llvm-6502/lib/IR/DebugInfo.cpp

1616 lines
47 KiB
C++
Raw Normal View History

//===--- DebugInfo.cpp - Debug Information Helper Classes -----------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements the helper classes used to build and interpret debug
// information in LLVM IR form.
//
//===----------------------------------------------------------------------===//
#include "llvm/IR/DebugInfo.h"
#include "LLVMContextImpl.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallString.h"
#include "llvm/ADT/StringSwitch.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DIBuilder.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/ValueHandle.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/Dwarf.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
using namespace llvm::dwarf;
//===----------------------------------------------------------------------===//
// DIDescriptor
//===----------------------------------------------------------------------===//
unsigned DIDescriptor::getFlag(StringRef Flag) {
return StringSwitch<unsigned>(Flag)
#define HANDLE_DI_FLAG(ID, NAME) .Case("DIFlag" #NAME, Flag##NAME)
#include "llvm/IR/DebugInfoFlags.def"
.Default(0);
}
const char *DIDescriptor::getFlagString(unsigned Flag) {
switch (Flag) {
default:
return "";
#define HANDLE_DI_FLAG(ID, NAME) \
case Flag##NAME: \
return "DIFlag" #NAME;
#include "llvm/IR/DebugInfoFlags.def"
}
}
unsigned DIDescriptor::splitFlags(unsigned Flags,
SmallVectorImpl<unsigned> &SplitFlags) {
// Accessibility flags need to be specially handled, since they're packed
// together.
if (unsigned A = Flags & FlagAccessibility) {
if (A == FlagPrivate)
SplitFlags.push_back(FlagPrivate);
else if (A == FlagProtected)
SplitFlags.push_back(FlagProtected);
else
SplitFlags.push_back(FlagPublic);
Flags &= ~A;
}
#define HANDLE_DI_FLAG(ID, NAME) \
if (unsigned Bit = Flags & ID) { \
SplitFlags.push_back(Bit); \
Flags &= ~Bit; \
}
#include "llvm/IR/DebugInfoFlags.def"
return Flags;
}
bool DIDescriptor::Verify() const {
return DbgNode &&
(DIDerivedType(DbgNode).Verify() ||
DICompositeType(DbgNode).Verify() || DIBasicType(DbgNode).Verify() ||
DIVariable(DbgNode).Verify() || DISubprogram(DbgNode).Verify() ||
DIGlobalVariable(DbgNode).Verify() || DIFile(DbgNode).Verify() ||
DICompileUnit(DbgNode).Verify() || DINameSpace(DbgNode).Verify() ||
DILexicalBlock(DbgNode).Verify() ||
DILexicalBlockFile(DbgNode).Verify() ||
DISubrange(DbgNode).Verify() || DIEnumerator(DbgNode).Verify() ||
DIObjCProperty(DbgNode).Verify() ||
DITemplateTypeParameter(DbgNode).Verify() ||
DITemplateValueParameter(DbgNode).Verify() ||
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
DIImportedEntity(DbgNode).Verify() || DIExpression(DbgNode).Verify());
}
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
static Metadata *getField(const MDNode *DbgNode, unsigned Elt) {
if (!DbgNode || Elt >= DbgNode->getNumOperands())
return nullptr;
return DbgNode->getOperand(Elt);
}
static MDNode *getNodeField(const MDNode *DbgNode, unsigned Elt) {
return dyn_cast_or_null<MDNode>(getField(DbgNode, Elt));
}
static StringRef getStringField(const MDNode *DbgNode, unsigned Elt) {
if (MDString *MDS = dyn_cast_or_null<MDString>(getField(DbgNode, Elt)))
return MDS->getString();
return StringRef();
}
StringRef DIDescriptor::getStringField(unsigned Elt) const {
return ::getStringField(DbgNode, Elt);
}
uint64_t DIDescriptor::getUInt64Field(unsigned Elt) const {
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
if (auto *C = getConstantField(Elt))
if (ConstantInt *CI = dyn_cast<ConstantInt>(C))
return CI->getZExtValue();
return 0;
}
int64_t DIDescriptor::getInt64Field(unsigned Elt) const {
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
if (auto *C = getConstantField(Elt))
if (ConstantInt *CI = dyn_cast<ConstantInt>(C))
return CI->getZExtValue();
return 0;
}
DIDescriptor DIDescriptor::getDescriptorField(unsigned Elt) const {
MDNode *Field = getNodeField(DbgNode, Elt);
return DIDescriptor(Field);
}
GlobalVariable *DIDescriptor::getGlobalVariableField(unsigned Elt) const {
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
return dyn_cast_or_null<GlobalVariable>(getConstantField(Elt));
}
Constant *DIDescriptor::getConstantField(unsigned Elt) const {
if (!DbgNode)
return nullptr;
if (Elt < DbgNode->getNumOperands())
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
if (auto *C =
dyn_cast_or_null<ConstantAsMetadata>(DbgNode->getOperand(Elt)))
return C->getValue();
return nullptr;
}
Function *DIDescriptor::getFunctionField(unsigned Elt) const {
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
return dyn_cast_or_null<Function>(getConstantField(Elt));
}
void DIDescriptor::replaceFunctionField(unsigned Elt, Function *F) {
if (!DbgNode)
return;
if (Elt < DbgNode->getNumOperands()) {
MDNode *Node = const_cast<MDNode *>(DbgNode);
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
Node->replaceOperandWith(Elt, F ? ConstantAsMetadata::get(F) : nullptr);
}
}
static unsigned DIVariableInlinedAtIndex = 4;
MDNode *DIVariable::getInlinedAt() const {
return getNodeField(DbgNode, DIVariableInlinedAtIndex);
}
/// \brief Return the size reported by the variable's type.
unsigned DIVariable::getSizeInBits(const DITypeIdentifierMap &Map) {
DIType Ty = getType().resolve(Map);
// Follow derived types until we reach a type that
// reports back a size.
while (Ty.isDerivedType() && !Ty.getSizeInBits()) {
DIDerivedType DT(&*Ty);
Ty = DT.getTypeDerivedFrom().resolve(Map);
}
assert(Ty.getSizeInBits() && "type with size 0");
return Ty.getSizeInBits();
}
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
uint64_t DIExpression::getElement(unsigned Idx) const {
unsigned I = Idx + 1;
assert(I < getNumHeaderFields() &&
"non-existing complex address element requested");
return getHeaderFieldAs<int64_t>(I);
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
}
bool DIExpression::isBitPiece() const {
unsigned N = getNumElements();
return N >=3 && getElement(N-3) == dwarf::DW_OP_bit_piece;
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
}
uint64_t DIExpression::getBitPieceOffset() const {
assert(isBitPiece() && "not a piece");
return getElement(getNumElements()-2);
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
}
uint64_t DIExpression::getBitPieceSize() const {
assert(isBitPiece() && "not a piece");
return getElement(getNumElements()-1);
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::iterator DIExpression::begin() const {
return DIExpression::iterator(*this);
}
DIExpression::iterator DIExpression::end() const {
return DIExpression::iterator();
}
DIExpression::Operand DIExpression::Operand::getNext() const {
iterator it(I);
return *(++it);
}
//===----------------------------------------------------------------------===//
// Predicates
//===----------------------------------------------------------------------===//
bool DIDescriptor::isSubroutineType() const {
return DbgNode && getTag() == dwarf::DW_TAG_subroutine_type;
}
bool DIDescriptor::isBasicType() const {
if (!DbgNode)
return false;
switch (getTag()) {
case dwarf::DW_TAG_base_type:
case dwarf::DW_TAG_unspecified_type:
return true;
default:
return false;
}
}
bool DIDescriptor::isDerivedType() const {
if (!DbgNode)
return false;
switch (getTag()) {
case dwarf::DW_TAG_typedef:
case dwarf::DW_TAG_pointer_type:
case dwarf::DW_TAG_ptr_to_member_type:
case dwarf::DW_TAG_reference_type:
case dwarf::DW_TAG_rvalue_reference_type:
case dwarf::DW_TAG_const_type:
case dwarf::DW_TAG_volatile_type:
case dwarf::DW_TAG_restrict_type:
case dwarf::DW_TAG_member:
case dwarf::DW_TAG_inheritance:
case dwarf::DW_TAG_friend:
return true;
default:
// CompositeTypes are currently modelled as DerivedTypes.
return isCompositeType();
}
}
bool DIDescriptor::isCompositeType() const {
if (!DbgNode)
return false;
switch (getTag()) {
case dwarf::DW_TAG_array_type:
case dwarf::DW_TAG_structure_type:
case dwarf::DW_TAG_union_type:
case dwarf::DW_TAG_enumeration_type:
case dwarf::DW_TAG_subroutine_type:
case dwarf::DW_TAG_class_type:
return true;
default:
return false;
}
}
bool DIDescriptor::isVariable() const {
if (!DbgNode)
return false;
switch (getTag()) {
case dwarf::DW_TAG_auto_variable:
case dwarf::DW_TAG_arg_variable:
return true;
default:
return false;
}
}
bool DIDescriptor::isType() const {
return isBasicType() || isCompositeType() || isDerivedType();
}
bool DIDescriptor::isSubprogram() const {
return DbgNode && getTag() == dwarf::DW_TAG_subprogram;
}
bool DIDescriptor::isGlobalVariable() const {
return DbgNode && getTag() == dwarf::DW_TAG_variable;
}
bool DIDescriptor::isScope() const {
if (!DbgNode)
return false;
switch (getTag()) {
case dwarf::DW_TAG_compile_unit:
case dwarf::DW_TAG_lexical_block:
case dwarf::DW_TAG_subprogram:
case dwarf::DW_TAG_namespace:
case dwarf::DW_TAG_file_type:
return true;
default:
break;
}
return isType();
}
bool DIDescriptor::isTemplateTypeParameter() const {
return DbgNode && getTag() == dwarf::DW_TAG_template_type_parameter;
}
bool DIDescriptor::isTemplateValueParameter() const {
return DbgNode && (getTag() == dwarf::DW_TAG_template_value_parameter ||
getTag() == dwarf::DW_TAG_GNU_template_template_param ||
getTag() == dwarf::DW_TAG_GNU_template_parameter_pack);
}
bool DIDescriptor::isCompileUnit() const {
return DbgNode && getTag() == dwarf::DW_TAG_compile_unit;
}
bool DIDescriptor::isFile() const {
return DbgNode && getTag() == dwarf::DW_TAG_file_type;
}
bool DIDescriptor::isNameSpace() const {
return DbgNode && getTag() == dwarf::DW_TAG_namespace;
}
bool DIDescriptor::isLexicalBlockFile() const {
return DbgNode && getTag() == dwarf::DW_TAG_lexical_block &&
DbgNode->getNumOperands() == 3 && getNumHeaderFields() == 2;
}
bool DIDescriptor::isLexicalBlock() const {
// FIXME: There are always exactly 4 header fields in DILexicalBlock, but
// something relies on this returning true for DILexicalBlockFile.
return DbgNode && getTag() == dwarf::DW_TAG_lexical_block &&
DbgNode->getNumOperands() == 3 &&
(getNumHeaderFields() == 2 || getNumHeaderFields() == 4);
}
bool DIDescriptor::isSubrange() const {
return DbgNode && getTag() == dwarf::DW_TAG_subrange_type;
}
bool DIDescriptor::isEnumerator() const {
return DbgNode && getTag() == dwarf::DW_TAG_enumerator;
}
bool DIDescriptor::isObjCProperty() const {
return DbgNode && getTag() == dwarf::DW_TAG_APPLE_property;
}
bool DIDescriptor::isImportedEntity() const {
return DbgNode && (getTag() == dwarf::DW_TAG_imported_module ||
getTag() == dwarf::DW_TAG_imported_declaration);
}
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
bool DIDescriptor::isExpression() const {
return DbgNode && (getTag() == dwarf::DW_TAG_expression);
}
//===----------------------------------------------------------------------===//
// Simple Descriptor Constructors and other Methods
//===----------------------------------------------------------------------===//
void DIDescriptor::replaceAllUsesWith(LLVMContext &VMContext, DIDescriptor D) {
assert(DbgNode && "Trying to replace an unverified type!");
// Since we use a TrackingVH for the node, its easy for clients to manufacture
// legitimate situations where they want to replaceAllUsesWith() on something
// which, due to uniquing, has merged with the source. We shield clients from
// this detail by allowing a value to be replaced with replaceAllUsesWith()
// itself.
const MDNode *DN = D;
if (DbgNode == DN) {
SmallVector<Metadata *, 10> Ops(DbgNode->op_begin(), DbgNode->op_end());
DN = MDNode::get(VMContext, Ops);
}
assert(DbgNode->isTemporary() && "Expected temporary node");
auto *Node = const_cast<MDNode *>(DbgNode);
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
Node->replaceAllUsesWith(const_cast<MDNode *>(DN));
MDNode::deleteTemporary(Node);
DbgNode = DN;
}
void DIDescriptor::replaceAllUsesWith(MDNode *D) {
assert(DbgNode && "Trying to replace an unverified type!");
assert(DbgNode != D && "This replacement should always happen");
assert(DbgNode->isTemporary() && "Expected temporary node");
auto *Node = const_cast<MDNode *>(DbgNode);
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
Node->replaceAllUsesWith(D);
MDNode::deleteTemporary(Node);
}
bool DICompileUnit::Verify() const {
if (!isCompileUnit())
return false;
// Don't bother verifying the compilation directory or producer string
// as those could be empty.
if (getFilename().empty())
return false;
return DbgNode->getNumOperands() == 7 && getNumHeaderFields() == 8;
}
bool DIObjCProperty::Verify() const {
if (!isObjCProperty())
return false;
// Don't worry about the rest of the strings for now.
return DbgNode->getNumOperands() == 3 && getNumHeaderFields() == 6;
}
/// \brief Check if a field at position Elt of a MDNode is a MDNode.
static bool fieldIsMDNode(const MDNode *DbgNode, unsigned Elt) {
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
Metadata *Fld = getField(DbgNode, Elt);
return !Fld || isa<MDNode>(Fld);
}
/// \brief Check if a field at position Elt of a MDNode is a MDString.
static bool fieldIsMDString(const MDNode *DbgNode, unsigned Elt) {
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
Metadata *Fld = getField(DbgNode, Elt);
return !Fld || isa<MDString>(Fld);
}
/// \brief Check if a value can be a reference to a type.
static bool isTypeRef(const Metadata *MD) {
if (!MD)
return true;
if (auto *S = dyn_cast<MDString>(MD))
return !S->getString().empty();
if (auto *N = dyn_cast<MDNode>(MD))
return DIType(N).isType();
return false;
}
/// \brief Check if referenced field might be a type.
static bool fieldIsTypeRef(const MDNode *DbgNode, unsigned Elt) {
return isTypeRef(dyn_cast_or_null<Metadata>(getField(DbgNode, Elt)));
}
/// \brief Check if a value can be a ScopeRef.
static bool isScopeRef(const Metadata *MD) {
if (!MD)
return true;
if (auto *S = dyn_cast<MDString>(MD))
return !S->getString().empty();
if (auto *N = dyn_cast<MDNode>(MD))
return DIScope(N).isScope();
return false;
}
/// \brief Check if a field at position Elt of a MDNode can be a ScopeRef.
static bool fieldIsScopeRef(const MDNode *DbgNode, unsigned Elt) {
return isScopeRef(dyn_cast_or_null<Metadata>(getField(DbgNode, Elt)));
}
#ifndef NDEBUG
/// \brief Check if a value can be a DescriptorRef.
static bool isDescriptorRef(const Metadata *MD) {
if (!MD)
return true;
if (auto *S = dyn_cast<MDString>(MD))
return !S->getString().empty();
return isa<MDNode>(MD);
}
#endif
bool DIType::Verify() const {
if (!isType())
return false;
// Make sure Context @ field 2 is MDNode.
if (!fieldIsScopeRef(DbgNode, 2))
return false;
// FIXME: Sink this into the various subclass verifies.
uint16_t Tag = getTag();
if (!isBasicType() && Tag != dwarf::DW_TAG_const_type &&
Tag != dwarf::DW_TAG_volatile_type && Tag != dwarf::DW_TAG_pointer_type &&
Tag != dwarf::DW_TAG_ptr_to_member_type &&
Tag != dwarf::DW_TAG_reference_type &&
Tag != dwarf::DW_TAG_rvalue_reference_type &&
Tag != dwarf::DW_TAG_restrict_type && Tag != dwarf::DW_TAG_array_type &&
Tag != dwarf::DW_TAG_enumeration_type &&
Tag != dwarf::DW_TAG_subroutine_type &&
Tag != dwarf::DW_TAG_inheritance && Tag != dwarf::DW_TAG_friend &&
getFilename().empty())
return false;
// DIType is abstract, it should be a BasicType, a DerivedType or
// a CompositeType.
if (isBasicType())
return DIBasicType(DbgNode).Verify();
else if (isCompositeType())
return DICompositeType(DbgNode).Verify();
else if (isDerivedType())
return DIDerivedType(DbgNode).Verify();
else
return false;
}
bool DIBasicType::Verify() const {
return isBasicType() && DbgNode->getNumOperands() == 3 &&
getNumHeaderFields() == 8;
}
bool DIDerivedType::Verify() const {
// Make sure DerivedFrom @ field 3 is TypeRef.
if (!fieldIsTypeRef(DbgNode, 3))
return false;
if (getTag() == dwarf::DW_TAG_ptr_to_member_type)
// Make sure ClassType @ field 4 is a TypeRef.
if (!fieldIsTypeRef(DbgNode, 4))
return false;
return isDerivedType() && DbgNode->getNumOperands() >= 4 &&
DbgNode->getNumOperands() <= 8 && getNumHeaderFields() >= 7 &&
getNumHeaderFields() <= 8;
}
bool DICompositeType::Verify() const {
if (!isCompositeType())
return false;
// Make sure DerivedFrom @ field 3 and ContainingType @ field 5 are TypeRef.
if (!fieldIsTypeRef(DbgNode, 3))
return false;
if (!fieldIsTypeRef(DbgNode, 5))
return false;
// Make sure the type identifier at field 7 is MDString, it can be null.
if (!fieldIsMDString(DbgNode, 7))
return false;
// A subroutine type can't be both & and &&.
if (isLValueReference() && isRValueReference())
return false;
return DbgNode->getNumOperands() == 8 && getNumHeaderFields() == 8;
}
bool DISubprogram::Verify() const {
if (!isSubprogram())
return false;
// Make sure context @ field 2 is a ScopeRef and type @ field 3 is a MDNode.
if (!fieldIsScopeRef(DbgNode, 2))
return false;
if (!fieldIsMDNode(DbgNode, 3))
return false;
// Containing type @ field 4.
if (!fieldIsTypeRef(DbgNode, 4))
return false;
// A subprogram can't be both & and &&.
if (isLValueReference() && isRValueReference())
return false;
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself. Let me tell you a tale... Originally committed in r211723 after discovering a nasty case of weird scoping due to inlining, this was reverted in r211724 after it fired in ASan/compiler-rt. (minor diversion where I accidentally committed/reverted again in r211871/r211873) After further testing and fixing bugs in ArgumentPromotion (r211872) and Inlining (r212065) it was recommitted in r212085. Reverted in r212089 after the sanitizer buildbots still showed problems. Fixed another bug in ArgumentPromotion (r212128) found by this assertion. Recommitted in r212205, reverted in r212226 after it crashed some more on sanitizer buildbots. Fix clang some more in r212761. Recommitted in r212776, reverted in r212793. ASan failures. Recommitted in r213391, reverted in r213432, trying to reproduce flakey ASan build failure. Fixed bugs in r213805 (ArgPromo + DebugInfo), r213952 (LiveDebugVariables strips dbg_value intrinsics in functions not described by debug info). Recommitted in r214761, reverted in r214999, flakey failure on Windows buildbot. Fixed DeadArgElimination + DebugInfo bug in r219210. Recommitted in r219215, reverted in r219512, failure on ObjC++ atomic properties in the test-suite on Darwin. Fixed ObjC++ atomic properties issue in Clang in r219690. [This commit is provided 'as is' with no hope that this is the last time I commit this change either expressed or implied] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@219702 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-14 18:22:52 +00:00
// If a DISubprogram has an llvm::Function*, then scope chains from all
// instructions within the function should lead to this DISubprogram.
if (auto *F = getFunction()) {
for (auto &BB : *F) {
for (auto &I : BB) {
DebugLoc DL = I.getDebugLoc();
if (DL.isUnknown())
continue;
MDNode *Scope = nullptr;
MDNode *IA = nullptr;
// walk the inlined-at scopes
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
while ((IA = DL.getInlinedAt()))
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself. Let me tell you a tale... Originally committed in r211723 after discovering a nasty case of weird scoping due to inlining, this was reverted in r211724 after it fired in ASan/compiler-rt. (minor diversion where I accidentally committed/reverted again in r211871/r211873) After further testing and fixing bugs in ArgumentPromotion (r211872) and Inlining (r212065) it was recommitted in r212085. Reverted in r212089 after the sanitizer buildbots still showed problems. Fixed another bug in ArgumentPromotion (r212128) found by this assertion. Recommitted in r212205, reverted in r212226 after it crashed some more on sanitizer buildbots. Fix clang some more in r212761. Recommitted in r212776, reverted in r212793. ASan failures. Recommitted in r213391, reverted in r213432, trying to reproduce flakey ASan build failure. Fixed bugs in r213805 (ArgPromo + DebugInfo), r213952 (LiveDebugVariables strips dbg_value intrinsics in functions not described by debug info). Recommitted in r214761, reverted in r214999, flakey failure on Windows buildbot. Fixed DeadArgElimination + DebugInfo bug in r219210. Recommitted in r219215, reverted in r219512, failure on ObjC++ atomic properties in the test-suite on Darwin. Fixed ObjC++ atomic properties issue in Clang in r219690. [This commit is provided 'as is' with no hope that this is the last time I commit this change either expressed or implied] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@219702 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-14 18:22:52 +00:00
DL = DebugLoc::getFromDILocation(IA);
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
DL.getScopeAndInlinedAt(Scope, IA);
if (!Scope)
return false;
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself. Let me tell you a tale... Originally committed in r211723 after discovering a nasty case of weird scoping due to inlining, this was reverted in r211724 after it fired in ASan/compiler-rt. (minor diversion where I accidentally committed/reverted again in r211871/r211873) After further testing and fixing bugs in ArgumentPromotion (r211872) and Inlining (r212065) it was recommitted in r212085. Reverted in r212089 after the sanitizer buildbots still showed problems. Fixed another bug in ArgumentPromotion (r212128) found by this assertion. Recommitted in r212205, reverted in r212226 after it crashed some more on sanitizer buildbots. Fix clang some more in r212761. Recommitted in r212776, reverted in r212793. ASan failures. Recommitted in r213391, reverted in r213432, trying to reproduce flakey ASan build failure. Fixed bugs in r213805 (ArgPromo + DebugInfo), r213952 (LiveDebugVariables strips dbg_value intrinsics in functions not described by debug info). Recommitted in r214761, reverted in r214999, flakey failure on Windows buildbot. Fixed DeadArgElimination + DebugInfo bug in r219210. Recommitted in r219215, reverted in r219512, failure on ObjC++ atomic properties in the test-suite on Darwin. Fixed ObjC++ atomic properties issue in Clang in r219690. [This commit is provided 'as is' with no hope that this is the last time I commit this change either expressed or implied] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@219702 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-14 18:22:52 +00:00
assert(!IA);
while (!DIDescriptor(Scope).isSubprogram()) {
DILexicalBlockFile D(Scope);
Scope = D.isLexicalBlockFile()
? D.getScope()
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
: DebugLoc::getFromDILexicalBlock(Scope).getScope();
if (!Scope)
return false;
DebugInfo: Ensure that all debug location scope chains from instructions within a function, lead to the function itself. Let me tell you a tale... Originally committed in r211723 after discovering a nasty case of weird scoping due to inlining, this was reverted in r211724 after it fired in ASan/compiler-rt. (minor diversion where I accidentally committed/reverted again in r211871/r211873) After further testing and fixing bugs in ArgumentPromotion (r211872) and Inlining (r212065) it was recommitted in r212085. Reverted in r212089 after the sanitizer buildbots still showed problems. Fixed another bug in ArgumentPromotion (r212128) found by this assertion. Recommitted in r212205, reverted in r212226 after it crashed some more on sanitizer buildbots. Fix clang some more in r212761. Recommitted in r212776, reverted in r212793. ASan failures. Recommitted in r213391, reverted in r213432, trying to reproduce flakey ASan build failure. Fixed bugs in r213805 (ArgPromo + DebugInfo), r213952 (LiveDebugVariables strips dbg_value intrinsics in functions not described by debug info). Recommitted in r214761, reverted in r214999, flakey failure on Windows buildbot. Fixed DeadArgElimination + DebugInfo bug in r219210. Recommitted in r219215, reverted in r219512, failure on ObjC++ atomic properties in the test-suite on Darwin. Fixed ObjC++ atomic properties issue in Clang in r219690. [This commit is provided 'as is' with no hope that this is the last time I commit this change either expressed or implied] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@219702 91177308-0d34-0410-b5e6-96231b3b80d8
2014-10-14 18:22:52 +00:00
}
if (!DISubprogram(Scope).describes(F))
return false;
}
}
}
return DbgNode->getNumOperands() == 9 && getNumHeaderFields() == 12;
}
bool DIGlobalVariable::Verify() const {
if (!isGlobalVariable())
return false;
if (getDisplayName().empty())
return false;
// Make sure context @ field 1 is an MDNode.
if (!fieldIsMDNode(DbgNode, 1))
return false;
// Make sure that type @ field 3 is a DITypeRef.
if (!fieldIsTypeRef(DbgNode, 3))
return false;
// Make sure StaticDataMemberDeclaration @ field 5 is MDNode.
if (!fieldIsMDNode(DbgNode, 5))
return false;
return DbgNode->getNumOperands() == 6 && getNumHeaderFields() == 7;
}
bool DIVariable::Verify() const {
if (!isVariable())
return false;
// Make sure context @ field 1 is an MDNode.
if (!fieldIsMDNode(DbgNode, 1))
return false;
// Make sure that type @ field 3 is a DITypeRef.
if (!fieldIsTypeRef(DbgNode, 3))
return false;
// Check the number of header fields, which is common between complex and
// simple variables.
if (getNumHeaderFields() != 4)
return false;
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
// Variable without an inline location.
if (DbgNode->getNumOperands() == 4)
return true;
// Variable with an inline location.
return getInlinedAt() != nullptr && DbgNode->getNumOperands() == 5;
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
}
bool DIExpression::Verify() const {
// Empty DIExpressions may be represented as a nullptr.
if (!DbgNode)
return true;
if (!(isExpression() && DbgNode->getNumOperands() == 1))
return false;
for (auto Op : *this)
switch (Op) {
case DW_OP_bit_piece:
// Must be the last element of the expression.
return std::distance(Op.getBase(), DIHeaderFieldIterator()) == 3;
case DW_OP_plus:
if (std::distance(Op.getBase(), DIHeaderFieldIterator()) < 2)
return false;
break;
case DW_OP_deref:
break;
default:
// Other operators are not yet supported by the backend.
return false;
}
return true;
}
bool DILocation::Verify() const {
return DbgNode && isa<MDLocation>(DbgNode);
}
bool DINameSpace::Verify() const {
if (!isNameSpace())
return false;
return DbgNode->getNumOperands() == 3 && getNumHeaderFields() == 3;
}
MDNode *DIFile::getFileNode() const { return getNodeField(DbgNode, 1); }
bool DIFile::Verify() const {
return isFile() && DbgNode->getNumOperands() == 2;
}
bool DIEnumerator::Verify() const {
return isEnumerator() && DbgNode->getNumOperands() == 1 &&
getNumHeaderFields() == 3;
}
bool DISubrange::Verify() const {
return isSubrange() && DbgNode->getNumOperands() == 1 &&
getNumHeaderFields() == 3;
}
bool DILexicalBlock::Verify() const {
return isLexicalBlock() && DbgNode->getNumOperands() == 3 &&
getNumHeaderFields() == 4;
}
bool DILexicalBlockFile::Verify() const {
return isLexicalBlockFile() && DbgNode->getNumOperands() == 3 &&
getNumHeaderFields() == 2;
}
bool DITemplateTypeParameter::Verify() const {
return isTemplateTypeParameter() && DbgNode->getNumOperands() == 4 &&
getNumHeaderFields() == 4;
}
bool DITemplateValueParameter::Verify() const {
return isTemplateValueParameter() && DbgNode->getNumOperands() == 5 &&
getNumHeaderFields() == 4;
}
bool DIImportedEntity::Verify() const {
return isImportedEntity() && DbgNode->getNumOperands() == 3 &&
getNumHeaderFields() == 3;
}
MDNode *DIDerivedType::getObjCProperty() const {
return getNodeField(DbgNode, 4);
}
MDString *DICompositeType::getIdentifier() const {
return cast_or_null<MDString>(getField(DbgNode, 7));
}
#ifndef NDEBUG
static void VerifySubsetOf(const MDNode *LHS, const MDNode *RHS) {
for (unsigned i = 0; i != LHS->getNumOperands(); ++i) {
// Skip the 'empty' list (that's a single i32 0, rather than truly empty).
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
if (i == 0 && mdconst::hasa<ConstantInt>(LHS->getOperand(i)))
continue;
const MDNode *E = cast<MDNode>(LHS->getOperand(i));
bool found = false;
for (unsigned j = 0; !found && j != RHS->getNumOperands(); ++j)
found = (E == cast<MDNode>(RHS->getOperand(j)));
assert(found && "Losing a member during member list replacement");
}
}
#endif
void DICompositeType::setArraysHelper(MDNode *Elements, MDNode *TParams) {
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
TrackingMDNodeRef N(*this);
if (Elements) {
#ifndef NDEBUG
// Check that the new list of members contains all the old members as well.
if (const MDNode *El = cast_or_null<MDNode>(N->getOperand(4)))
VerifySubsetOf(El, Elements);
#endif
N->replaceOperandWith(4, Elements);
}
if (TParams)
N->replaceOperandWith(6, TParams);
DbgNode = N;
}
DIScopeRef DIScope::getRef() const {
if (!isCompositeType())
return DIScopeRef(*this);
DICompositeType DTy(DbgNode);
if (!DTy.getIdentifier())
return DIScopeRef(*this);
return DIScopeRef(DTy.getIdentifier());
}
void DICompositeType::setContainingType(DICompositeType ContainingType) {
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
TrackingMDNodeRef N(*this);
N->replaceOperandWith(5, ContainingType.getRef());
DbgNode = N;
}
bool DIVariable::isInlinedFnArgument(const Function *CurFn) {
assert(CurFn && "Invalid function");
if (!getContext().isSubprogram())
return false;
// This variable is not inlined function argument if its scope
// does not describe current function.
return !DISubprogram(getContext()).describes(CurFn);
}
bool DISubprogram::describes(const Function *F) {
assert(F && "Invalid function");
if (F == getFunction())
return true;
StringRef Name = getLinkageName();
if (Name.empty())
Name = getName();
if (F->getName() == Name)
return true;
return false;
}
MDNode *DISubprogram::getVariablesNodes() const {
return getNodeField(DbgNode, 8);
}
DIArray DISubprogram::getVariables() const {
return DIArray(getNodeField(DbgNode, 8));
}
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
Metadata *DITemplateValueParameter::getValue() const {
return DbgNode->getOperand(3);
}
DIScopeRef DIScope::getContext() const {
if (isType())
return DIType(DbgNode).getContext();
if (isSubprogram())
return DIScopeRef(DISubprogram(DbgNode).getContext());
if (isLexicalBlock())
return DIScopeRef(DILexicalBlock(DbgNode).getContext());
if (isLexicalBlockFile())
return DIScopeRef(DILexicalBlockFile(DbgNode).getContext());
if (isNameSpace())
return DIScopeRef(DINameSpace(DbgNode).getContext());
assert((isFile() || isCompileUnit()) && "Unhandled type of scope.");
return DIScopeRef(nullptr);
}
StringRef DIScope::getName() const {
if (isType())
return DIType(DbgNode).getName();
if (isSubprogram())
return DISubprogram(DbgNode).getName();
if (isNameSpace())
return DINameSpace(DbgNode).getName();
assert((isLexicalBlock() || isLexicalBlockFile() || isFile() ||
isCompileUnit()) &&
"Unhandled type of scope.");
return StringRef();
}
StringRef DIScope::getFilename() const {
if (!DbgNode)
return StringRef();
return ::getStringField(getNodeField(DbgNode, 1), 0);
}
StringRef DIScope::getDirectory() const {
if (!DbgNode)
return StringRef();
return ::getStringField(getNodeField(DbgNode, 1), 1);
}
DIArray DICompileUnit::getEnumTypes() const {
if (!DbgNode || DbgNode->getNumOperands() < 7)
return DIArray();
return DIArray(getNodeField(DbgNode, 2));
}
DIArray DICompileUnit::getRetainedTypes() const {
if (!DbgNode || DbgNode->getNumOperands() < 7)
return DIArray();
return DIArray(getNodeField(DbgNode, 3));
}
DIArray DICompileUnit::getSubprograms() const {
if (!DbgNode || DbgNode->getNumOperands() < 7)
return DIArray();
return DIArray(getNodeField(DbgNode, 4));
}
DIArray DICompileUnit::getGlobalVariables() const {
if (!DbgNode || DbgNode->getNumOperands() < 7)
return DIArray();
return DIArray(getNodeField(DbgNode, 5));
}
DIArray DICompileUnit::getImportedEntities() const {
if (!DbgNode || DbgNode->getNumOperands() < 7)
return DIArray();
return DIArray(getNodeField(DbgNode, 6));
}
void DICompileUnit::replaceSubprograms(DIArray Subprograms) {
assert(Verify() && "Expected compile unit");
if (Subprograms == getSubprograms())
return;
const_cast<MDNode *>(DbgNode)->replaceOperandWith(4, Subprograms);
}
void DICompileUnit::replaceGlobalVariables(DIArray GlobalVariables) {
assert(Verify() && "Expected compile unit");
if (GlobalVariables == getGlobalVariables())
return;
const_cast<MDNode *>(DbgNode)->replaceOperandWith(5, GlobalVariables);
}
DILocation DILocation::copyWithNewScope(LLVMContext &Ctx,
DILexicalBlockFile NewScope) {
assert(Verify());
assert(NewScope && "Expected valid scope");
const auto *Old = cast<MDLocation>(DbgNode);
return DILocation(MDLocation::get(Ctx, Old->getLine(), Old->getColumn(),
NewScope, Old->getInlinedAt()));
}
unsigned DILocation::computeNewDiscriminator(LLVMContext &Ctx) {
std::pair<const char *, unsigned> Key(getFilename().data(), getLineNumber());
return ++Ctx.pImpl->DiscriminatorTable[Key];
}
DIVariable llvm::createInlinedVariable(MDNode *DV, MDNode *InlinedScope,
LLVMContext &VMContext) {
assert(DIVariable(DV).Verify() && "Expected a DIVariable");
if (!InlinedScope)
return cleanseInlinedVariable(DV, VMContext);
// Insert inlined scope.
SmallVector<Metadata *, 8> Elts(DV->op_begin(),
DV->op_begin() + DIVariableInlinedAtIndex);
Elts.push_back(InlinedScope);
DIVariable Inlined(MDNode::get(VMContext, Elts));
assert(Inlined.Verify() && "Expected to create a DIVariable");
return Inlined;
}
DIVariable llvm::cleanseInlinedVariable(MDNode *DV, LLVMContext &VMContext) {
assert(DIVariable(DV).Verify() && "Expected a DIVariable");
if (!DIVariable(DV).getInlinedAt())
return DIVariable(DV);
// Remove inlined scope.
SmallVector<Metadata *, 8> Elts(DV->op_begin(),
DV->op_begin() + DIVariableInlinedAtIndex);
DIVariable Cleansed(MDNode::get(VMContext, Elts));
assert(Cleansed.Verify() && "Expected to create a DIVariable");
return Cleansed;
}
DISubprogram llvm::getDISubprogram(const MDNode *Scope) {
DIDescriptor D(Scope);
if (D.isSubprogram())
return DISubprogram(Scope);
if (D.isLexicalBlockFile())
return getDISubprogram(DILexicalBlockFile(Scope).getContext());
if (D.isLexicalBlock())
return getDISubprogram(DILexicalBlock(Scope).getContext());
return DISubprogram();
}
DISubprogram llvm::getDISubprogram(const Function *F) {
// We look for the first instr that has a debug annotation leading back to F.
for (auto &BB : *F) {
auto Inst = std::find_if(BB.begin(), BB.end(), [](const Instruction &Inst) {
return !Inst.getDebugLoc().isUnknown();
});
if (Inst == BB.end())
continue;
DebugLoc DLoc = Inst->getDebugLoc();
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
const MDNode *Scope = DLoc.getScopeNode();
DISubprogram Subprogram = getDISubprogram(Scope);
return Subprogram.describes(F) ? Subprogram : DISubprogram();
}
return DISubprogram();
}
DICompositeType llvm::getDICompositeType(DIType T) {
if (T.isCompositeType())
return DICompositeType(T);
if (T.isDerivedType()) {
// This function is currently used by dragonegg and dragonegg does
// not generate identifier for types, so using an empty map to resolve
// DerivedFrom should be fine.
DITypeIdentifierMap EmptyMap;
return getDICompositeType(
DIDerivedType(T).getTypeDerivedFrom().resolve(EmptyMap));
}
return DICompositeType();
}
DITypeIdentifierMap
llvm::generateDITypeIdentifierMap(const NamedMDNode *CU_Nodes) {
DITypeIdentifierMap Map;
for (unsigned CUi = 0, CUe = CU_Nodes->getNumOperands(); CUi != CUe; ++CUi) {
DICompileUnit CU(CU_Nodes->getOperand(CUi));
DIArray Retain = CU.getRetainedTypes();
for (unsigned Ti = 0, Te = Retain.getNumElements(); Ti != Te; ++Ti) {
if (!Retain.getElement(Ti).isCompositeType())
continue;
DICompositeType Ty(Retain.getElement(Ti));
if (MDString *TypeId = Ty.getIdentifier()) {
// Definition has priority over declaration.
// Try to insert (TypeId, Ty) to Map.
std::pair<DITypeIdentifierMap::iterator, bool> P =
Map.insert(std::make_pair(TypeId, Ty));
// If TypeId already exists in Map and this is a definition, replace
// whatever we had (declaration or definition) with the definition.
if (!P.second && !Ty.isForwardDecl())
P.first->second = Ty;
}
}
}
return Map;
}
//===----------------------------------------------------------------------===//
// DebugInfoFinder implementations.
//===----------------------------------------------------------------------===//
void DebugInfoFinder::reset() {
CUs.clear();
SPs.clear();
GVs.clear();
TYs.clear();
Scopes.clear();
NodesSeen.clear();
TypeIdentifierMap.clear();
TypeMapInitialized = false;
}
void DebugInfoFinder::InitializeTypeMap(const Module &M) {
if (!TypeMapInitialized)
if (NamedMDNode *CU_Nodes = M.getNamedMetadata("llvm.dbg.cu")) {
TypeIdentifierMap = generateDITypeIdentifierMap(CU_Nodes);
TypeMapInitialized = true;
}
}
void DebugInfoFinder::processModule(const Module &M) {
InitializeTypeMap(M);
if (NamedMDNode *CU_Nodes = M.getNamedMetadata("llvm.dbg.cu")) {
for (unsigned i = 0, e = CU_Nodes->getNumOperands(); i != e; ++i) {
DICompileUnit CU(CU_Nodes->getOperand(i));
addCompileUnit(CU);
DIArray GVs = CU.getGlobalVariables();
for (unsigned i = 0, e = GVs.getNumElements(); i != e; ++i) {
DIGlobalVariable DIG(GVs.getElement(i));
if (addGlobalVariable(DIG)) {
processScope(DIG.getContext());
processType(DIG.getType().resolve(TypeIdentifierMap));
}
}
DIArray SPs = CU.getSubprograms();
for (unsigned i = 0, e = SPs.getNumElements(); i != e; ++i)
processSubprogram(DISubprogram(SPs.getElement(i)));
DIArray EnumTypes = CU.getEnumTypes();
for (unsigned i = 0, e = EnumTypes.getNumElements(); i != e; ++i)
processType(DIType(EnumTypes.getElement(i)));
DIArray RetainedTypes = CU.getRetainedTypes();
for (unsigned i = 0, e = RetainedTypes.getNumElements(); i != e; ++i)
processType(DIType(RetainedTypes.getElement(i)));
DIArray Imports = CU.getImportedEntities();
for (unsigned i = 0, e = Imports.getNumElements(); i != e; ++i) {
DIImportedEntity Import = DIImportedEntity(Imports.getElement(i));
DIDescriptor Entity = Import.getEntity().resolve(TypeIdentifierMap);
if (Entity.isType())
processType(DIType(Entity));
else if (Entity.isSubprogram())
processSubprogram(DISubprogram(Entity));
else if (Entity.isNameSpace())
processScope(DINameSpace(Entity).getContext());
}
}
}
}
void DebugInfoFinder::processLocation(const Module &M, DILocation Loc) {
if (!Loc)
return;
InitializeTypeMap(M);
processScope(Loc.getScope());
processLocation(M, Loc.getOrigLocation());
}
void DebugInfoFinder::processType(DIType DT) {
if (!addType(DT))
return;
processScope(DT.getContext().resolve(TypeIdentifierMap));
if (DT.isCompositeType()) {
DICompositeType DCT(DT);
processType(DCT.getTypeDerivedFrom().resolve(TypeIdentifierMap));
if (DT.isSubroutineType()) {
DITypeArray DTA = DISubroutineType(DT).getTypeArray();
for (unsigned i = 0, e = DTA.getNumElements(); i != e; ++i)
processType(DTA.getElement(i).resolve(TypeIdentifierMap));
return;
}
DIArray DA = DCT.getElements();
for (unsigned i = 0, e = DA.getNumElements(); i != e; ++i) {
DIDescriptor D = DA.getElement(i);
if (D.isType())
processType(DIType(D));
else if (D.isSubprogram())
processSubprogram(DISubprogram(D));
}
} else if (DT.isDerivedType()) {
DIDerivedType DDT(DT);
processType(DDT.getTypeDerivedFrom().resolve(TypeIdentifierMap));
}
}
void DebugInfoFinder::processScope(DIScope Scope) {
if (Scope.isType()) {
DIType Ty(Scope);
processType(Ty);
return;
}
if (Scope.isCompileUnit()) {
addCompileUnit(DICompileUnit(Scope));
return;
}
if (Scope.isSubprogram()) {
processSubprogram(DISubprogram(Scope));
return;
}
if (!addScope(Scope))
return;
if (Scope.isLexicalBlock()) {
DILexicalBlock LB(Scope);
processScope(LB.getContext());
} else if (Scope.isLexicalBlockFile()) {
DILexicalBlockFile LBF = DILexicalBlockFile(Scope);
processScope(LBF.getScope());
} else if (Scope.isNameSpace()) {
DINameSpace NS(Scope);
processScope(NS.getContext());
}
}
void DebugInfoFinder::processSubprogram(DISubprogram SP) {
if (!addSubprogram(SP))
return;
processScope(SP.getContext().resolve(TypeIdentifierMap));
processType(SP.getType());
DIArray TParams = SP.getTemplateParams();
for (unsigned I = 0, E = TParams.getNumElements(); I != E; ++I) {
DIDescriptor Element = TParams.getElement(I);
if (Element.isTemplateTypeParameter()) {
DITemplateTypeParameter TType(Element);
processType(TType.getType().resolve(TypeIdentifierMap));
} else if (Element.isTemplateValueParameter()) {
DITemplateValueParameter TVal(Element);
processType(TVal.getType().resolve(TypeIdentifierMap));
}
}
}
void DebugInfoFinder::processDeclare(const Module &M,
const DbgDeclareInst *DDI) {
MDNode *N = dyn_cast<MDNode>(DDI->getVariable());
if (!N)
return;
InitializeTypeMap(M);
DIDescriptor DV(N);
if (!DV.isVariable())
return;
if (!NodesSeen.insert(DV).second)
return;
processScope(DIVariable(N).getContext());
processType(DIVariable(N).getType().resolve(TypeIdentifierMap));
}
void DebugInfoFinder::processValue(const Module &M, const DbgValueInst *DVI) {
MDNode *N = dyn_cast<MDNode>(DVI->getVariable());
if (!N)
return;
InitializeTypeMap(M);
DIDescriptor DV(N);
if (!DV.isVariable())
return;
if (!NodesSeen.insert(DV).second)
return;
processScope(DIVariable(N).getContext());
processType(DIVariable(N).getType().resolve(TypeIdentifierMap));
}
bool DebugInfoFinder::addType(DIType DT) {
if (!DT)
return false;
if (!NodesSeen.insert(DT).second)
return false;
TYs.push_back(DT);
return true;
}
bool DebugInfoFinder::addCompileUnit(DICompileUnit CU) {
if (!CU)
return false;
if (!NodesSeen.insert(CU).second)
return false;
CUs.push_back(CU);
return true;
}
bool DebugInfoFinder::addGlobalVariable(DIGlobalVariable DIG) {
if (!DIG)
return false;
if (!NodesSeen.insert(DIG).second)
return false;
GVs.push_back(DIG);
return true;
}
bool DebugInfoFinder::addSubprogram(DISubprogram SP) {
if (!SP)
return false;
if (!NodesSeen.insert(SP).second)
return false;
SPs.push_back(SP);
return true;
}
bool DebugInfoFinder::addScope(DIScope Scope) {
if (!Scope)
return false;
// FIXME: Ocaml binding generates a scope with no content, we treat it
// as null for now.
if (Scope->getNumOperands() == 0)
return false;
if (!NodesSeen.insert(Scope).second)
return false;
Scopes.push_back(Scope);
return true;
}
//===----------------------------------------------------------------------===//
// DIDescriptor: dump routines for all descriptors.
//===----------------------------------------------------------------------===//
void DIDescriptor::dump() const {
print(dbgs());
dbgs() << '\n';
}
void DIDescriptor::print(raw_ostream &OS) const {
if (!DbgNode)
return;
if (const char *Tag = dwarf::TagString(getTag()))
OS << "[ " << Tag << " ]";
if (this->isSubrange()) {
DISubrange(DbgNode).printInternal(OS);
} else if (this->isCompileUnit()) {
DICompileUnit(DbgNode).printInternal(OS);
} else if (this->isFile()) {
DIFile(DbgNode).printInternal(OS);
} else if (this->isEnumerator()) {
DIEnumerator(DbgNode).printInternal(OS);
} else if (this->isBasicType()) {
DIType(DbgNode).printInternal(OS);
} else if (this->isDerivedType()) {
DIDerivedType(DbgNode).printInternal(OS);
} else if (this->isCompositeType()) {
DICompositeType(DbgNode).printInternal(OS);
} else if (this->isSubprogram()) {
DISubprogram(DbgNode).printInternal(OS);
} else if (this->isGlobalVariable()) {
DIGlobalVariable(DbgNode).printInternal(OS);
} else if (this->isVariable()) {
DIVariable(DbgNode).printInternal(OS);
} else if (this->isObjCProperty()) {
DIObjCProperty(DbgNode).printInternal(OS);
} else if (this->isNameSpace()) {
DINameSpace(DbgNode).printInternal(OS);
} else if (this->isScope()) {
DIScope(DbgNode).printInternal(OS);
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
} else if (this->isExpression()) {
DIExpression(DbgNode).printInternal(OS);
}
}
void DISubrange::printInternal(raw_ostream &OS) const {
int64_t Count = getCount();
if (Count != -1)
OS << " [" << getLo() << ", " << Count - 1 << ']';
else
OS << " [unbounded]";
}
void DIScope::printInternal(raw_ostream &OS) const {
OS << " [" << getDirectory() << "/" << getFilename() << ']';
}
void DICompileUnit::printInternal(raw_ostream &OS) const {
DIScope::printInternal(OS);
OS << " [";
unsigned Lang = getLanguage();
if (const char *LangStr = dwarf::LanguageString(Lang))
OS << LangStr;
else
(OS << "lang 0x").write_hex(Lang);
OS << ']';
}
void DIEnumerator::printInternal(raw_ostream &OS) const {
OS << " [" << getName() << " :: " << getEnumValue() << ']';
}
void DIType::printInternal(raw_ostream &OS) const {
if (!DbgNode)
return;
StringRef Res = getName();
if (!Res.empty())
OS << " [" << Res << "]";
// TODO: Print context?
OS << " [line " << getLineNumber() << ", size " << getSizeInBits()
<< ", align " << getAlignInBits() << ", offset " << getOffsetInBits();
if (isBasicType())
if (const char *Enc =
dwarf::AttributeEncodingString(DIBasicType(DbgNode).getEncoding()))
OS << ", enc " << Enc;
OS << "]";
if (isPrivate())
OS << " [private]";
else if (isProtected())
OS << " [protected]";
else if (isPublic())
OS << " [public]";
if (isArtificial())
OS << " [artificial]";
if (isForwardDecl())
OS << " [decl]";
else if (getTag() == dwarf::DW_TAG_structure_type ||
getTag() == dwarf::DW_TAG_union_type ||
getTag() == dwarf::DW_TAG_enumeration_type ||
getTag() == dwarf::DW_TAG_class_type)
OS << " [def]";
if (isVector())
OS << " [vector]";
if (isStaticMember())
OS << " [static]";
if (isLValueReference())
OS << " [reference]";
if (isRValueReference())
OS << " [rvalue reference]";
}
void DIDerivedType::printInternal(raw_ostream &OS) const {
DIType::printInternal(OS);
OS << " [from " << getTypeDerivedFrom().getName() << ']';
}
void DICompositeType::printInternal(raw_ostream &OS) const {
DIType::printInternal(OS);
DIArray A = getElements();
OS << " [" << A.getNumElements() << " elements]";
}
void DINameSpace::printInternal(raw_ostream &OS) const {
StringRef Name = getName();
if (!Name.empty())
OS << " [" << Name << ']';
OS << " [line " << getLineNumber() << ']';
}
void DISubprogram::printInternal(raw_ostream &OS) const {
// TODO : Print context
OS << " [line " << getLineNumber() << ']';
if (isLocalToUnit())
OS << " [local]";
if (isDefinition())
OS << " [def]";
if (getScopeLineNumber() != getLineNumber())
OS << " [scope " << getScopeLineNumber() << "]";
if (isPrivate())
OS << " [private]";
else if (isProtected())
OS << " [protected]";
else if (isPublic())
OS << " [public]";
if (isLValueReference())
OS << " [reference]";
if (isRValueReference())
OS << " [rvalue reference]";
StringRef Res = getName();
if (!Res.empty())
OS << " [" << Res << ']';
}
void DIGlobalVariable::printInternal(raw_ostream &OS) const {
StringRef Res = getName();
if (!Res.empty())
OS << " [" << Res << ']';
OS << " [line " << getLineNumber() << ']';
// TODO : Print context
if (isLocalToUnit())
OS << " [local]";
if (isDefinition())
OS << " [def]";
}
void DIVariable::printInternal(raw_ostream &OS) const {
StringRef Res = getName();
if (!Res.empty())
OS << " [" << Res << ']';
OS << " [line " << getLineNumber() << ']';
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
}
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
void DIExpression::printInternal(raw_ostream &OS) const {
for (auto Op : *this) {
OS << " [" << OperationEncodingString(Op);
switch (Op) {
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
case DW_OP_plus: {
OS << " " << Op.getArg(1);
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
break;
}
case DW_OP_bit_piece: {
OS << " offset=" << Op.getArg(1) << ", size=" << Op.getArg(2);
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
break;
}
case DW_OP_deref:
// No arguments.
break;
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
default:
llvm_unreachable("unhandled operation");
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
}
OS << "]";
}
}
void DIObjCProperty::printInternal(raw_ostream &OS) const {
StringRef Name = getObjCPropertyName();
if (!Name.empty())
OS << " [" << Name << ']';
OS << " [line " << getLineNumber() << ", properties " << getUnsignedField(6)
<< ']';
}
static void printDebugLoc(DebugLoc DL, raw_ostream &CommentOS,
const LLVMContext &Ctx) {
if (!DL.isUnknown()) { // Print source line info.
DIScope Scope(DL.getScope(Ctx));
assert(Scope.isScope() && "Scope of a DebugLoc should be a DIScope.");
// Omit the directory, because it's likely to be long and uninteresting.
CommentOS << Scope.getFilename();
CommentOS << ':' << DL.getLine();
if (DL.getCol() != 0)
CommentOS << ':' << DL.getCol();
DebugLoc InlinedAtDL = DebugLoc::getFromDILocation(DL.getInlinedAt(Ctx));
if (!InlinedAtDL.isUnknown()) {
CommentOS << " @[ ";
printDebugLoc(InlinedAtDL, CommentOS, Ctx);
CommentOS << " ]";
}
}
}
void DIVariable::printExtendedName(raw_ostream &OS) const {
const LLVMContext &Ctx = DbgNode->getContext();
StringRef Res = getName();
if (!Res.empty())
OS << Res << "," << getLineNumber();
if (MDNode *InlinedAt = getInlinedAt()) {
DebugLoc InlinedAtDL = DebugLoc::getFromDILocation(InlinedAt);
if (!InlinedAtDL.isUnknown()) {
OS << " @[";
printDebugLoc(InlinedAtDL, OS, Ctx);
OS << "]";
}
}
}
template <> DIRef<DIDescriptor>::DIRef(const Metadata *V) : Val(V) {
assert(isDescriptorRef(V) &&
"DIDescriptorRef should be a MDString or MDNode");
}
template <> DIRef<DIScope>::DIRef(const Metadata *V) : Val(V) {
assert(isScopeRef(V) && "DIScopeRef should be a MDString or MDNode");
}
template <> DIRef<DIType>::DIRef(const Metadata *V) : Val(V) {
assert(isTypeRef(V) && "DITypeRef should be a MDString or MDNode");
}
template <>
DIDescriptorRef DIDescriptor::getFieldAs<DIDescriptorRef>(unsigned Elt) const {
return DIDescriptorRef(cast_or_null<Metadata>(getField(DbgNode, Elt)));
}
template <>
DIScopeRef DIDescriptor::getFieldAs<DIScopeRef>(unsigned Elt) const {
return DIScopeRef(cast_or_null<Metadata>(getField(DbgNode, Elt)));
}
template <> DITypeRef DIDescriptor::getFieldAs<DITypeRef>(unsigned Elt) const {
return DITypeRef(cast_or_null<Metadata>(getField(DbgNode, Elt)));
}
bool llvm::StripDebugInfo(Module &M) {
bool Changed = false;
// Remove all of the calls to the debugger intrinsics, and remove them from
// the module.
if (Function *Declare = M.getFunction("llvm.dbg.declare")) {
while (!Declare->use_empty()) {
[C++11] Add range based accessors for the Use-Def chain of a Value. This requires a number of steps. 1) Move value_use_iterator into the Value class as an implementation detail 2) Change it to actually be a *Use* iterator rather than a *User* iterator. 3) Add an adaptor which is a User iterator that always looks through the Use to the User. 4) Wrap these in Value::use_iterator and Value::user_iterator typedefs. 5) Add the range adaptors as Value::uses() and Value::users(). 6) Update *all* of the callers to correctly distinguish between whether they wanted a use_iterator (and to explicitly dig out the User when needed), or a user_iterator which makes the Use itself totally opaque. Because #6 requires churning essentially everything that walked the Use-Def chains, I went ahead and added all of the range adaptors and switched them to range-based loops where appropriate. Also because the renaming requires at least churning every line of code, it didn't make any sense to split these up into multiple commits -- all of which would touch all of the same lies of code. The result is still not quite optimal. The Value::use_iterator is a nice regular iterator, but Value::user_iterator is an iterator over User*s rather than over the User objects themselves. As a consequence, it fits a bit awkwardly into the range-based world and it has the weird extra-dereferencing 'operator->' that so many of our iterators have. I think this could be fixed by providing something which transforms a range of T&s into a range of T*s, but that *can* be separated into another patch, and it isn't yet 100% clear whether this is the right move. However, this change gets us most of the benefit and cleans up a substantial amount of code around Use and User. =] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203364 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-09 03:16:01 +00:00
CallInst *CI = cast<CallInst>(Declare->user_back());
CI->eraseFromParent();
}
Declare->eraseFromParent();
Changed = true;
}
if (Function *DbgVal = M.getFunction("llvm.dbg.value")) {
while (!DbgVal->use_empty()) {
[C++11] Add range based accessors for the Use-Def chain of a Value. This requires a number of steps. 1) Move value_use_iterator into the Value class as an implementation detail 2) Change it to actually be a *Use* iterator rather than a *User* iterator. 3) Add an adaptor which is a User iterator that always looks through the Use to the User. 4) Wrap these in Value::use_iterator and Value::user_iterator typedefs. 5) Add the range adaptors as Value::uses() and Value::users(). 6) Update *all* of the callers to correctly distinguish between whether they wanted a use_iterator (and to explicitly dig out the User when needed), or a user_iterator which makes the Use itself totally opaque. Because #6 requires churning essentially everything that walked the Use-Def chains, I went ahead and added all of the range adaptors and switched them to range-based loops where appropriate. Also because the renaming requires at least churning every line of code, it didn't make any sense to split these up into multiple commits -- all of which would touch all of the same lies of code. The result is still not quite optimal. The Value::use_iterator is a nice regular iterator, but Value::user_iterator is an iterator over User*s rather than over the User objects themselves. As a consequence, it fits a bit awkwardly into the range-based world and it has the weird extra-dereferencing 'operator->' that so many of our iterators have. I think this could be fixed by providing something which transforms a range of T&s into a range of T*s, but that *can* be separated into another patch, and it isn't yet 100% clear whether this is the right move. However, this change gets us most of the benefit and cleans up a substantial amount of code around Use and User. =] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203364 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-09 03:16:01 +00:00
CallInst *CI = cast<CallInst>(DbgVal->user_back());
CI->eraseFromParent();
}
DbgVal->eraseFromParent();
Changed = true;
}
for (Module::named_metadata_iterator NMI = M.named_metadata_begin(),
NME = M.named_metadata_end(); NMI != NME;) {
NamedMDNode *NMD = NMI;
++NMI;
if (NMD->getName().startswith("llvm.dbg.")) {
NMD->eraseFromParent();
Changed = true;
}
}
for (Module::iterator MI = M.begin(), ME = M.end(); MI != ME; ++MI)
for (Function::iterator FI = MI->begin(), FE = MI->end(); FI != FE;
++FI)
for (BasicBlock::iterator BI = FI->begin(), BE = FI->end(); BI != BE;
++BI) {
if (!BI->getDebugLoc().isUnknown()) {
Changed = true;
BI->setDebugLoc(DebugLoc());
}
}
return Changed;
}
unsigned llvm::getDebugMetadataVersionFromModule(const Module &M) {
if (auto *Val = mdconst::dyn_extract_or_null<ConstantInt>(
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
M.getModuleFlag("Debug Info Version")))
return Val->getZExtValue();
return 0;
}
llvm::DenseMap<const llvm::Function *, llvm::DISubprogram>
llvm::makeSubprogramMap(const Module &M) {
DenseMap<const Function *, DISubprogram> R;
NamedMDNode *CU_Nodes = M.getNamedMetadata("llvm.dbg.cu");
if (!CU_Nodes)
return R;
for (MDNode *N : CU_Nodes->operands()) {
DICompileUnit CUNode(N);
DIArray SPs = CUNode.getSubprograms();
for (unsigned i = 0, e = SPs.getNumElements(); i != e; ++i) {
DISubprogram SP(SPs.getElement(i));
if (Function *F = SP.getFunction())
R.insert(std::make_pair(F, SP));
}
}
return R;
}