
DITypeArray ParameterTypes,
unsigned Flags = 0);
/// createArtificialType - Create a new DIType with "artificial" flag set.
DIType createArtificialType(DIType Ty);
/// createObjectPointerType - Create a new DIType with the "object pointer"
/// flag set.
DIType createObjectPointerType(DIType Ty);
/// \brief Create a permanent forward-declared type.
DICompositeType createForwardDecl(unsigned Tag, StringRef Name,
DIDescriptor Scope, DIFile F,
unsigned Line, unsigned RuntimeLang = 0,
uint64_t SizeInBits = 0,
uint64_t AlignInBits = 0,
StringRef UniqueIdentifier = StringRef());
/// \brief Create a temporary forward-declared type.
DICompositeType createReplaceableForwardDecl(
unsigned Tag, StringRef Name, DIDescriptor Scope, DIFile F,
unsigned Line, unsigned RuntimeLang = 0, uint64_t SizeInBits = 0,
uint64_t AlignInBits = 0, StringRef UniqueIdentifier = StringRef());
/// retainType - Retain DIType in a module even if it is not referenced
/// through debug info anchors.
void retainType(DIType T);
/// createUnspecifiedParameter - Create unspecified parameter type
/// for a subroutine type.
DIBasicType createUnspecifiedParameter();
/// getOrCreateArray - Get a DIArray, create one if required.
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
DIArray getOrCreateArray(ArrayRef<Metadata *> Elements);
/// getOrCreateTypeArray - Get a DITypeArray, create one if required.
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@223802 91177308-0d34-0410-b5e6-96231b3b80d8
2014-12-09 18:38:53 +00:00
DITypeArray getOrCreateTypeArray(ArrayRef<Metadata *> Elements);
/// getOrCreateSubrange - Create a descriptor for a value range. This
/// implicitly uniques the values returned.
DISubrange getOrCreateSubrange(int64_t Lo, int64_t Count);
/// createGlobalVariable - Create a new descriptor for the specified
/// variable.
/// @param Context Variable scope.
/// @param Name Name of the variable.
/// @param LinkageName Mangled name of the variable.
/// @param File File where this variable is defined.
/// @param LineNo Line number.
/// @param Ty Variable Type.
/// @param isLocalToUnit Boolean flag indicate whether this variable is
/// externally visible or not.
/// @param Val llvm::Value of the variable.
/// @param Decl Reference to the corresponding declaration.
DIGlobalVariable createGlobalVariable(DIDescriptor Context, StringRef Name,
StringRef LinkageName, DIFile File,
unsigned LineNo, DITypeRef Ty,
bool isLocalToUnit,
llvm::Constant *Val,
MDNode *Decl = nullptr);
/// createTempGlobalVariableFwdDecl - Identical to createGlobalVariable
/// except that the resulting DbgNode is temporary and meant to be RAUWed.
DIGlobalVariable createTempGlobalVariableFwdDecl(
DIDescriptor Context, StringRef Name, StringRef LinkageName,
DIFile File, unsigned LineNo, DITypeRef Ty, bool isLocalToUnit,
llvm::Constant *Val, MDNode *Decl = nullptr);
/// createLocalVariable - Create a new descriptor for the specified
/// local variable.
/// @param Tag Dwarf TAG. Usually DW_TAG_auto_variable or
/// DW_TAG_arg_variable.
/// @param Scope Variable scope.
/// @param Name Variable name.
/// @param File File where this variable is defined.
/// @param LineNo Line number.
/// @param Ty Variable Type
/// @param AlwaysPreserve Boolean. Set to true if debug info for this
/// variable should be preserved in optimized build.
/// @param Flags Flags, e.g. artificial variable.
/// @param ArgNo If this variable is an argument then this argument's
/// number. 1 indicates 1st argument.
DIVariable createLocalVariable(unsigned Tag, DIDescriptor Scope,
StringRef Name,
DIFile File, unsigned LineNo,
DITypeRef Ty, bool AlwaysPreserve = false,
unsigned Flags = 0,
unsigned ArgNo = 0);
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
/// createExpression - Create a new descriptor for the specified
/// variable which has a complex address expression for its address.
/// @param Addr An array of complex address operations.
DIExpression createExpression(ArrayRef<uint64_t> Addr = None);
DIExpression createExpression(ArrayRef<int64_t> Addr);
/// createBitPieceExpression - Create a descriptor to describe one part
/// of aggregate variable that is fragmented across multiple Values.
///
/// @param OffsetInBits Offset of the piece in bits.
/// @param SizeInBits Size of the piece in bits.
DIExpression createBitPieceExpression(unsigned OffsetInBits,
unsigned SizeInBits);
/// createFunction - Create a new descriptor for the specified subprogram.
/// See comments in DISubprogram for descriptions of these fields.
/// @param Scope Function scope.
/// @param Name Function name.
/// @param LinkageName Mangled function name.
/// @param File File where this variable is defined.
/// @param LineNo Line number.
/// @param Ty Function type.
/// @param isLocalToUnit True if this function is not externally visible.
/// @param isDefinition True if this is a function definition.
/// @param ScopeLine Set to the beginning of the scope this starts
/// @param Flags e.g. is this function prototyped or not.
/// These flags are used to emit dwarf attributes.
/// @param isOptimized True if optimization is ON.
/// @param Fn llvm::Function pointer.
/// @param TParam Function template parameters.
DISubprogram createFunction(DIDescriptor Scope, StringRef Name,
StringRef LinkageName,
DIFile File, unsigned LineNo,
DICompositeType Ty, bool isLocalToUnit,
bool isDefinition,
unsigned ScopeLine,
unsigned Flags = 0,
bool isOptimized = false,
Function *Fn = nullptr,
MDNode *TParam = nullptr,
MDNode *Decl = nullptr);
/// createTempFunctionFwdDecl - Identical to createFunction,
/// except that the resulting DbgNode is meant to be RAUWed.
DISubprogram createTempFunctionFwdDecl(DIDescriptor Scope, StringRef Name,
StringRef LinkageName,
DIFile File, unsigned LineNo,
DICompositeType Ty, bool isLocalToUnit,
bool isDefinition,
unsigned ScopeLine,
unsigned Flags = 0,
bool isOptimized = false,
Function *Fn = nullptr,
MDNode *TParam = nullptr,
MDNode *Decl = nullptr);
/// FIXME: this is added for dragonegg. Once we update dragonegg
/// to call resolve function, this will be removed.
DISubprogram createFunction(DIScopeRef Scope, StringRef Name,
StringRef LinkageName,
DIFile File, unsigned LineNo,
DICompositeType Ty, bool isLocalToUnit,
bool isDefinition,
unsigned ScopeLine,
unsigned Flags = 0,
bool isOptimized = false,
Function *Fn = nullptr,
MDNode *TParam = nullptr,
MDNode *Decl = nullptr);
/// createMethod - Create a new descriptor for the specified C++ method.
/// See comments in DISubprogram for descriptions of these fields.
/// @param Scope Function scope.
/// @param Name Function name.
/// @param LinkageName Mangled function name.
/// @param File File where this variable is defined.
/// @param LineNo Line number.
/// @param Ty Function type.
/// @param isLocalToUnit True if this function is not externally visible..
/// @param isDefinition True if this is a function definition.
/// @param Virtuality Attributes describing virtualness. e.g. pure
/// virtual function.
/// @param VTableIndex Index no of this method in virtual table.
/// @param VTableHolder Type that holds vtable.
/// @param Flags e.g. is this function prototyped or not.
/// This flags are used to emit dwarf attributes.
/// @param isOptimized True if optimization is ON.
/// @param Fn llvm::Function pointer.
/// @param TParam Function template parameters.
DISubprogram createMethod(DIDescriptor Scope, StringRef Name,
StringRef LinkageName,
DIFile File, unsigned LineNo,
DICompositeType Ty, bool isLocalToUnit,
bool isDefinition,
unsigned Virtuality = 0, unsigned VTableIndex = 0,
DIType VTableHolder = DIType(),
unsigned Flags = 0,
bool isOptimized = false,
Function *Fn = nullptr,
MDNode *TParam = nullptr);
/// createNameSpace - This creates new descriptor for a namespace
/// with the specified parent scope.
/// @param Scope Namespace scope
/// @param Name Name of this namespace
/// @param File Source file
/// @param LineNo Line number
DINameSpace createNameSpace(DIDescriptor Scope, StringRef Name,
DIFile File, unsigned LineNo);
/// createLexicalBlockFile - This creates a descriptor for a lexical
/// block with a new file attached. This merely extends the existing
/// lexical block as it crosses a file.
/// @param Scope Lexical block.
/// @param File Source file.
/// @param Discriminator DWARF path discriminator value.
DILexicalBlockFile createLexicalBlockFile(DIDescriptor Scope, DIFile File,
unsigned Discriminator = 0);
/// createLexicalBlock - This creates a descriptor for a lexical block
/// with the specified parent context.
/// @param Scope Parent lexical scope.
/// @param File Source file.
/// @param Line Line number.
/// @param Col Column number.
DILexicalBlock createLexicalBlock(DIDescriptor Scope, DIFile File,
unsigned Line, unsigned Col);
/// \brief Create a descriptor for an imported module.
/// @param Context The scope this module is imported into
/// @param NS The namespace being imported here
/// @param Line Line number
DIImportedEntity createImportedModule(DIScope Context, DINameSpace NS,
unsigned Line);
/// \brief Create a descriptor for an imported module.
/// @param Context The scope this module is imported into
/// @param NS An aliased namespace
/// @param Line Line number
DIImportedEntity createImportedModule(DIScope Context, DIImportedEntity NS,
unsigned Line);
/// \brief Create a descriptor for an imported function.
/// @param Context The scope this module is imported into
/// @param Decl The declaration (or definition) of a function, type, or
/// variable
/// @param Line Line number
DIImportedEntity createImportedDeclaration(DIScope Context, DIDescriptor Decl,
unsigned Line,
StringRef Name = StringRef());
DIImportedEntity createImportedDeclaration(DIScope Context,
DIImportedEntity NS,
unsigned Line,
StringRef Name = StringRef());
/// insertDeclare - Insert a new llvm.dbg.declare intrinsic call.
/// @param Storage llvm::Value of the variable
/// @param VarInfo Variable's debug info descriptor.
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
/// @param Expr A complex location expression.
/// @param InsertAtEnd Location for the new intrinsic.
Instruction *insertDeclare(llvm::Value *Storage, DIVariable VarInfo,
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
DIExpression Expr, BasicBlock *InsertAtEnd);
/// insertDeclare - Insert a new llvm.dbg.declare intrinsic call.
/// @param Storage llvm::Value of the variable
/// @param VarInfo Variable's debug info descriptor.
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
/// @param Expr A complex location expression.
/// @param InsertBefore Location for the new intrinsic.
Instruction *insertDeclare(llvm::Value *Storage, DIVariable VarInfo,
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
DIExpression Expr, Instruction *InsertBefore);
/// insertDbgValueIntrinsic - Insert a new llvm.dbg.value intrinsic call.
/// @param Val llvm::Value of the variable
/// @param Offset Offset
/// @param VarInfo Variable's debug info descriptor.
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
/// @param Expr A complex location expression.
/// @param InsertAtEnd Location for the new intrinsic.
Instruction *insertDbgValueIntrinsic(llvm::Value *Val, uint64_t Offset,
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
DIVariable VarInfo, DIExpression Expr,
BasicBlock *InsertAtEnd);
/// insertDbgValueIntrinsic - Insert a new llvm.dbg.value intrinsic call.
/// @param Val llvm::Value of the variable
/// @param Offset Offset
/// @param VarInfo Variable's debug info descriptor.
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
/// @param Expr A complex location expression.
/// @param InsertBefore Location for the new intrinsic.
Instruction *insertDbgValueIntrinsic(llvm::Value *Val, uint64_t Offset,
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218787 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-01 18:55:02 +00:00
DIVariable VarInfo, DIExpression Expr,
Instruction *InsertBefore);
/// \brief Replace the vtable holder in the given composite type.
///
/// If this creates a self reference, it may orphan some unresolved cycles
/// in the operands of \c T, so \a DIBuilder needs to track that.
void replaceVTableHolder(DICompositeType &T, DICompositeType VTableHolder);
/// \brief Replace arrays on a composite type.
///
/// If \c T is resolved, but the arrays aren't -- which can happen if \c T
/// has a self-reference -- \a DIBuilder needs to track the array to
/// resolve cycles.
void replaceArrays(DICompositeType &T, DIArray Elements,
DIArray TParems = DIArray());
};
} // end namespace llvm
#endif