llvm-project/llvm/lib/IR/DIBuilder.cpp

1011 lines
42 KiB
C++
Raw Normal View History

//===--- DIBuilder.cpp - Debug Information Builder ------------------------===//
//
// 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 DIBuilder.
//
//===----------------------------------------------------------------------===//
#include "llvm/IR/DIBuilder.h"
#include "llvm/IR/IRBuilder.h"
#include "LLVMContextImpl.h"
#include "llvm/ADT/Optional.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/BinaryFormat/Dwarf.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DebugInfo.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
using namespace llvm;
using namespace llvm::dwarf;
cl::opt<bool>
UseDbgAddr("use-dbg-addr",
llvm::cl::desc("Use llvm.dbg.addr for all local variables"),
cl::init(false), cl::Hidden);
DIBuilder::DIBuilder(Module &m, bool AllowUnresolvedNodes, DICompileUnit *CU)
: M(m), VMContext(M.getContext()), CUNode(CU),
DeclareFn(nullptr), ValueFn(nullptr), LabelFn(nullptr),
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
AllowUnresolvedNodes(AllowUnresolvedNodes) {}
void DIBuilder::trackIfUnresolved(MDNode *N) {
if (!N)
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
return;
if (N->isResolved())
return;
assert(AllowUnresolvedNodes && "Cannot handle unresolved nodes");
UnresolvedNodes.emplace_back(N);
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
}
void DIBuilder::finalizeSubprogram(DISubprogram *SP) {
MDTuple *Temp = SP->getRetainedNodes().get();
if (!Temp || !Temp->isTemporary())
return;
SmallVector<Metadata *, 16> RetainedNodes;
auto PV = PreservedVariables.find(SP);
if (PV != PreservedVariables.end())
RetainedNodes.append(PV->second.begin(), PV->second.end());
auto PL = PreservedLabels.find(SP);
if (PL != PreservedLabels.end())
RetainedNodes.append(PL->second.begin(), PL->second.end());
DINodeArray Node = getOrCreateArray(RetainedNodes);
TempMDTuple(Temp)->replaceAllUsesWith(Node.get());
}
void DIBuilder::finalize() {
if (!CUNode) {
assert(!AllowUnresolvedNodes &&
"creating type nodes without a CU is not supported");
return;
}
CUNode->replaceEnumTypes(MDTuple::get(VMContext, AllEnumTypes));
SmallVector<Metadata *, 16> RetainValues;
// Declarations and definitions of the same type may be retained. Some
// clients RAUW these pairs, leaving duplicates in the retained types
// list. Use a set to remove the duplicates while we transform the
// TrackingVHs back into Values.
SmallPtrSet<Metadata *, 16> RetainSet;
for (unsigned I = 0, E = AllRetainTypes.size(); I < E; I++)
if (RetainSet.insert(AllRetainTypes[I]).second)
RetainValues.push_back(AllRetainTypes[I]);
if (!RetainValues.empty())
CUNode->replaceRetainedTypes(MDTuple::get(VMContext, RetainValues));
DISubprogramArray SPs = MDTuple::get(VMContext, AllSubprograms);
for (auto *SP : SPs)
finalizeSubprogram(SP);
for (auto *N : RetainValues)
if (auto *SP = dyn_cast<DISubprogram>(N))
finalizeSubprogram(SP);
if (!AllGVs.empty())
CUNode->replaceGlobalVariables(MDTuple::get(VMContext, AllGVs));
if (!AllImportedModules.empty())
CUNode->replaceImportedEntities(MDTuple::get(
VMContext, SmallVector<Metadata *, 16>(AllImportedModules.begin(),
AllImportedModules.end())));
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
for (const auto &I : AllMacrosPerParent) {
// DIMacroNode's with nullptr parent are DICompileUnit direct children.
if (!I.first) {
CUNode->replaceMacros(MDTuple::get(VMContext, I.second.getArrayRef()));
continue;
}
// Otherwise, it must be a temporary DIMacroFile that need to be resolved.
auto *TMF = cast<DIMacroFile>(I.first);
auto *MF = DIMacroFile::get(VMContext, dwarf::DW_MACINFO_start_file,
TMF->getLine(), TMF->getFile(),
getOrCreateMacroArray(I.second.getArrayRef()));
replaceTemporary(llvm::TempDIMacroNode(TMF), MF);
}
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
// Now that all temp nodes have been replaced or deleted, resolve remaining
// cycles.
for (const auto &N : UnresolvedNodes)
if (N && !N->isResolved())
N->resolveCycles();
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
UnresolvedNodes.clear();
// Can't handle unresolved nodes anymore.
AllowUnresolvedNodes = false;
}
/// If N is compile unit return NULL otherwise return N.
static DIScope *getNonCompileUnitScope(DIScope *N) {
if (!N || isa<DICompileUnit>(N))
return nullptr;
return cast<DIScope>(N);
}
DICompileUnit *DIBuilder::createCompileUnit(
unsigned Lang, DIFile *File, StringRef Producer, bool isOptimized,
StringRef Flags, unsigned RunTimeVer, StringRef SplitName,
DICompileUnit::DebugEmissionKind Kind, uint64_t DWOId,
bool SplitDebugInlining, bool DebugInfoForProfiling, bool GnuPubnames) {
assert(((Lang <= dwarf::DW_LANG_Fortran08 && Lang >= dwarf::DW_LANG_C89) ||
(Lang <= dwarf::DW_LANG_hi_user && Lang >= dwarf::DW_LANG_lo_user)) &&
"Invalid Language tag");
assert(!CUNode && "Can only make one compile unit per DIBuilder instance");
CUNode = DICompileUnit::getDistinct(
VMContext, Lang, File, Producer, isOptimized, Flags, RunTimeVer,
SplitName, Kind, nullptr, nullptr, nullptr, nullptr, nullptr, DWOId,
SplitDebugInlining, DebugInfoForProfiling, GnuPubnames);
// Create a named metadata so that it is easier to find cu in a module.
NamedMDNode *NMD = M.getOrInsertNamedMetadata("llvm.dbg.cu");
NMD->addOperand(CUNode);
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
trackIfUnresolved(CUNode);
return CUNode;
}
static DIImportedEntity *
createImportedModule(LLVMContext &C, dwarf::Tag Tag, DIScope *Context,
Metadata *NS, DIFile *File, unsigned Line, StringRef Name,
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
SmallVectorImpl<TrackingMDNodeRef> &AllImportedModules) {
if (Line)
assert(File && "Source location has line number but no file");
unsigned EntitiesCount = C.pImpl->DIImportedEntitys.size();
auto *M =
DIImportedEntity::get(C, Tag, Context, DINodeRef(NS), File, Line, Name);
if (EntitiesCount < C.pImpl->DIImportedEntitys.size())
// A new Imported Entity was just added to the context.
// Add it to the Imported Modules list.
AllImportedModules.emplace_back(M);
return M;
}
DIImportedEntity *DIBuilder::createImportedModule(DIScope *Context,
DINamespace *NS, DIFile *File,
unsigned Line) {
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_module,
Context, NS, File, Line, StringRef(),
AllImportedModules);
}
DIImportedEntity *DIBuilder::createImportedModule(DIScope *Context,
DIImportedEntity *NS,
DIFile *File, unsigned Line) {
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_module,
Context, NS, File, Line, StringRef(),
AllImportedModules);
}
DIImportedEntity *DIBuilder::createImportedModule(DIScope *Context, DIModule *M,
DIFile *File, unsigned Line) {
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_module,
Context, M, File, Line, StringRef(),
AllImportedModules);
}
DIImportedEntity *DIBuilder::createImportedDeclaration(DIScope *Context,
DINode *Decl,
DIFile *File,
unsigned Line,
StringRef Name) {
// Make sure to use the unique identifier based metadata reference for
// types that have one.
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_declaration,
Context, Decl, File, Line, Name,
AllImportedModules);
}
DIFile *DIBuilder::createFile(StringRef Filename, StringRef Directory,
Optional<DIFile::ChecksumInfo<StringRef>> CS,
Optional<StringRef> Source) {
return DIFile::get(VMContext, Filename, Directory, CS, Source);
}
DIMacro *DIBuilder::createMacro(DIMacroFile *Parent, unsigned LineNumber,
unsigned MacroType, StringRef Name,
StringRef Value) {
assert(!Name.empty() && "Unable to create macro without name");
assert((MacroType == dwarf::DW_MACINFO_undef ||
MacroType == dwarf::DW_MACINFO_define) &&
"Unexpected macro type");
auto *M = DIMacro::get(VMContext, MacroType, LineNumber, Name, Value);
AllMacrosPerParent[Parent].insert(M);
return M;
}
DIMacroFile *DIBuilder::createTempMacroFile(DIMacroFile *Parent,
unsigned LineNumber, DIFile *File) {
auto *MF = DIMacroFile::getTemporary(VMContext, dwarf::DW_MACINFO_start_file,
LineNumber, File, DIMacroNodeArray())
.release();
AllMacrosPerParent[Parent].insert(MF);
// Add the new temporary DIMacroFile to the macro per parent map as a parent.
// This is needed to assure DIMacroFile with no children to have an entry in
// the map. Otherwise, it will not be resolved in DIBuilder::finalize().
AllMacrosPerParent.insert({MF, {}});
return MF;
}
DIEnumerator *DIBuilder::createEnumerator(StringRef Name, int64_t Val,
bool IsUnsigned) {
assert(!Name.empty() && "Unable to create enumerator without name");
return DIEnumerator::get(VMContext, Val, IsUnsigned, Name);
}
DIBasicType *DIBuilder::createUnspecifiedType(StringRef Name) {
assert(!Name.empty() && "Unable to create type without name");
return DIBasicType::get(VMContext, dwarf::DW_TAG_unspecified_type, Name);
}
DIBasicType *DIBuilder::createNullPtrType() {
return createUnspecifiedType("decltype(nullptr)");
}
DIBasicType *DIBuilder::createBasicType(StringRef Name, uint64_t SizeInBits,
unsigned Encoding) {
assert(!Name.empty() && "Unable to create type without name");
return DIBasicType::get(VMContext, dwarf::DW_TAG_base_type, Name, SizeInBits,
0, Encoding);
}
DIDerivedType *DIBuilder::createQualifiedType(unsigned Tag, DIType *FromTy) {
return DIDerivedType::get(VMContext, Tag, "", nullptr, 0, nullptr, FromTy, 0,
0, 0, None, DINode::FlagZero);
}
DIDerivedType *DIBuilder::createPointerType(
DIType *PointeeTy,
uint64_t SizeInBits,
uint32_t AlignInBits,
Optional<unsigned> DWARFAddressSpace,
StringRef Name) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// FIXME: Why is there a name here?
return DIDerivedType::get(VMContext, dwarf::DW_TAG_pointer_type, Name,
nullptr, 0, nullptr, PointeeTy, SizeInBits,
AlignInBits, 0, DWARFAddressSpace,
DINode::FlagZero);
}
DIDerivedType *DIBuilder::createMemberPointerType(DIType *PointeeTy,
DIType *Base,
uint64_t SizeInBits,
uint32_t AlignInBits,
DINode::DIFlags Flags) {
return DIDerivedType::get(VMContext, dwarf::DW_TAG_ptr_to_member_type, "",
nullptr, 0, nullptr, PointeeTy, SizeInBits,
AlignInBits, 0, None, Flags, Base);
}
DIDerivedType *DIBuilder::createReferenceType(
unsigned Tag, DIType *RTy,
uint64_t SizeInBits,
uint32_t AlignInBits,
Optional<unsigned> DWARFAddressSpace) {
assert(RTy && "Unable to create reference type");
return DIDerivedType::get(VMContext, Tag, "", nullptr, 0, nullptr, RTy,
SizeInBits, AlignInBits, 0, DWARFAddressSpace,
DINode::FlagZero);
}
DIDerivedType *DIBuilder::createTypedef(DIType *Ty, StringRef Name,
DIFile *File, unsigned LineNo,
DIScope *Context) {
return DIDerivedType::get(VMContext, dwarf::DW_TAG_typedef, Name, File,
LineNo, getNonCompileUnitScope(Context), Ty, 0, 0,
0, None, DINode::FlagZero);
}
DIDerivedType *DIBuilder::createFriend(DIType *Ty, DIType *FriendTy) {
assert(Ty && "Invalid type!");
assert(FriendTy && "Invalid friend type!");
return DIDerivedType::get(VMContext, dwarf::DW_TAG_friend, "", nullptr, 0, Ty,
FriendTy, 0, 0, 0, None, DINode::FlagZero);
}
DIDerivedType *DIBuilder::createInheritance(DIType *Ty, DIType *BaseTy,
uint64_t BaseOffset,
uint32_t VBPtrOffset,
DINode::DIFlags Flags) {
assert(Ty && "Unable to create inheritance");
Metadata *ExtraData = ConstantAsMetadata::get(
ConstantInt::get(IntegerType::get(VMContext, 32), VBPtrOffset));
return DIDerivedType::get(VMContext, dwarf::DW_TAG_inheritance, "", nullptr,
0, Ty, BaseTy, 0, 0, BaseOffset, None,
Flags, ExtraData);
}
DIDerivedType *DIBuilder::createMemberType(DIScope *Scope, StringRef Name,
DIFile *File, unsigned LineNumber,
uint64_t SizeInBits,
uint32_t AlignInBits,
uint64_t OffsetInBits,
DINode::DIFlags Flags, DIType *Ty) {
return DIDerivedType::get(VMContext, dwarf::DW_TAG_member, Name, File,
LineNumber, getNonCompileUnitScope(Scope), Ty,
SizeInBits, AlignInBits, OffsetInBits, None, Flags);
}
static ConstantAsMetadata *getConstantOrNull(Constant *C) {
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
if (C)
return ConstantAsMetadata::get(C);
return nullptr;
}
DIDerivedType *DIBuilder::createVariantMemberType(DIScope *Scope, StringRef Name,
DIFile *File, unsigned LineNumber,
uint64_t SizeInBits,
uint32_t AlignInBits,
uint64_t OffsetInBits,
Constant *Discriminant,
DINode::DIFlags Flags, DIType *Ty) {
return DIDerivedType::get(VMContext, dwarf::DW_TAG_member, Name, File,
LineNumber, getNonCompileUnitScope(Scope), Ty,
SizeInBits, AlignInBits, OffsetInBits, None, Flags,
getConstantOrNull(Discriminant));
}
DIDerivedType *DIBuilder::createBitFieldMemberType(
DIScope *Scope, StringRef Name, DIFile *File, unsigned LineNumber,
uint64_t SizeInBits, uint64_t OffsetInBits, uint64_t StorageOffsetInBits,
DINode::DIFlags Flags, DIType *Ty) {
Flags |= DINode::FlagBitField;
return DIDerivedType::get(
VMContext, dwarf::DW_TAG_member, Name, File, LineNumber,
getNonCompileUnitScope(Scope), Ty, SizeInBits, /* AlignInBits */ 0,
OffsetInBits, None, Flags,
ConstantAsMetadata::get(ConstantInt::get(IntegerType::get(VMContext, 64),
StorageOffsetInBits)));
}
DIDerivedType *
DIBuilder::createStaticMemberType(DIScope *Scope, StringRef Name, DIFile *File,
unsigned LineNumber, DIType *Ty,
DINode::DIFlags Flags, llvm::Constant *Val,
uint32_t AlignInBits) {
Flags |= DINode::FlagStaticMember;
return DIDerivedType::get(VMContext, dwarf::DW_TAG_member, Name, File,
LineNumber, getNonCompileUnitScope(Scope), Ty, 0,
AlignInBits, 0, None, Flags,
getConstantOrNull(Val));
}
DIDerivedType *
DIBuilder::createObjCIVar(StringRef Name, DIFile *File, unsigned LineNumber,
uint64_t SizeInBits, uint32_t AlignInBits,
uint64_t OffsetInBits, DINode::DIFlags Flags,
DIType *Ty, MDNode *PropertyNode) {
return DIDerivedType::get(VMContext, dwarf::DW_TAG_member, Name, File,
LineNumber, getNonCompileUnitScope(File), Ty,
SizeInBits, AlignInBits, OffsetInBits, None, Flags,
PropertyNode);
}
DIObjCProperty *
DIBuilder::createObjCProperty(StringRef Name, DIFile *File, unsigned LineNumber,
StringRef GetterName, StringRef SetterName,
unsigned PropertyAttributes, DIType *Ty) {
return DIObjCProperty::get(VMContext, Name, File, LineNumber, GetterName,
SetterName, PropertyAttributes, Ty);
}
DITemplateTypeParameter *
DIBuilder::createTemplateTypeParameter(DIScope *Context, StringRef Name,
DIType *Ty) {
assert((!Context || isa<DICompileUnit>(Context)) && "Expected compile unit");
return DITemplateTypeParameter::get(VMContext, Name, Ty);
}
static DITemplateValueParameter *
createTemplateValueParameterHelper(LLVMContext &VMContext, unsigned Tag,
DIScope *Context, StringRef Name, DIType *Ty,
Metadata *MD) {
assert((!Context || isa<DICompileUnit>(Context)) && "Expected compile unit");
return DITemplateValueParameter::get(VMContext, Tag, Name, Ty, MD);
}
DITemplateValueParameter *
DIBuilder::createTemplateValueParameter(DIScope *Context, StringRef Name,
DIType *Ty, Constant *Val) {
return createTemplateValueParameterHelper(
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
VMContext, dwarf::DW_TAG_template_value_parameter, Context, Name, Ty,
getConstantOrNull(Val));
}
DITemplateValueParameter *
DIBuilder::createTemplateTemplateParameter(DIScope *Context, StringRef Name,
DIType *Ty, StringRef Val) {
return createTemplateValueParameterHelper(
VMContext, dwarf::DW_TAG_GNU_template_template_param, Context, Name, Ty,
MDString::get(VMContext, Val));
}
DITemplateValueParameter *
DIBuilder::createTemplateParameterPack(DIScope *Context, StringRef Name,
DIType *Ty, DINodeArray Val) {
return createTemplateValueParameterHelper(
VMContext, dwarf::DW_TAG_GNU_template_parameter_pack, Context, Name, Ty,
Val.get());
}
DICompositeType *DIBuilder::createClassType(
DIScope *Context, StringRef Name, DIFile *File, unsigned LineNumber,
uint64_t SizeInBits, uint32_t AlignInBits, uint64_t OffsetInBits,
DINode::DIFlags Flags, DIType *DerivedFrom, DINodeArray Elements,
DIType *VTableHolder, MDNode *TemplateParams, StringRef UniqueIdentifier) {
assert((!Context || isa<DIScope>(Context)) &&
"createClassType should be called with a valid Context");
auto *R = DICompositeType::get(
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
VMContext, dwarf::DW_TAG_structure_type, Name, File, LineNumber,
getNonCompileUnitScope(Context), DerivedFrom, SizeInBits, AlignInBits,
OffsetInBits, Flags, Elements, 0, VTableHolder,
cast_or_null<MDTuple>(TemplateParams), UniqueIdentifier);
trackIfUnresolved(R);
return R;
}
DICompositeType *DIBuilder::createStructType(
DIScope *Context, StringRef Name, DIFile *File, unsigned LineNumber,
uint64_t SizeInBits, uint32_t AlignInBits, DINode::DIFlags Flags,
DIType *DerivedFrom, DINodeArray Elements, unsigned RunTimeLang,
DIType *VTableHolder, StringRef UniqueIdentifier) {
auto *R = DICompositeType::get(
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
VMContext, dwarf::DW_TAG_structure_type, Name, File, LineNumber,
getNonCompileUnitScope(Context), DerivedFrom, SizeInBits, AlignInBits, 0,
Flags, Elements, RunTimeLang, VTableHolder, nullptr, UniqueIdentifier);
trackIfUnresolved(R);
return R;
}
DICompositeType *DIBuilder::createUnionType(
DIScope *Scope, StringRef Name, DIFile *File, unsigned LineNumber,
uint64_t SizeInBits, uint32_t AlignInBits, DINode::DIFlags Flags,
DINodeArray Elements, unsigned RunTimeLang, StringRef UniqueIdentifier) {
auto *R = DICompositeType::get(
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
VMContext, dwarf::DW_TAG_union_type, Name, File, LineNumber,
getNonCompileUnitScope(Scope), nullptr, SizeInBits, AlignInBits, 0, Flags,
Elements, RunTimeLang, nullptr, nullptr, UniqueIdentifier);
trackIfUnresolved(R);
return R;
}
DICompositeType *DIBuilder::createVariantPart(
DIScope *Scope, StringRef Name, DIFile *File, unsigned LineNumber,
uint64_t SizeInBits, uint32_t AlignInBits, DINode::DIFlags Flags,
DIDerivedType *Discriminator, DINodeArray Elements, StringRef UniqueIdentifier) {
auto *R = DICompositeType::get(
VMContext, dwarf::DW_TAG_variant_part, Name, File, LineNumber,
getNonCompileUnitScope(Scope), nullptr, SizeInBits, AlignInBits, 0, Flags,
Elements, 0, nullptr, nullptr, UniqueIdentifier, Discriminator);
trackIfUnresolved(R);
return R;
}
DISubroutineType *DIBuilder::createSubroutineType(DITypeRefArray ParameterTypes,
DINode::DIFlags Flags,
unsigned CC) {
return DISubroutineType::get(VMContext, Flags, CC, ParameterTypes);
}
DICompositeType *DIBuilder::createEnumerationType(
DIScope *Scope, StringRef Name, DIFile *File, unsigned LineNumber,
uint64_t SizeInBits, uint32_t AlignInBits, DINodeArray Elements,
DIType *UnderlyingType, StringRef UniqueIdentifier, bool IsFixed) {
auto *CTy = DICompositeType::get(
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
VMContext, dwarf::DW_TAG_enumeration_type, Name, File, LineNumber,
getNonCompileUnitScope(Scope), UnderlyingType, SizeInBits, AlignInBits, 0,
IsFixed ? DINode::FlagFixedEnum : DINode::FlagZero, Elements, 0, nullptr,
nullptr, UniqueIdentifier);
AllEnumTypes.push_back(CTy);
trackIfUnresolved(CTy);
return CTy;
}
DICompositeType *DIBuilder::createArrayType(uint64_t Size,
uint32_t AlignInBits, DIType *Ty,
DINodeArray Subscripts) {
auto *R = DICompositeType::get(VMContext, dwarf::DW_TAG_array_type, "",
nullptr, 0, nullptr, Ty, Size, AlignInBits, 0,
DINode::FlagZero, Subscripts, 0, nullptr);
trackIfUnresolved(R);
return R;
}
DICompositeType *DIBuilder::createVectorType(uint64_t Size,
uint32_t AlignInBits, DIType *Ty,
DINodeArray Subscripts) {
auto *R = DICompositeType::get(VMContext, dwarf::DW_TAG_array_type, "",
nullptr, 0, nullptr, Ty, Size, AlignInBits, 0,
DINode::FlagVector, Subscripts, 0, nullptr);
trackIfUnresolved(R);
return R;
}
static DIType *createTypeWithFlags(LLVMContext &Context, DIType *Ty,
DINode::DIFlags FlagsToSet) {
auto NewTy = Ty->clone();
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
NewTy->setFlags(NewTy->getFlags() | FlagsToSet);
return MDNode::replaceWithUniqued(std::move(NewTy));
}
DIType *DIBuilder::createArtificialType(DIType *Ty) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// FIXME: Restrict this to the nodes where it's valid.
if (Ty->isArtificial())
return Ty;
return createTypeWithFlags(VMContext, Ty, DINode::FlagArtificial);
}
DIType *DIBuilder::createObjectPointerType(DIType *Ty) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// FIXME: Restrict this to the nodes where it's valid.
if (Ty->isObjectPointer())
return Ty;
DINode::DIFlags Flags = DINode::FlagObjectPointer | DINode::FlagArtificial;
return createTypeWithFlags(VMContext, Ty, Flags);
}
void DIBuilder::retainType(DIScope *T) {
assert(T && "Expected non-null type");
assert((isa<DIType>(T) || (isa<DISubprogram>(T) &&
cast<DISubprogram>(T)->isDefinition() == false)) &&
"Expected type or subprogram declaration");
AllRetainTypes.emplace_back(T);
}
DIBasicType *DIBuilder::createUnspecifiedParameter() { return nullptr; }
DICompositeType *
DIBuilder::createForwardDecl(unsigned Tag, StringRef Name, DIScope *Scope,
DIFile *F, unsigned Line, unsigned RuntimeLang,
uint64_t SizeInBits, uint32_t AlignInBits,
StringRef UniqueIdentifier) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// FIXME: Define in terms of createReplaceableForwardDecl() by calling
// replaceWithUniqued().
auto *RetTy = DICompositeType::get(
VMContext, Tag, Name, F, Line, getNonCompileUnitScope(Scope), nullptr,
SizeInBits, AlignInBits, 0, DINode::FlagFwdDecl, nullptr, RuntimeLang,
nullptr, nullptr, UniqueIdentifier);
trackIfUnresolved(RetTy);
return RetTy;
}
DICompositeType *DIBuilder::createReplaceableCompositeType(
unsigned Tag, StringRef Name, DIScope *Scope, DIFile *F, unsigned Line,
unsigned RuntimeLang, uint64_t SizeInBits, uint32_t AlignInBits,
DINode::DIFlags Flags, StringRef UniqueIdentifier) {
auto *RetTy =
DICompositeType::getTemporary(
VMContext, Tag, Name, F, Line, getNonCompileUnitScope(Scope), nullptr,
SizeInBits, AlignInBits, 0, Flags, nullptr, RuntimeLang, nullptr,
nullptr, UniqueIdentifier)
.release();
trackIfUnresolved(RetTy);
2013-07-03 02:37:35 +08:00
return RetTy;
}
DINodeArray DIBuilder::getOrCreateArray(ArrayRef<Metadata *> Elements) {
return MDTuple::get(VMContext, Elements);
}
DIMacroNodeArray
DIBuilder::getOrCreateMacroArray(ArrayRef<Metadata *> Elements) {
return MDTuple::get(VMContext, Elements);
}
DITypeRefArray DIBuilder::getOrCreateTypeArray(ArrayRef<Metadata *> Elements) {
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
SmallVector<llvm::Metadata *, 16> Elts;
for (unsigned i = 0, e = Elements.size(); i != e; ++i) {
if (Elements[i] && isa<MDNode>(Elements[i]))
Elts.push_back(cast<DIType>(Elements[i]));
else
Elts.push_back(Elements[i]);
}
return DITypeRefArray(MDNode::get(VMContext, Elts));
}
DISubrange *DIBuilder::getOrCreateSubrange(int64_t Lo, int64_t Count) {
return DISubrange::get(VMContext, Count, Lo);
}
DISubrange *DIBuilder::getOrCreateSubrange(int64_t Lo, Metadata *CountNode) {
return DISubrange::get(VMContext, CountNode, Lo);
}
static void checkGlobalVariableScope(DIScope *Context) {
#ifndef NDEBUG
if (auto *CT =
dyn_cast_or_null<DICompositeType>(getNonCompileUnitScope(Context)))
assert(CT->getIdentifier().empty() &&
"Context of a global variable should not be a type with identifier");
#endif
}
DIGlobalVariableExpression *DIBuilder::createGlobalVariableExpression(
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *F,
unsigned LineNumber, DIType *Ty, bool isLocalToUnit, DIExpression *Expr,
MDNode *Decl, uint32_t AlignInBits) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
checkGlobalVariableScope(Context);
auto *GV = DIGlobalVariable::getDistinct(
VMContext, cast_or_null<DIScope>(Context), Name, LinkageName, F,
LineNumber, Ty, isLocalToUnit, true, cast_or_null<DIDerivedType>(Decl),
AlignInBits);
if (!Expr)
Expr = createExpression();
auto *N = DIGlobalVariableExpression::get(VMContext, GV, Expr);
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
AllGVs.push_back(N);
return N;
}
DIGlobalVariable *DIBuilder::createTempGlobalVariableFwdDecl(
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *F,
unsigned LineNumber, DIType *Ty, bool isLocalToUnit, MDNode *Decl,
uint32_t AlignInBits) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
checkGlobalVariableScope(Context);
return DIGlobalVariable::getTemporary(
VMContext, cast_or_null<DIScope>(Context), Name, LinkageName, F,
LineNumber, Ty, isLocalToUnit, false,
cast_or_null<DIDerivedType>(Decl), AlignInBits)
.release();
}
static DILocalVariable *createLocalVariable(
LLVMContext &VMContext,
DenseMap<MDNode *, SmallVector<TrackingMDNodeRef, 1>> &PreservedVariables,
DIScope *Scope, StringRef Name, unsigned ArgNo, DIFile *File,
unsigned LineNo, DIType *Ty, bool AlwaysPreserve, DINode::DIFlags Flags,
uint32_t AlignInBits) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// FIXME: Why getNonCompileUnitScope()?
// FIXME: Why is "!Context" okay here?
// FIXME: Why doesn't this check for a subprogram or lexical block (AFAICT
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// the only valid scopes)?
DIScope *Context = getNonCompileUnitScope(Scope);
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
auto *Node =
DILocalVariable::get(VMContext, cast_or_null<DILocalScope>(Context), Name,
File, LineNo, Ty, ArgNo, Flags, AlignInBits);
if (AlwaysPreserve) {
// The optimizer may remove local variables. If there is an interest
// to preserve variable info in such situation then stash it in a
// named mdnode.
DISubprogram *Fn = getDISubprogram(Scope);
assert(Fn && "Missing subprogram for local variable");
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
PreservedVariables[Fn].emplace_back(Node);
}
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
return Node;
}
DILocalVariable *DIBuilder::createAutoVariable(DIScope *Scope, StringRef Name,
DIFile *File, unsigned LineNo,
DIType *Ty, bool AlwaysPreserve,
DINode::DIFlags Flags,
uint32_t AlignInBits) {
return createLocalVariable(VMContext, PreservedVariables, Scope, Name,
/* ArgNo */ 0, File, LineNo, Ty, AlwaysPreserve,
Flags, AlignInBits);
}
DILocalVariable *DIBuilder::createParameterVariable(
DIScope *Scope, StringRef Name, unsigned ArgNo, DIFile *File,
unsigned LineNo, DIType *Ty, bool AlwaysPreserve, DINode::DIFlags Flags) {
assert(ArgNo && "Expected non-zero argument number for parameter");
return createLocalVariable(VMContext, PreservedVariables, Scope, Name, ArgNo,
File, LineNo, Ty, AlwaysPreserve, Flags,
/* AlignInBits */0);
}
DILabel *DIBuilder::createLabel(
DIScope *Scope, StringRef Name, DIFile *File,
unsigned LineNo, bool AlwaysPreserve) {
DIScope *Context = getNonCompileUnitScope(Scope);
auto *Node =
DILabel::get(VMContext, cast_or_null<DILocalScope>(Context), Name,
File, LineNo);
if (AlwaysPreserve) {
/// The optimizer may remove labels. If there is an interest
/// to preserve label info in such situation then append it to
/// the list of retained nodes of the DISubprogram.
DISubprogram *Fn = getDISubprogram(Scope);
assert(Fn && "Missing subprogram for label");
PreservedLabels[Fn].emplace_back(Node);
}
return Node;
}
DIExpression *DIBuilder::createExpression(ArrayRef<uint64_t> Addr) {
return DIExpression::get(VMContext, Addr);
}
DIExpression *DIBuilder::createExpression(ArrayRef<int64_t> Signed) {
// TODO: Remove the callers of this signed version and delete.
SmallVector<uint64_t, 8> Addr(Signed.begin(), Signed.end());
return createExpression(Addr);
}
template <class... Ts>
static DISubprogram *getSubprogram(bool IsDistinct, Ts &&... Args) {
if (IsDistinct)
return DISubprogram::getDistinct(std::forward<Ts>(Args)...);
return DISubprogram::get(std::forward<Ts>(Args)...);
}
DISubprogram *DIBuilder::createFunction(
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *File,
unsigned LineNo, DISubroutineType *Ty, bool isLocalToUnit,
bool isDefinition, unsigned ScopeLine, DINode::DIFlags Flags,
bool isOptimized, DITemplateParameterArray TParams, DISubprogram *Decl,
DITypeArray ThrownTypes) {
auto *Node = getSubprogram(
/* IsDistinct = */ isDefinition, VMContext,
getNonCompileUnitScope(Context), Name, LinkageName, File, LineNo, Ty,
isLocalToUnit, isDefinition, ScopeLine, nullptr, 0, 0, 0, Flags,
isOptimized, isDefinition ? CUNode : nullptr, TParams, Decl,
MDTuple::getTemporary(VMContext, None).release(), ThrownTypes);
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
if (isDefinition)
AllSubprograms.push_back(Node);
trackIfUnresolved(Node);
return Node;
}
DISubprogram *DIBuilder::createTempFunctionFwdDecl(
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *File,
unsigned LineNo, DISubroutineType *Ty, bool isLocalToUnit,
bool isDefinition, unsigned ScopeLine, DINode::DIFlags Flags,
bool isOptimized, DITemplateParameterArray TParams, DISubprogram *Decl,
DITypeArray ThrownTypes) {
return DISubprogram::getTemporary(
VMContext, getNonCompileUnitScope(Context), Name, LinkageName,
File, LineNo, Ty, isLocalToUnit, isDefinition, ScopeLine, nullptr,
0, 0, 0, Flags, isOptimized, isDefinition ? CUNode : nullptr,
TParams, Decl, nullptr, ThrownTypes)
.release();
}
DISubprogram *DIBuilder::createMethod(
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *F,
unsigned LineNo, DISubroutineType *Ty, bool isLocalToUnit,
bool isDefinition, unsigned VK, unsigned VIndex, int ThisAdjustment,
DIType *VTableHolder, DINode::DIFlags Flags, bool isOptimized,
DITemplateParameterArray TParams, DITypeArray ThrownTypes) {
assert(getNonCompileUnitScope(Context) &&
"Methods should have both a Context and a context that isn't "
"the compile unit.");
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// FIXME: Do we want to use different scope/lines?
auto *SP = getSubprogram(
/* IsDistinct = */ isDefinition, VMContext, cast<DIScope>(Context), Name,
LinkageName, F, LineNo, Ty, isLocalToUnit, isDefinition, LineNo,
VTableHolder, VK, VIndex, ThisAdjustment, Flags, isOptimized,
isDefinition ? CUNode : nullptr, TParams, nullptr, nullptr, ThrownTypes);
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
if (isDefinition)
AllSubprograms.push_back(SP);
trackIfUnresolved(SP);
return SP;
}
DINamespace *DIBuilder::createNameSpace(DIScope *Scope, StringRef Name,
bool ExportSymbols) {
// It is okay to *not* make anonymous top-level namespaces distinct, because
// all nodes that have an anonymous namespace as their parent scope are
// guaranteed to be unique and/or are linked to their containing
// DICompileUnit. This decision is an explicit tradeoff of link time versus
// memory usage versus code simplicity and may get revisited in the future.
return DINamespace::get(VMContext, getNonCompileUnitScope(Scope), Name,
ExportSymbols);
}
DIModule *DIBuilder::createModule(DIScope *Scope, StringRef Name,
StringRef ConfigurationMacros,
StringRef IncludePath,
StringRef ISysRoot) {
return DIModule::get(VMContext, getNonCompileUnitScope(Scope), Name,
ConfigurationMacros, IncludePath, ISysRoot);
}
DILexicalBlockFile *DIBuilder::createLexicalBlockFile(DIScope *Scope,
DIFile *File,
unsigned Discriminator) {
return DILexicalBlockFile::get(VMContext, Scope, File, Discriminator);
}
DILexicalBlock *DIBuilder::createLexicalBlock(DIScope *Scope, DIFile *File,
unsigned Line, unsigned Col) {
DebugInfo: Move new hierarchy into place Move the specialized metadata nodes for the new debug info hierarchy into place, finishing off PR22464. I've done bootstraps (and all that) and I'm confident this commit is NFC as far as DWARF output is concerned. Let me know if I'm wrong :). The code changes are fairly mechanical: - Bumped the "Debug Info Version". - `DIBuilder` now creates the appropriate subclass of `MDNode`. - Subclasses of DIDescriptor now expect to hold their "MD" counterparts (e.g., `DIBasicType` expects `MDBasicType`). - Deleted a ton of dead code in `AsmWriter.cpp` and `DebugInfo.cpp` for printing comments. - Big update to LangRef to describe the nodes in the new hierarchy. Feel free to make it better. Testcase changes are enormous. There's an accompanying clang commit on its way. If you have out-of-tree debug info testcases, I just broke your build. - `upgrade-specialized-nodes.sh` is attached to PR22564. I used it to update all the IR testcases. - Unfortunately I failed to find way to script the updates to CHECK lines, so I updated all of these by hand. This was fairly painful, since the old CHECKs are difficult to reason about. That's one of the benefits of the new hierarchy. This work isn't quite finished, BTW. The `DIDescriptor` subclasses are almost empty wrappers, but not quite: they still have loose casting checks (see the `RETURN_FROM_RAW()` macro). Once they're completely gutted, I'll rename the "MD" classes to "DI" and kill the wrappers. I also expect to make a few schema changes now that it's easier to reason about everything. llvm-svn: 231082
2015-03-04 01:24:31 +08:00
// Make these distinct, to avoid merging two lexical blocks on the same
// file/line/column.
return DILexicalBlock::getDistinct(VMContext, getNonCompileUnitScope(Scope),
File, Line, Col);
}
Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
DIExpression *Expr, const DILocation *DL,
Instruction *InsertBefore) {
return insertDeclare(Storage, VarInfo, Expr, DL, InsertBefore->getParent(),
InsertBefore);
}
Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
DIExpression *Expr, const DILocation *DL,
BasicBlock *InsertAtEnd) {
// If this block already has a terminator then insert this intrinsic before
// the terminator. Otherwise, put it at the end of the block.
Instruction *InsertBefore = InsertAtEnd->getTerminator();
return insertDeclare(Storage, VarInfo, Expr, DL, InsertAtEnd, InsertBefore);
}
Instruction *DIBuilder::insertLabel(DILabel *LabelInfo, const DILocation *DL,
Instruction *InsertBefore) {
return insertLabel(
LabelInfo, DL, InsertBefore ? InsertBefore->getParent() : nullptr,
InsertBefore);
}
Instruction *DIBuilder::insertLabel(DILabel *LabelInfo, const DILocation *DL,
BasicBlock *InsertAtEnd) {
return insertLabel(LabelInfo, DL, InsertAtEnd, nullptr);
}
Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V,
DILocalVariable *VarInfo,
DIExpression *Expr,
const DILocation *DL,
Instruction *InsertBefore) {
return insertDbgValueIntrinsic(
V, VarInfo, Expr, DL, InsertBefore ? InsertBefore->getParent() : nullptr,
InsertBefore);
}
Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V,
DILocalVariable *VarInfo,
DIExpression *Expr,
const DILocation *DL,
BasicBlock *InsertAtEnd) {
return insertDbgValueIntrinsic(V, VarInfo, Expr, DL, InsertAtEnd, nullptr);
}
/// Return an IRBuilder for inserting dbg.declare and dbg.value intrinsics. This
/// abstracts over the various ways to specify an insert position.
static IRBuilder<> getIRBForDbgInsertion(const DILocation *DL,
BasicBlock *InsertBB,
Instruction *InsertBefore) {
IRBuilder<> B(DL->getContext());
if (InsertBefore)
B.SetInsertPoint(InsertBefore);
else if (InsertBB)
B.SetInsertPoint(InsertBB);
B.SetCurrentDebugLocation(DL);
return B;
}
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
static Value *getDbgIntrinsicValueImpl(LLVMContext &VMContext, Value *V) {
assert(V && "no value passed to dbg intrinsic");
return MetadataAsValue::get(VMContext, ValueAsMetadata::get(V));
}
static Function *getDeclareIntrin(Module &M) {
return Intrinsic::getDeclaration(&M, UseDbgAddr ? Intrinsic::dbg_addr
: Intrinsic::dbg_declare);
}
Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
DIExpression *Expr, const DILocation *DL,
BasicBlock *InsertBB, Instruction *InsertBefore) {
assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.declare");
assert(DL && "Expected debug loc");
assert(DL->getScope()->getSubprogram() ==
VarInfo->getScope()->getSubprogram() &&
"Expected matching subprograms");
if (!DeclareFn)
DeclareFn = getDeclareIntrin(M);
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.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
trackIfUnresolved(VarInfo);
trackIfUnresolved(Expr);
Value *Args[] = {getDbgIntrinsicValueImpl(VMContext, Storage),
MetadataAsValue::get(VMContext, VarInfo),
MetadataAsValue::get(VMContext, Expr)};
IRBuilder<> B = getIRBForDbgInsertion(DL, InsertBB, InsertBefore);
return B.CreateCall(DeclareFn, Args);
}
Instruction *DIBuilder::insertDbgValueIntrinsic(
Value *V, DILocalVariable *VarInfo, DIExpression *Expr,
const DILocation *DL, BasicBlock *InsertBB, Instruction *InsertBefore) {
assert(V && "no value passed to dbg.value");
assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.value");
assert(DL && "Expected debug loc");
assert(DL->getScope()->getSubprogram() ==
VarInfo->getScope()->getSubprogram() &&
"Expected matching subprograms");
if (!ValueFn)
ValueFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_value);
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
trackIfUnresolved(VarInfo);
trackIfUnresolved(Expr);
Value *Args[] = {getDbgIntrinsicValueImpl(VMContext, V),
MetadataAsValue::get(VMContext, VarInfo),
MetadataAsValue::get(VMContext, Expr)};
IRBuilder<> B = getIRBForDbgInsertion(DL, InsertBB, InsertBefore);
return B.CreateCall(ValueFn, Args);
}
Instruction *DIBuilder::insertLabel(
DILabel *LabelInfo, const DILocation *DL,
BasicBlock *InsertBB, Instruction *InsertBefore) {
assert(LabelInfo && "empty or invalid DILabel* passed to dbg.label");
assert(DL && "Expected debug loc");
assert(DL->getScope()->getSubprogram() ==
LabelInfo->getScope()->getSubprogram() &&
"Expected matching subprograms");
if (!LabelFn)
LabelFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_label);
trackIfUnresolved(LabelInfo);
Value *Args[] = {MetadataAsValue::get(VMContext, LabelInfo)};
IRBuilder<> B = getIRBForDbgInsertion(DL, InsertBB, InsertBefore);
return B.CreateCall(LabelFn, Args);
}
void DIBuilder::replaceVTableHolder(DICompositeType *&T,
DIType *VTableHolder) {
{
TypedTrackingMDRef<DICompositeType> N(T);
N->replaceVTableHolder(VTableHolder);
T = N.get();
}
// If this didn't create a self-reference, just return.
if (T != VTableHolder)
return;
// Look for unresolved operands. T will drop RAUW support, orphaning any
// cycles underneath it.
if (T->isResolved())
for (const MDOperand &O : T->operands())
if (auto *N = dyn_cast_or_null<MDNode>(O))
trackIfUnresolved(N);
}
void DIBuilder::replaceArrays(DICompositeType *&T, DINodeArray Elements,
DINodeArray TParams) {
{
TypedTrackingMDRef<DICompositeType> N(T);
if (Elements)
N->replaceElements(Elements);
if (TParams)
N->replaceTemplateParams(DITemplateParameterArray(TParams));
T = N.get();
}
// If T isn't resolved, there's no problem.
if (!T->isResolved())
return;
// If T is resolved, it may be due to a self-reference cycle. Track the
// arrays explicitly if they're unresolved, or else the cycles will be
// orphaned.
if (Elements)
trackIfUnresolved(Elements.get());
if (TParams)
trackIfUnresolved(TParams.get());
}