Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
//===--- 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.
|
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
2014-03-06 08:22:06 +08:00
|
|
|
#include "llvm/IR/DIBuilder.h"
|
2012-12-04 00:50:05 +08:00
|
|
|
#include "llvm/ADT/STLExtras.h"
|
2013-01-02 19:36:10 +08:00
|
|
|
#include "llvm/IR/Constants.h"
|
2014-03-06 08:46:21 +08:00
|
|
|
#include "llvm/IR/DebugInfo.h"
|
2013-01-02 19:36:10 +08:00
|
|
|
#include "llvm/IR/IntrinsicInst.h"
|
|
|
|
#include "llvm/IR/Module.h"
|
2012-04-03 08:43:49 +08:00
|
|
|
#include "llvm/Support/Debug.h"
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
#include "llvm/Support/Dwarf.h"
|
2016-03-13 19:11:39 +08:00
|
|
|
#include "LLVMContextImpl.h"
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
|
|
|
|
using namespace llvm;
|
|
|
|
using namespace llvm::dwarf;
|
|
|
|
|
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
|
|
|
DIBuilder::DIBuilder(Module &m, bool AllowUnresolvedNodes)
|
2015-07-03 06:32:52 +08:00
|
|
|
: M(m), VMContext(M.getContext()), CUNode(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
|
|
|
DeclareFn(nullptr), ValueFn(nullptr),
|
|
|
|
AllowUnresolvedNodes(AllowUnresolvedNodes) {}
|
|
|
|
|
|
|
|
void DIBuilder::trackIfUnresolved(MDNode *N) {
|
2015-01-20 03:09:14 +08:00
|
|
|
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;
|
2015-01-20 03:09:14 +08:00
|
|
|
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
|
|
|
}
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
|
2011-08-16 07:00:00 +08:00
|
|
|
void DIBuilder::finalize() {
|
2015-07-07 00:22:12 +08:00
|
|
|
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]);
|
2015-07-07 00:36:02 +08:00
|
|
|
|
|
|
|
if (!RetainValues.empty())
|
|
|
|
CUNode->replaceRetainedTypes(MDTuple::get(VMContext, RetainValues));
|
2015-07-07 00:22:12 +08:00
|
|
|
|
|
|
|
DISubprogramArray SPs = MDTuple::get(VMContext, AllSubprograms);
|
2016-04-15 23:57:41 +08:00
|
|
|
auto resolveVariables = [&](DISubprogram *SP) {
|
2016-04-21 04:03:59 +08:00
|
|
|
MDTuple *Temp = SP->getVariables().get();
|
|
|
|
if (!Temp)
|
|
|
|
return;
|
|
|
|
|
|
|
|
SmallVector<Metadata *, 4> Variables;
|
|
|
|
|
|
|
|
auto PV = PreservedVariables.find(SP);
|
|
|
|
if (PV != PreservedVariables.end())
|
|
|
|
Variables.append(PV->second.begin(), PV->second.end());
|
|
|
|
|
|
|
|
DINodeArray AV = getOrCreateArray(Variables);
|
|
|
|
TempMDTuple(Temp)->replaceAllUsesWith(AV.get());
|
2016-04-15 23:57:41 +08:00
|
|
|
};
|
|
|
|
for (auto *SP : SPs)
|
|
|
|
resolveVariables(SP);
|
|
|
|
for (auto *N : RetainValues)
|
|
|
|
if (auto *SP = dyn_cast<DISubprogram>(N))
|
|
|
|
resolveVariables(SP);
|
2011-08-17 06:09:43 +08:00
|
|
|
|
2015-07-07 00:36:02 +08:00
|
|
|
if (!AllGVs.empty())
|
|
|
|
CUNode->replaceGlobalVariables(MDTuple::get(VMContext, AllGVs));
|
2013-04-22 14:12:31 +08:00
|
|
|
|
2015-07-07 00:36:02 +08:00
|
|
|
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
|
|
|
|
2017-01-12 23:49:46 +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)
|
2015-01-20 07:13:14 +08:00
|
|
|
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;
|
2011-08-17 06:09:43 +08:00
|
|
|
}
|
|
|
|
|
2014-10-02 05:32:15 +08:00
|
|
|
/// If N is compile unit return NULL otherwise return N.
|
2015-04-30 00:38:44 +08:00
|
|
|
static DIScope *getNonCompileUnitScope(DIScope *N) {
|
|
|
|
if (!N || isa<DICompileUnit>(N))
|
2014-04-09 14:08:46 +08:00
|
|
|
return nullptr;
|
2015-04-30 00:38:44 +08:00
|
|
|
return cast<DIScope>(N);
|
2011-08-16 07:00:00 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompileUnit *DIBuilder::createCompileUnit(
|
2016-12-15 04:24:54 +08:00
|
|
|
unsigned Lang, DIFile *File, StringRef Producer, bool isOptimized,
|
|
|
|
StringRef Flags, unsigned RunTimeVer, StringRef SplitName,
|
2016-08-25 02:29:49 +08:00
|
|
|
DICompileUnit::DebugEmissionKind Kind, uint64_t DWOId,
|
Change debug-info-for-profiling from a TargetOption to a function attribute.
Summary: LTO requires the debug-info-for-profiling to be a function attribute.
Reviewers: echristo, mehdi_amini, dblaikie, probinson, aprantl
Reviewed By: mehdi_amini, dblaikie, aprantl
Subscribers: aprantl, probinson, ahatanak, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D29203
llvm-svn: 293833
2017-02-02 06:45:09 +08:00
|
|
|
bool SplitDebugInlining, bool DebugInfoForProfiling) {
|
2014-02-27 09:24:56 +08:00
|
|
|
|
2015-02-07 14:35:30 +08:00
|
|
|
assert(((Lang <= dwarf::DW_LANG_Fortran08 && Lang >= dwarf::DW_LANG_C89) ||
|
2012-01-11 02:18:52 +08:00
|
|
|
(Lang <= dwarf::DW_LANG_hi_user && Lang >= dwarf::DW_LANG_lo_user)) &&
|
|
|
|
"Invalid Language tag");
|
2011-08-17 06:09:43 +08:00
|
|
|
|
2015-07-03 06:32:52 +08:00
|
|
|
assert(!CUNode && "Can only make one compile unit per DIBuilder instance");
|
|
|
|
CUNode = DICompileUnit::getDistinct(
|
2016-12-15 04:24:54 +08:00
|
|
|
VMContext, Lang, File, Producer, isOptimized, Flags, RunTimeVer,
|
|
|
|
SplitName, Kind, nullptr, nullptr, nullptr, nullptr, nullptr, DWOId,
|
Change debug-info-for-profiling from a TargetOption to a function attribute.
Summary: LTO requires the debug-info-for-profiling to be a function attribute.
Reviewers: echristo, mehdi_amini, dblaikie, probinson, aprantl
Reviewed By: mehdi_amini, dblaikie, aprantl
Subscribers: aprantl, probinson, ahatanak, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D29203
llvm-svn: 293833
2017-02-02 06:45:09 +08:00
|
|
|
SplitDebugInlining, DebugInfoForProfiling);
|
2011-05-04 00:18:28 +08:00
|
|
|
|
|
|
|
// Create a named metadata so that it is easier to find cu in a module.
|
2016-04-09 06:43:03 +08:00
|
|
|
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);
|
2015-04-07 07:18:49 +08:00
|
|
|
return CUNode;
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
static DIImportedEntity *
|
|
|
|
createImportedModule(LLVMContext &C, dwarf::Tag Tag, DIScope *Context,
|
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
|
|
|
Metadata *NS, unsigned Line, StringRef Name,
|
|
|
|
SmallVectorImpl<TrackingMDNodeRef> &AllImportedModules) {
|
2016-03-13 19:11:39 +08:00
|
|
|
unsigned EntitiesCount = C.pImpl->DIImportedEntitys.size();
|
2015-04-30 00:38:44 +08:00
|
|
|
auto *M = DIImportedEntity::get(C, Tag, Context, DINodeRef(NS), Line, Name);
|
2016-03-13 19:11:39 +08:00
|
|
|
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);
|
2013-05-08 05:35:53 +08:00
|
|
|
return M;
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIImportedEntity *DIBuilder::createImportedModule(DIScope *Context,
|
|
|
|
DINamespace *NS,
|
|
|
|
unsigned Line) {
|
2014-04-06 14:29:01 +08:00
|
|
|
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_module,
|
|
|
|
Context, NS, Line, StringRef(), AllImportedModules);
|
2013-05-21 06:50:35 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIImportedEntity *DIBuilder::createImportedModule(DIScope *Context,
|
|
|
|
DIImportedEntity *NS,
|
|
|
|
unsigned Line) {
|
2014-04-06 14:29:01 +08:00
|
|
|
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_module,
|
|
|
|
Context, NS, Line, StringRef(), AllImportedModules);
|
2013-05-21 06:50:35 +08:00
|
|
|
}
|
|
|
|
|
2015-06-30 07:03:47 +08:00
|
|
|
DIImportedEntity *DIBuilder::createImportedModule(DIScope *Context, DIModule *M,
|
|
|
|
unsigned Line) {
|
|
|
|
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_module,
|
|
|
|
Context, M, Line, StringRef(), AllImportedModules);
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIImportedEntity *DIBuilder::createImportedDeclaration(DIScope *Context,
|
|
|
|
DINode *Decl,
|
2015-04-17 00:36:23 +08:00
|
|
|
unsigned Line,
|
|
|
|
StringRef Name) {
|
2014-11-07 01:46:55 +08:00
|
|
|
// Make sure to use the unique identifier based metadata reference for
|
|
|
|
// types that have one.
|
2014-04-06 14:29:01 +08:00
|
|
|
return ::createImportedModule(VMContext, dwarf::DW_TAG_imported_declaration,
|
2016-04-24 05:08:00 +08:00
|
|
|
Context, Decl, Line, Name, AllImportedModules);
|
2013-04-22 14:12:31 +08:00
|
|
|
}
|
|
|
|
|
2016-12-25 18:12:09 +08:00
|
|
|
DIFile *DIBuilder::createFile(StringRef Filename, StringRef Directory,
|
|
|
|
DIFile::ChecksumKind CSKind, StringRef Checksum) {
|
|
|
|
return DIFile::get(VMContext, Filename, Directory, CSKind, Checksum);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2017-01-12 23:49:46 +08:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIEnumerator *DIBuilder::createEnumerator(StringRef Name, int64_t Val) {
|
2011-09-13 02:26:08 +08:00
|
|
|
assert(!Name.empty() && "Unable to create enumerator without name");
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIEnumerator::get(VMContext, Val, Name);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIBasicType *DIBuilder::createUnspecifiedType(StringRef Name) {
|
2011-09-15 07:13:28 +08:00
|
|
|
assert(!Name.empty() && "Unable to create type without name");
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIBasicType::get(VMContext, dwarf::DW_TAG_unspecified_type, Name);
|
2011-09-15 07:13:28 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIBasicType *DIBuilder::createNullPtrType() {
|
2013-06-28 06:50:59 +08:00
|
|
|
return createUnspecifiedType("decltype(nullptr)");
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIBasicType *DIBuilder::createBasicType(StringRef Name, uint64_t SizeInBits,
|
2015-04-17 00:36:23 +08:00
|
|
|
unsigned Encoding) {
|
2011-09-13 02:26:08 +08:00
|
|
|
assert(!Name.empty() && "Unable to create type without name");
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIBasicType::get(VMContext, dwarf::DW_TAG_base_type, Name, SizeInBits,
|
2016-10-20 08:13:12 +08:00
|
|
|
0, Encoding);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIDerivedType *DIBuilder::createQualifiedType(unsigned Tag, DIType *FromTy) {
|
2016-04-24 05:08:00 +08:00
|
|
|
return DIDerivedType::get(VMContext, Tag, "", nullptr, 0, nullptr, FromTy, 0,
|
2016-09-06 18:46:28 +08:00
|
|
|
0, 0, DINode::FlagZero);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIDerivedType *DIBuilder::createPointerType(DIType *PointeeTy,
|
2015-04-17 00:36:23 +08:00
|
|
|
uint64_t SizeInBits,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint32_t AlignInBits,
|
2015-04-17 00:36:23 +08:00
|
|
|
StringRef Name) {
|
2015-03-04 01:24:31 +08:00
|
|
|
// FIXME: Why is there a name here?
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_pointer_type, Name,
|
2016-04-24 05:08:00 +08:00
|
|
|
nullptr, 0, nullptr, PointeeTy, SizeInBits,
|
2016-09-06 18:46:28 +08:00
|
|
|
AlignInBits, 0, DINode::FlagZero);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIDerivedType *DIBuilder::createMemberPointerType(DIType *PointeeTy,
|
|
|
|
DIType *Base,
|
2015-04-17 00:36:23 +08:00
|
|
|
uint64_t SizeInBits,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint32_t AlignInBits,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::DIFlags Flags) {
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_ptr_to_member_type, "",
|
2016-04-24 05:08:00 +08:00
|
|
|
nullptr, 0, nullptr, PointeeTy, SizeInBits,
|
2016-06-18 05:31:33 +08:00
|
|
|
AlignInBits, 0, Flags, Base);
|
2013-01-07 13:51:15 +08:00
|
|
|
}
|
|
|
|
|
2015-11-16 15:57:32 +08:00
|
|
|
DIDerivedType *DIBuilder::createReferenceType(unsigned Tag, DIType *RTy,
|
|
|
|
uint64_t SizeInBits,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint32_t AlignInBits) {
|
2015-04-07 07:18:49 +08:00
|
|
|
assert(RTy && "Unable to create reference type");
|
2016-04-24 05:08:00 +08:00
|
|
|
return DIDerivedType::get(VMContext, Tag, "", nullptr, 0, nullptr, RTy,
|
2016-09-06 18:46:28 +08:00
|
|
|
SizeInBits, AlignInBits, 0, DINode::FlagZero);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIDerivedType *DIBuilder::createTypedef(DIType *Ty, StringRef Name,
|
|
|
|
DIFile *File, unsigned LineNo,
|
|
|
|
DIScope *Context) {
|
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_typedef, Name, File,
|
2016-04-24 05:08:00 +08:00
|
|
|
LineNo, getNonCompileUnitScope(Context), Ty, 0, 0,
|
2016-09-06 18:46:28 +08:00
|
|
|
0, DINode::FlagZero);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIDerivedType *DIBuilder::createFriend(DIType *Ty, DIType *FriendTy) {
|
2015-04-07 07:18:49 +08:00
|
|
|
assert(Ty && "Invalid type!");
|
|
|
|
assert(FriendTy && "Invalid friend type!");
|
2016-04-24 05:08:00 +08:00
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_friend, "", nullptr, 0, Ty,
|
2016-09-06 18:46:28 +08:00
|
|
|
FriendTy, 0, 0, 0, DINode::FlagZero);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIDerivedType *DIBuilder::createInheritance(DIType *Ty, DIType *BaseTy,
|
2015-04-17 00:36:23 +08:00
|
|
|
uint64_t BaseOffset,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::DIFlags Flags) {
|
2015-04-07 07:18:49 +08:00
|
|
|
assert(Ty && "Unable to create inheritance");
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_inheritance, "", nullptr,
|
2016-04-24 05:08:00 +08:00
|
|
|
0, Ty, BaseTy, 0, 0, BaseOffset, Flags);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIDerivedType *DIBuilder::createMemberType(DIScope *Scope, StringRef Name,
|
|
|
|
DIFile *File, unsigned LineNumber,
|
2015-04-17 00:36:23 +08:00
|
|
|
uint64_t SizeInBits,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint32_t AlignInBits,
|
2015-04-17 00:36:23 +08:00
|
|
|
uint64_t OffsetInBits,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::DIFlags Flags, DIType *Ty) {
|
2016-04-24 05:08:00 +08:00
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_member, Name, File,
|
|
|
|
LineNumber, getNonCompileUnitScope(Scope), Ty,
|
|
|
|
SizeInBits, AlignInBits, OffsetInBits, Flags);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
|
|
|
|
2015-03-27 08:34:10 +08:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2016-06-30 11:00:20 +08:00
|
|
|
DIDerivedType *DIBuilder::createBitFieldMemberType(
|
|
|
|
DIScope *Scope, StringRef Name, DIFile *File, unsigned LineNumber,
|
2016-10-20 08:13:12 +08:00
|
|
|
uint64_t SizeInBits, uint64_t OffsetInBits, uint64_t StorageOffsetInBits,
|
|
|
|
DINode::DIFlags Flags, DIType *Ty) {
|
2016-06-30 11:00:20 +08:00
|
|
|
Flags |= DINode::FlagBitField;
|
|
|
|
return DIDerivedType::get(
|
|
|
|
VMContext, dwarf::DW_TAG_member, Name, File, LineNumber,
|
2016-10-20 08:13:12 +08:00
|
|
|
getNonCompileUnitScope(Scope), Ty, SizeInBits, /* AlignInBits */ 0,
|
|
|
|
OffsetInBits, Flags,
|
|
|
|
ConstantAsMetadata::get(ConstantInt::get(IntegerType::get(VMContext, 64),
|
|
|
|
StorageOffsetInBits)));
|
2016-06-30 11:00:20 +08:00
|
|
|
}
|
|
|
|
|
2016-09-07 01:03:02 +08:00
|
|
|
DIDerivedType *
|
|
|
|
DIBuilder::createStaticMemberType(DIScope *Scope, StringRef Name, DIFile *File,
|
|
|
|
unsigned LineNumber, DIType *Ty,
|
2016-10-20 08:13:12 +08:00
|
|
|
DINode::DIFlags Flags, llvm::Constant *Val,
|
|
|
|
uint32_t AlignInBits) {
|
2015-04-30 00:38:44 +08:00
|
|
|
Flags |= DINode::FlagStaticMember;
|
2016-04-24 05:08:00 +08:00
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_member, Name, File,
|
2016-10-20 08:13:12 +08:00
|
|
|
LineNumber, getNonCompileUnitScope(Scope), Ty, 0,
|
|
|
|
AlignInBits, 0, Flags, getConstantOrNull(Val));
|
2013-01-16 09:22:23 +08:00
|
|
|
}
|
|
|
|
|
2016-09-07 01:03:02 +08:00
|
|
|
DIDerivedType *
|
|
|
|
DIBuilder::createObjCIVar(StringRef Name, DIFile *File, unsigned LineNumber,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint64_t SizeInBits, uint32_t AlignInBits,
|
2016-09-07 01:03:02 +08:00
|
|
|
uint64_t OffsetInBits, DINode::DIFlags Flags,
|
|
|
|
DIType *Ty, MDNode *PropertyNode) {
|
2016-04-24 05:08:00 +08:00
|
|
|
return DIDerivedType::get(VMContext, dwarf::DW_TAG_member, Name, File,
|
|
|
|
LineNumber, getNonCompileUnitScope(File), Ty,
|
|
|
|
SizeInBits, AlignInBits, OffsetInBits, Flags,
|
|
|
|
PropertyNode);
|
2011-04-16 08:11:51 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIObjCProperty *
|
|
|
|
DIBuilder::createObjCProperty(StringRef Name, DIFile *File, unsigned LineNumber,
|
2013-10-16 07:31:31 +08:00
|
|
|
StringRef GetterName, StringRef SetterName,
|
2015-04-30 00:38:44 +08:00
|
|
|
unsigned PropertyAttributes, DIType *Ty) {
|
|
|
|
return DIObjCProperty::get(VMContext, Name, File, LineNumber, GetterName,
|
2016-04-24 05:08:00 +08:00
|
|
|
SetterName, PropertyAttributes, Ty);
|
2012-02-04 08:59:25 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DITemplateTypeParameter *
|
|
|
|
DIBuilder::createTemplateTypeParameter(DIScope *Context, StringRef Name,
|
|
|
|
DIType *Ty) {
|
|
|
|
assert((!Context || isa<DICompileUnit>(Context)) && "Expected compile unit");
|
2016-04-24 05:08:00 +08:00
|
|
|
return DITemplateTypeParameter::get(VMContext, Name, Ty);
|
2011-02-03 05:38:25 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
static DITemplateValueParameter *
|
2015-02-13 11:35:29 +08:00
|
|
|
createTemplateValueParameterHelper(LLVMContext &VMContext, unsigned Tag,
|
2015-04-30 00:38:44 +08:00
|
|
|
DIScope *Context, StringRef Name, DIType *Ty,
|
2015-04-17 00:36:23 +08:00
|
|
|
Metadata *MD) {
|
2015-04-30 00:38:44 +08:00
|
|
|
assert((!Context || isa<DICompileUnit>(Context)) && "Expected compile unit");
|
2016-04-24 05:08:00 +08:00
|
|
|
return DITemplateValueParameter::get(VMContext, Tag, Name, Ty, MD);
|
2011-02-03 06:35:53 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DITemplateValueParameter *
|
|
|
|
DIBuilder::createTemplateValueParameter(DIScope *Context, StringRef Name,
|
|
|
|
DIType *Ty, Constant *Val) {
|
2014-11-15 08:05:04 +08:00
|
|
|
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,
|
2015-02-13 11:35:29 +08:00
|
|
|
getConstantOrNull(Val));
|
2013-06-23 02:59:11 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DITemplateValueParameter *
|
|
|
|
DIBuilder::createTemplateTemplateParameter(DIScope *Context, StringRef Name,
|
|
|
|
DIType *Ty, StringRef Val) {
|
2014-11-15 08:05:04 +08:00
|
|
|
return createTemplateValueParameterHelper(
|
|
|
|
VMContext, dwarf::DW_TAG_GNU_template_template_param, Context, Name, Ty,
|
2015-02-13 11:35:29 +08:00
|
|
|
MDString::get(VMContext, Val));
|
2013-06-23 02:59:11 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DITemplateValueParameter *
|
|
|
|
DIBuilder::createTemplateParameterPack(DIScope *Context, StringRef Name,
|
|
|
|
DIType *Ty, DINodeArray Val) {
|
2014-11-15 08:05:04 +08:00
|
|
|
return createTemplateValueParameterHelper(
|
|
|
|
VMContext, dwarf::DW_TAG_GNU_template_parameter_pack, Context, Name, Ty,
|
2015-04-08 00:50:39 +08:00
|
|
|
Val.get());
|
2013-06-23 02:59:11 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompositeType *DIBuilder::createClassType(
|
|
|
|
DIScope *Context, StringRef Name, DIFile *File, unsigned LineNumber,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint64_t SizeInBits, uint32_t AlignInBits, uint64_t OffsetInBits,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::DIFlags Flags, DIType *DerivedFrom, DINodeArray Elements,
|
2015-04-30 00:38:44 +08:00
|
|
|
DIType *VTableHolder, MDNode *TemplateParams, StringRef UniqueIdentifier) {
|
|
|
|
assert((!Context || isa<DIScope>(Context)) &&
|
2013-03-12 07:21:19 +08:00
|
|
|
"createClassType should be called with a valid Context");
|
2015-04-17 00:36:23 +08:00
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
auto *R = DICompositeType::get(
|
2015-03-04 01:24:31 +08:00
|
|
|
VMContext, dwarf::DW_TAG_structure_type, Name, File, LineNumber,
|
2016-04-24 05:08:00 +08:00
|
|
|
getNonCompileUnitScope(Context), DerivedFrom, SizeInBits, AlignInBits,
|
|
|
|
OffsetInBits, Flags, Elements, 0, VTableHolder,
|
2015-04-07 03:03:45 +08:00
|
|
|
cast_or_null<MDTuple>(TemplateParams), UniqueIdentifier);
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(R);
|
2013-03-12 07:21:19 +08:00
|
|
|
return R;
|
2012-07-06 10:35:57 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompositeType *DIBuilder::createStructType(
|
|
|
|
DIScope *Context, StringRef Name, DIFile *File, unsigned LineNumber,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint64_t SizeInBits, uint32_t AlignInBits, DINode::DIFlags Flags,
|
2015-04-30 00:38:44 +08:00
|
|
|
DIType *DerivedFrom, DINodeArray Elements, unsigned RunTimeLang,
|
|
|
|
DIType *VTableHolder, StringRef UniqueIdentifier) {
|
|
|
|
auto *R = DICompositeType::get(
|
2015-03-04 01:24:31 +08:00
|
|
|
VMContext, dwarf::DW_TAG_structure_type, Name, File, LineNumber,
|
2016-04-24 05:08:00 +08:00
|
|
|
getNonCompileUnitScope(Context), DerivedFrom, SizeInBits, AlignInBits, 0,
|
|
|
|
Flags, Elements, RunTimeLang, VTableHolder, nullptr, UniqueIdentifier);
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(R);
|
2013-03-12 07:21:19 +08:00
|
|
|
return R;
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompositeType *DIBuilder::createUnionType(
|
|
|
|
DIScope *Scope, StringRef Name, DIFile *File, unsigned LineNumber,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint64_t SizeInBits, uint32_t AlignInBits, DINode::DIFlags Flags,
|
2015-04-30 00:38:44 +08:00
|
|
|
DINodeArray Elements, unsigned RunTimeLang, StringRef UniqueIdentifier) {
|
|
|
|
auto *R = DICompositeType::get(
|
2015-03-04 01:24:31 +08:00
|
|
|
VMContext, dwarf::DW_TAG_union_type, Name, File, LineNumber,
|
2016-04-24 05:08:00 +08:00
|
|
|
getNonCompileUnitScope(Scope), nullptr, SizeInBits, AlignInBits, 0, Flags,
|
|
|
|
Elements, RunTimeLang, nullptr, nullptr, UniqueIdentifier);
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(R);
|
2013-08-30 07:17:54 +08:00
|
|
|
return R;
|
2010-12-08 09:50:15 +08:00
|
|
|
}
|
|
|
|
|
2015-10-15 14:56:10 +08:00
|
|
|
DISubroutineType *DIBuilder::createSubroutineType(DITypeRefArray ParameterTypes,
|
2016-09-07 01:03:02 +08:00
|
|
|
DINode::DIFlags Flags,
|
|
|
|
unsigned CC) {
|
2016-06-09 04:34:29 +08:00
|
|
|
return DISubroutineType::get(VMContext, Flags, CC, ParameterTypes);
|
2010-12-08 09:50:15 +08:00
|
|
|
}
|
|
|
|
|
2015-07-16 01:01:41 +08:00
|
|
|
DICompositeType *DIBuilder::createExternalTypeRef(unsigned Tag, DIFile *File,
|
|
|
|
StringRef UniqueIdentifier) {
|
|
|
|
assert(!UniqueIdentifier.empty() && "external type ref without uid");
|
2016-04-24 05:08:00 +08:00
|
|
|
return DICompositeType::get(VMContext, Tag, "", nullptr, 0, nullptr, nullptr,
|
|
|
|
0, 0, 0, DINode::FlagExternalTypeRef, nullptr, 0,
|
|
|
|
nullptr, nullptr, UniqueIdentifier);
|
2015-07-16 01:01:41 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompositeType *DIBuilder::createEnumerationType(
|
|
|
|
DIScope *Scope, StringRef Name, DIFile *File, unsigned LineNumber,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint64_t SizeInBits, uint32_t AlignInBits, DINodeArray Elements,
|
2015-04-30 00:38:44 +08:00
|
|
|
DIType *UnderlyingType, StringRef UniqueIdentifier) {
|
|
|
|
auto *CTy = DICompositeType::get(
|
2015-03-04 01:24:31 +08:00
|
|
|
VMContext, dwarf::DW_TAG_enumeration_type, Name, File, LineNumber,
|
2016-04-24 05:08:00 +08:00
|
|
|
getNonCompileUnitScope(Scope), UnderlyingType, SizeInBits, AlignInBits, 0,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::FlagZero, Elements, 0, nullptr, nullptr, UniqueIdentifier);
|
2013-11-19 07:33:32 +08:00
|
|
|
AllEnumTypes.push_back(CTy);
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(CTy);
|
2013-11-19 07:33:32 +08:00
|
|
|
return CTy;
|
2010-12-08 09:50:15 +08:00
|
|
|
}
|
|
|
|
|
2016-10-18 22:31:22 +08:00
|
|
|
DICompositeType *DIBuilder::createArrayType(uint64_t Size,
|
|
|
|
uint32_t AlignInBits, DIType *Ty,
|
2015-04-30 00:38:44 +08:00
|
|
|
DINodeArray Subscripts) {
|
|
|
|
auto *R = DICompositeType::get(VMContext, dwarf::DW_TAG_array_type, "",
|
2016-04-24 05:08:00 +08:00
|
|
|
nullptr, 0, nullptr, Ty, Size, AlignInBits, 0,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::FlagZero, Subscripts, 0, nullptr);
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(R);
|
|
|
|
return R;
|
2010-12-08 09:50:15 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompositeType *DIBuilder::createVectorType(uint64_t Size,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint32_t AlignInBits, DIType *Ty,
|
2015-04-30 00:38:44 +08:00
|
|
|
DINodeArray Subscripts) {
|
2016-04-24 05:08:00 +08:00
|
|
|
auto *R = DICompositeType::get(VMContext, dwarf::DW_TAG_array_type, "",
|
|
|
|
nullptr, 0, nullptr, Ty, Size, AlignInBits, 0,
|
|
|
|
DINode::FlagVector, Subscripts, 0, nullptr);
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(R);
|
|
|
|
return R;
|
2010-12-08 09:50:15 +08:00
|
|
|
}
|
2010-12-08 07:25:47 +08:00
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
static DIType *createTypeWithFlags(LLVMContext &Context, DIType *Ty,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::DIFlags FlagsToSet) {
|
2015-04-17 00:36:23 +08:00
|
|
|
auto NewTy = Ty->clone();
|
2015-03-04 01:24:31 +08:00
|
|
|
NewTy->setFlags(NewTy->getFlags() | FlagsToSet);
|
|
|
|
return MDNode::replaceWithUniqued(std::move(NewTy));
|
2014-10-04 04:01:09 +08:00
|
|
|
}
|
2014-10-03 06:15:31 +08:00
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIType *DIBuilder::createArtificialType(DIType *Ty) {
|
2015-03-04 01:24:31 +08:00
|
|
|
// FIXME: Restrict this to the nodes where it's valid.
|
2015-04-16 09:01:28 +08:00
|
|
|
if (Ty->isArtificial())
|
2014-10-04 04:01:09 +08:00
|
|
|
return Ty;
|
2015-04-30 00:38:44 +08:00
|
|
|
return createTypeWithFlags(VMContext, Ty, DINode::FlagArtificial);
|
Introduce DIBuilder. It is intended to be a front-end friendly interface to emit debuggging information entries in LLVM IR.
To create debugging information for a pointer, using DIBUilder front-end just needs
DBuilder.CreatePointerType(Ty, Size);
instead of
DebugFactory.CreateDerivedType(llvm::dwarf::DW_TAG_pointer_type,
TheCU, "", getOrCreateMainFile(),
0, Size, 0, 0, 0, OCTy);
llvm-svn: 118248
2010-11-04 23:01:38 +08:00
|
|
|
}
|
2010-12-08 07:25:47 +08:00
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIType *DIBuilder::createObjectPointerType(DIType *Ty) {
|
2015-03-04 01:24:31 +08:00
|
|
|
// FIXME: Restrict this to the nodes where it's valid.
|
2015-04-16 09:01:28 +08:00
|
|
|
if (Ty->isObjectPointer())
|
2012-09-13 07:36:19 +08:00
|
|
|
return Ty;
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::DIFlags Flags = DINode::FlagObjectPointer | DINode::FlagArtificial;
|
2014-10-04 04:01:09 +08:00
|
|
|
return createTypeWithFlags(VMContext, Ty, Flags);
|
2012-09-13 07:36:19 +08:00
|
|
|
}
|
|
|
|
|
2016-04-15 23:57:41 +08:00
|
|
|
void DIBuilder::retainType(DIScope *T) {
|
2015-04-16 09:01:28 +08:00
|
|
|
assert(T && "Expected non-null type");
|
2016-04-15 23:57:41 +08:00
|
|
|
assert((isa<DIType>(T) || (isa<DISubprogram>(T) &&
|
|
|
|
cast<DISubprogram>(T)->isDefinition() == false)) &&
|
|
|
|
"Expected type or subprogram declaration");
|
2015-03-28 07:00:49 +08:00
|
|
|
AllRetainTypes.emplace_back(T);
|
|
|
|
}
|
2010-12-08 09:50:15 +08:00
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIBasicType *DIBuilder::createUnspecifiedParameter() { return nullptr; }
|
2010-12-08 09:50:15 +08:00
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompositeType *
|
|
|
|
DIBuilder::createForwardDecl(unsigned Tag, StringRef Name, DIScope *Scope,
|
|
|
|
DIFile *F, unsigned Line, unsigned RuntimeLang,
|
2016-10-18 22:31:22 +08:00
|
|
|
uint64_t SizeInBits, uint32_t AlignInBits,
|
2013-10-16 07:31:31 +08:00
|
|
|
StringRef UniqueIdentifier) {
|
2015-03-04 01:24:31 +08:00
|
|
|
// FIXME: Define in terms of createReplaceableForwardDecl() by calling
|
|
|
|
// replaceWithUniqued().
|
2015-04-30 00:38:44 +08:00
|
|
|
auto *RetTy = DICompositeType::get(
|
2016-04-24 05:08:00 +08:00
|
|
|
VMContext, Tag, Name, F, Line, getNonCompileUnitScope(Scope), nullptr,
|
|
|
|
SizeInBits, AlignInBits, 0, DINode::FlagFwdDecl, nullptr, RuntimeLang,
|
|
|
|
nullptr, nullptr, UniqueIdentifier);
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(RetTy);
|
2014-05-06 11:41:57 +08:00
|
|
|
return RetTy;
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DICompositeType *DIBuilder::createReplaceableCompositeType(
|
|
|
|
unsigned Tag, StringRef Name, DIScope *Scope, DIFile *F, unsigned Line,
|
2016-10-18 22:31:22 +08:00
|
|
|
unsigned RuntimeLang, uint64_t SizeInBits, uint32_t AlignInBits,
|
2016-09-06 18:46:28 +08:00
|
|
|
DINode::DIFlags Flags, StringRef UniqueIdentifier) {
|
2016-04-24 05:08:00 +08:00
|
|
|
auto *RetTy =
|
|
|
|
DICompositeType::getTemporary(
|
|
|
|
VMContext, Tag, Name, F, Line, getNonCompileUnitScope(Scope), nullptr,
|
|
|
|
SizeInBits, AlignInBits, 0, Flags, nullptr, RuntimeLang, nullptr,
|
|
|
|
nullptr, UniqueIdentifier)
|
|
|
|
.release();
|
2015-02-18 03:17:39 +08:00
|
|
|
trackIfUnresolved(RetTy);
|
2013-07-03 02:37:35 +08:00
|
|
|
return RetTy;
|
2012-02-08 08:22:26 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DINodeArray DIBuilder::getOrCreateArray(ArrayRef<Metadata *> Elements) {
|
2015-04-17 00:36:23 +08:00
|
|
|
return MDTuple::get(VMContext, Elements);
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2017-01-12 23:49:46 +08:00
|
|
|
DIMacroNodeArray
|
|
|
|
DIBuilder::getOrCreateMacroArray(ArrayRef<Metadata *> Elements) {
|
|
|
|
return MDTuple::get(VMContext, Elements);
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
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;
|
2014-07-29 03:33:20 +08:00
|
|
|
for (unsigned i = 0, e = Elements.size(); i != e; ++i) {
|
|
|
|
if (Elements[i] && isa<MDNode>(Elements[i]))
|
2016-04-24 05:08:00 +08:00
|
|
|
Elts.push_back(cast<DIType>(Elements[i]));
|
2014-07-29 03:33:20 +08:00
|
|
|
else
|
|
|
|
Elts.push_back(Elements[i]);
|
|
|
|
}
|
2015-04-30 00:38:44 +08:00
|
|
|
return DITypeRefArray(MDNode::get(VMContext, Elts));
|
2014-07-29 03:33:20 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DISubrange *DIBuilder::getOrCreateSubrange(int64_t Lo, int64_t Count) {
|
|
|
|
return DISubrange::get(VMContext, Count, Lo);
|
2010-12-08 09:50:15 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
static void checkGlobalVariableScope(DIScope *Context) {
|
2015-04-07 07:34:41 +08:00
|
|
|
#ifndef NDEBUG
|
2015-04-16 09:01:28 +08:00
|
|
|
if (auto *CT =
|
2015-04-30 00:38:44 +08:00
|
|
|
dyn_cast_or_null<DICompositeType>(getNonCompileUnitScope(Context)))
|
2015-04-16 09:01:28 +08:00
|
|
|
assert(CT->getIdentifier().empty() &&
|
2014-11-22 03:47:48 +08:00
|
|
|
"Context of a global variable should not be a type with identifier");
|
2015-04-07 07:34:41 +08:00
|
|
|
#endif
|
2014-09-17 17:28:34 +08:00
|
|
|
}
|
|
|
|
|
2016-12-20 10:09:43 +08:00
|
|
|
DIGlobalVariableExpression *DIBuilder::createGlobalVariableExpression(
|
2015-04-30 00:38:44 +08:00
|
|
|
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *F,
|
2016-12-20 10:09:43 +08:00
|
|
|
unsigned LineNumber, DIType *Ty, bool isLocalToUnit, DIExpression *Expr,
|
|
|
|
MDNode *Decl, uint32_t AlignInBits) {
|
2015-03-04 01:24:31 +08:00
|
|
|
checkGlobalVariableScope(Context);
|
|
|
|
|
2016-12-20 10:09:43 +08:00
|
|
|
auto *GV = DIGlobalVariable::getDistinct(
|
2016-04-24 06:29:09 +08:00
|
|
|
VMContext, cast_or_null<DIScope>(Context), Name, LinkageName, F,
|
2016-12-20 10:09:43 +08:00
|
|
|
LineNumber, Ty, isLocalToUnit, true, cast_or_null<DIDerivedType>(Decl),
|
|
|
|
AlignInBits);
|
|
|
|
auto *N = DIGlobalVariableExpression::get(VMContext, GV, Expr);
|
2015-03-04 01:24:31 +08:00
|
|
|
AllGVs.push_back(N);
|
|
|
|
return N;
|
2014-09-17 17:28:34 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIGlobalVariable *DIBuilder::createTempGlobalVariableFwdDecl(
|
|
|
|
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *F,
|
2016-12-20 10:09:43 +08:00
|
|
|
unsigned LineNumber, DIType *Ty, bool isLocalToUnit, MDNode *Decl,
|
|
|
|
uint32_t AlignInBits) {
|
2015-03-04 01:24:31 +08:00
|
|
|
checkGlobalVariableScope(Context);
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIGlobalVariable::getTemporary(
|
|
|
|
VMContext, cast_or_null<DIScope>(Context), Name, LinkageName, F,
|
2016-12-20 10:09:43 +08:00
|
|
|
LineNumber, Ty, isLocalToUnit, false,
|
2016-10-20 08:13:12 +08:00
|
|
|
cast_or_null<DIDerivedType>(Decl), AlignInBits)
|
2015-04-16 09:37:00 +08:00
|
|
|
.release();
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2015-08-01 01:55:53 +08:00
|
|
|
static DILocalVariable *createLocalVariable(
|
|
|
|
LLVMContext &VMContext,
|
2016-04-21 04:14:09 +08:00
|
|
|
DenseMap<MDNode *, SmallVector<TrackingMDNodeRef, 1>> &PreservedVariables,
|
2015-08-01 01:55:53 +08:00
|
|
|
DIScope *Scope, StringRef Name, unsigned ArgNo, DIFile *File,
|
2016-10-20 08:13:12 +08:00
|
|
|
unsigned LineNo, DIType *Ty, bool AlwaysPreserve, DINode::DIFlags Flags,
|
|
|
|
uint32_t AlignInBits) {
|
2015-03-04 01:24:31 +08:00
|
|
|
// FIXME: Why getNonCompileUnitScope()?
|
|
|
|
// FIXME: Why is "!Context" okay here?
|
2015-07-11 07:26:02 +08:00
|
|
|
// FIXME: Why doesn't this check for a subprogram or lexical block (AFAICT
|
2015-03-04 01:24:31 +08:00
|
|
|
// the only valid scopes)?
|
2015-04-30 00:38:44 +08:00
|
|
|
DIScope *Context = getNonCompileUnitScope(Scope);
|
2015-03-04 01:24:31 +08:00
|
|
|
|
2015-08-01 02:58:39 +08:00
|
|
|
auto *Node =
|
|
|
|
DILocalVariable::get(VMContext, cast_or_null<DILocalScope>(Context), Name,
|
2016-10-20 08:13:12 +08:00
|
|
|
File, LineNo, Ty, ArgNo, Flags, AlignInBits);
|
2010-12-08 07:58:00 +08:00
|
|
|
if (AlwaysPreserve) {
|
2015-07-11 07:26:02 +08:00
|
|
|
// The optimizer may remove local variables. If there is an interest
|
2010-12-08 07:58:00 +08:00
|
|
|
// to preserve variable info in such situation then stash it in a
|
|
|
|
// named mdnode.
|
2015-04-30 00:38:44 +08:00
|
|
|
DISubprogram *Fn = getDISubprogram(Scope);
|
2014-10-16 00:11:41 +08:00
|
|
|
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);
|
2010-12-08 07:58:00 +08:00
|
|
|
}
|
2015-03-04 01:24:31 +08:00
|
|
|
return Node;
|
2010-12-08 07:58:00 +08:00
|
|
|
}
|
|
|
|
|
2015-08-01 01:55:53 +08:00
|
|
|
DILocalVariable *DIBuilder::createAutoVariable(DIScope *Scope, StringRef Name,
|
|
|
|
DIFile *File, unsigned LineNo,
|
|
|
|
DIType *Ty, bool AlwaysPreserve,
|
2016-10-20 08:13:12 +08:00
|
|
|
DINode::DIFlags Flags,
|
|
|
|
uint32_t AlignInBits) {
|
2015-08-01 01:55:53 +08:00
|
|
|
return createLocalVariable(VMContext, PreservedVariables, Scope, Name,
|
|
|
|
/* ArgNo */ 0, File, LineNo, Ty, AlwaysPreserve,
|
2016-10-20 08:13:12 +08:00
|
|
|
Flags, AlignInBits);
|
2015-08-01 01:55:53 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
DILocalVariable *DIBuilder::createParameterVariable(
|
|
|
|
DIScope *Scope, StringRef Name, unsigned ArgNo, DIFile *File,
|
2016-09-06 18:46:28 +08:00
|
|
|
unsigned LineNo, DIType *Ty, bool AlwaysPreserve, DINode::DIFlags Flags) {
|
2015-08-01 01:55:53 +08:00
|
|
|
assert(ArgNo && "Expected non-zero argument number for parameter");
|
|
|
|
return createLocalVariable(VMContext, PreservedVariables, Scope, Name, ArgNo,
|
2016-10-20 08:13:12 +08:00
|
|
|
File, LineNo, Ty, AlwaysPreserve, Flags,
|
|
|
|
/* AlignInBits */0);
|
2015-08-01 01:55:53 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIExpression *DIBuilder::createExpression(ArrayRef<uint64_t> Addr) {
|
|
|
|
return DIExpression::get(VMContext, Addr);
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DIExpression *DIBuilder::createExpression(ArrayRef<int64_t> Signed) {
|
2015-02-10 06:13:27 +08:00
|
|
|
// TODO: Remove the callers of this signed version and delete.
|
|
|
|
SmallVector<uint64_t, 8> Addr(Signed.begin(), Signed.end());
|
|
|
|
return createExpression(Addr);
|
|
|
|
}
|
|
|
|
|
2016-12-06 02:04:47 +08:00
|
|
|
DIExpression *DIBuilder::createFragmentExpression(unsigned OffsetInBytes,
|
2015-04-30 00:38:44 +08:00
|
|
|
unsigned SizeInBytes) {
|
2016-12-06 02:04:47 +08:00
|
|
|
uint64_t Addr[] = {dwarf::DW_OP_LLVM_fragment, OffsetInBytes, SizeInBytes};
|
2015-04-30 00:38:44 +08:00
|
|
|
return DIExpression::get(VMContext, Addr);
|
2014-10-02 05:32:12 +08:00
|
|
|
}
|
|
|
|
|
2015-08-27 06:50:16 +08:00
|
|
|
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)...);
|
|
|
|
}
|
|
|
|
|
2015-11-06 06:03:56 +08:00
|
|
|
DISubprogram *DIBuilder::createFunction(
|
|
|
|
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *File,
|
|
|
|
unsigned LineNo, DISubroutineType *Ty, bool isLocalToUnit,
|
2016-09-07 01:03:02 +08:00
|
|
|
bool isDefinition, unsigned ScopeLine, DINode::DIFlags Flags,
|
|
|
|
bool isOptimized, DITemplateParameterArray TParams, DISubprogram *Decl) {
|
2016-04-15 23:57:41 +08:00
|
|
|
auto *Node = getSubprogram(
|
|
|
|
/* IsDistinct = */ isDefinition, VMContext,
|
2016-04-24 05:08:00 +08:00
|
|
|
getNonCompileUnitScope(Context), Name, LinkageName, File, LineNo, Ty,
|
2016-07-01 10:41:21 +08:00
|
|
|
isLocalToUnit, isDefinition, ScopeLine, nullptr, 0, 0, 0, Flags,
|
|
|
|
isOptimized, isDefinition ? CUNode : nullptr, TParams, Decl,
|
2016-04-15 23:57:41 +08:00
|
|
|
MDTuple::getTemporary(VMContext, None).release());
|
2015-03-04 01:24:31 +08:00
|
|
|
|
|
|
|
if (isDefinition)
|
|
|
|
AllSubprograms.push_back(Node);
|
|
|
|
trackIfUnresolved(Node);
|
|
|
|
return Node;
|
2014-09-17 17:28:34 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DISubprogram *DIBuilder::createTempFunctionFwdDecl(
|
|
|
|
DIScope *Context, StringRef Name, StringRef LinkageName, DIFile *File,
|
|
|
|
unsigned LineNo, DISubroutineType *Ty, bool isLocalToUnit,
|
2016-09-07 01:03:02 +08:00
|
|
|
bool isDefinition, unsigned ScopeLine, DINode::DIFlags Flags,
|
|
|
|
bool isOptimized, DITemplateParameterArray TParams, DISubprogram *Decl) {
|
2015-04-30 00:38:44 +08:00
|
|
|
return DISubprogram::getTemporary(
|
2016-04-24 05:08:00 +08:00
|
|
|
VMContext, getNonCompileUnitScope(Context), Name, LinkageName,
|
|
|
|
File, LineNo, Ty, isLocalToUnit, isDefinition, ScopeLine, nullptr,
|
2016-07-01 10:41:21 +08:00
|
|
|
0, 0, 0, Flags, isOptimized, isDefinition ? CUNode : nullptr,
|
|
|
|
TParams, Decl, nullptr)
|
2015-04-30 00:38:44 +08:00
|
|
|
.release();
|
|
|
|
}
|
|
|
|
|
2016-09-07 01:03:02 +08:00
|
|
|
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) {
|
2013-10-16 07:31:36 +08:00
|
|
|
assert(getNonCompileUnitScope(Context) &&
|
|
|
|
"Methods should have both a Context and a context that isn't "
|
|
|
|
"the compile unit.");
|
2015-03-04 01:24:31 +08:00
|
|
|
// FIXME: Do we want to use different scope/lines?
|
2015-11-06 06:03:56 +08:00
|
|
|
auto *SP = getSubprogram(
|
2016-04-24 05:08:00 +08:00
|
|
|
/* IsDistinct = */ isDefinition, VMContext, cast<DIScope>(Context), Name,
|
|
|
|
LinkageName, F, LineNo, Ty, isLocalToUnit, isDefinition, LineNo,
|
2016-07-01 10:41:21 +08:00
|
|
|
VTableHolder, VK, VIndex, ThisAdjustment, Flags, isOptimized,
|
2016-04-24 05:08:00 +08:00
|
|
|
isDefinition ? CUNode : nullptr, TParams, nullptr, nullptr);
|
2015-03-04 01:24:31 +08:00
|
|
|
|
2013-02-18 15:10:22 +08:00
|
|
|
if (isDefinition)
|
2015-04-07 07:18:49 +08:00
|
|
|
AllSubprograms.push_back(SP);
|
|
|
|
trackIfUnresolved(SP);
|
|
|
|
return SP;
|
2010-12-09 04:42:44 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DINamespace *DIBuilder::createNameSpace(DIScope *Scope, StringRef Name,
|
2016-11-04 03:42:02 +08:00
|
|
|
DIFile *File, unsigned LineNo,
|
|
|
|
bool ExportSymbols) {
|
2015-04-30 00:38:44 +08:00
|
|
|
return DINamespace::get(VMContext, getNonCompileUnitScope(Scope), File, Name,
|
2016-11-04 03:42:02 +08:00
|
|
|
LineNo, ExportSymbols);
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2015-06-30 07:03:47 +08:00
|
|
|
DIModule *DIBuilder::createModule(DIScope *Scope, StringRef Name,
|
|
|
|
StringRef ConfigurationMacros,
|
|
|
|
StringRef IncludePath,
|
|
|
|
StringRef ISysRoot) {
|
|
|
|
return DIModule::get(VMContext, getNonCompileUnitScope(Scope), Name,
|
|
|
|
ConfigurationMacros, IncludePath, ISysRoot);
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DILexicalBlockFile *DIBuilder::createLexicalBlockFile(DIScope *Scope,
|
|
|
|
DIFile *File,
|
|
|
|
unsigned Discriminator) {
|
|
|
|
return DILexicalBlockFile::get(VMContext, Scope, File, Discriminator);
|
2011-10-12 06:59:11 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
DILexicalBlock *DIBuilder::createLexicalBlock(DIScope *Scope, DIFile *File,
|
|
|
|
unsigned Line, unsigned Col) {
|
2015-03-04 01:24:31 +08:00
|
|
|
// Make these distinct, to avoid merging two lexical blocks on the same
|
|
|
|
// file/line/column.
|
2015-04-30 00:38:44 +08:00
|
|
|
return DILexicalBlock::getDistinct(VMContext, getNonCompileUnitScope(Scope),
|
2015-04-16 07:19:27 +08:00
|
|
|
File, Line, Col);
|
2010-12-08 09:50:15 +08:00
|
|
|
}
|
|
|
|
|
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));
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
static Instruction *withDebugLoc(Instruction *I, const DILocation *DL) {
|
|
|
|
I->setDebugLoc(const_cast<DILocation *>(DL));
|
2015-04-16 05:18:07 +08:00
|
|
|
return I;
|
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
|
|
|
|
DIExpression *Expr, const DILocation *DL,
|
2010-12-08 07:25:47 +08:00
|
|
|
Instruction *InsertBefore) {
|
2015-04-30 00:38:44 +08:00
|
|
|
assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.declare");
|
2015-04-16 05:18:07 +08:00
|
|
|
assert(DL && "Expected debug loc");
|
|
|
|
assert(DL->getScope()->getSubprogram() ==
|
|
|
|
VarInfo->getScope()->getSubprogram() &&
|
|
|
|
"Expected matching subprograms");
|
2010-12-08 07:25:47 +08:00
|
|
|
if (!DeclareFn)
|
|
|
|
DeclareFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_declare);
|
|
|
|
|
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)};
|
2015-04-16 05:18:07 +08:00
|
|
|
return withDebugLoc(CallInst::Create(DeclareFn, Args, "", InsertBefore), DL);
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
Instruction *DIBuilder::insertDeclare(Value *Storage, DILocalVariable *VarInfo,
|
|
|
|
DIExpression *Expr, const DILocation *DL,
|
2010-12-08 07:25:47 +08:00
|
|
|
BasicBlock *InsertAtEnd) {
|
2015-04-30 00:38:44 +08:00
|
|
|
assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.declare");
|
2015-04-16 05:18:07 +08:00
|
|
|
assert(DL && "Expected debug loc");
|
|
|
|
assert(DL->getScope()->getSubprogram() ==
|
|
|
|
VarInfo->getScope()->getSubprogram() &&
|
|
|
|
"Expected matching subprograms");
|
2010-12-08 07:25:47 +08:00
|
|
|
if (!DeclareFn)
|
|
|
|
DeclareFn = Intrinsic::getDeclaration(&M, Intrinsic::dbg_declare);
|
|
|
|
|
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)};
|
2010-12-08 07:25:47 +08:00
|
|
|
|
|
|
|
// If this block already has a terminator then insert this intrinsic
|
|
|
|
// before the terminator.
|
|
|
|
if (TerminatorInst *T = InsertAtEnd->getTerminator())
|
2015-04-16 05:18:07 +08:00
|
|
|
return withDebugLoc(CallInst::Create(DeclareFn, Args, "", T), DL);
|
|
|
|
return withDebugLoc(CallInst::Create(DeclareFn, Args, "", InsertAtEnd), DL);
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2011-02-23 02:56:12 +08:00
|
|
|
Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
|
2015-04-30 00:38:44 +08:00
|
|
|
DILocalVariable *VarInfo,
|
|
|
|
DIExpression *Expr,
|
|
|
|
const DILocation *DL,
|
2010-12-08 07:25:47 +08:00
|
|
|
Instruction *InsertBefore) {
|
|
|
|
assert(V && "no value passed to dbg.value");
|
2015-04-30 00:38:44 +08:00
|
|
|
assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.value");
|
2015-04-16 05:18:07 +08:00
|
|
|
assert(DL && "Expected debug loc");
|
|
|
|
assert(DL->getScope()->getSubprogram() ==
|
|
|
|
VarInfo->getScope()->getSubprogram() &&
|
|
|
|
"Expected matching subprograms");
|
2010-12-08 07:25:47 +08:00
|
|
|
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),
|
|
|
|
ConstantInt::get(Type::getInt64Ty(VMContext), Offset),
|
|
|
|
MetadataAsValue::get(VMContext, VarInfo),
|
|
|
|
MetadataAsValue::get(VMContext, Expr)};
|
2015-04-16 05:18:07 +08:00
|
|
|
return withDebugLoc(CallInst::Create(ValueFn, Args, "", InsertBefore), DL);
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
|
|
|
|
2011-02-23 02:56:12 +08:00
|
|
|
Instruction *DIBuilder::insertDbgValueIntrinsic(Value *V, uint64_t Offset,
|
2015-04-30 00:38:44 +08:00
|
|
|
DILocalVariable *VarInfo,
|
|
|
|
DIExpression *Expr,
|
|
|
|
const DILocation *DL,
|
2010-12-08 07:25:47 +08:00
|
|
|
BasicBlock *InsertAtEnd) {
|
|
|
|
assert(V && "no value passed to dbg.value");
|
2015-04-30 00:38:44 +08:00
|
|
|
assert(VarInfo && "empty or invalid DILocalVariable* passed to dbg.value");
|
2015-04-16 05:18:07 +08:00
|
|
|
assert(DL && "Expected debug loc");
|
|
|
|
assert(DL->getScope()->getSubprogram() ==
|
|
|
|
VarInfo->getScope()->getSubprogram() &&
|
|
|
|
"Expected matching subprograms");
|
2010-12-08 07:25:47 +08:00
|
|
|
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),
|
|
|
|
ConstantInt::get(Type::getInt64Ty(VMContext), Offset),
|
|
|
|
MetadataAsValue::get(VMContext, VarInfo),
|
|
|
|
MetadataAsValue::get(VMContext, Expr)};
|
2015-04-16 05:18:07 +08:00
|
|
|
|
|
|
|
return withDebugLoc(CallInst::Create(ValueFn, Args, "", InsertAtEnd), DL);
|
2010-12-08 07:25:47 +08:00
|
|
|
}
|
2014-12-18 08:46:16 +08:00
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
void DIBuilder::replaceVTableHolder(DICompositeType *&T,
|
|
|
|
DICompositeType *VTableHolder) {
|
2015-04-07 12:12:02 +08:00
|
|
|
{
|
2015-04-30 00:38:44 +08:00
|
|
|
TypedTrackingMDRef<DICompositeType> N(T);
|
2016-04-24 05:08:00 +08:00
|
|
|
N->replaceVTableHolder(VTableHolder);
|
2015-04-07 12:12:02 +08:00
|
|
|
T = N.get();
|
|
|
|
}
|
2014-12-18 08:46:16 +08:00
|
|
|
|
|
|
|
// If this didn't create a self-reference, just return.
|
|
|
|
if (T != VTableHolder)
|
|
|
|
return;
|
|
|
|
|
2015-02-12 01:45:10 +08:00
|
|
|
// 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);
|
2014-12-18 08:46:16 +08:00
|
|
|
}
|
|
|
|
|
2015-04-30 00:38:44 +08:00
|
|
|
void DIBuilder::replaceArrays(DICompositeType *&T, DINodeArray Elements,
|
|
|
|
DINodeArray TParams) {
|
2015-04-07 12:12:02 +08:00
|
|
|
{
|
2015-04-30 00:38:44 +08:00
|
|
|
TypedTrackingMDRef<DICompositeType> N(T);
|
2015-04-07 12:12:02 +08:00
|
|
|
if (Elements)
|
2015-04-07 12:14:33 +08:00
|
|
|
N->replaceElements(Elements);
|
2015-04-07 12:12:02 +08:00
|
|
|
if (TParams)
|
2015-04-30 00:38:44 +08:00
|
|
|
N->replaceTemplateParams(DITemplateParameterArray(TParams));
|
2015-04-07 12:12:02 +08:00
|
|
|
T = N.get();
|
|
|
|
}
|
2014-12-18 08:46:16 +08:00
|
|
|
|
|
|
|
// If T isn't resolved, there's no problem.
|
|
|
|
if (!T->isResolved())
|
|
|
|
return;
|
|
|
|
|
2015-07-11 07:26:02 +08:00
|
|
|
// If T is resolved, it may be due to a self-reference cycle. Track the
|
2014-12-18 08:46:16 +08:00
|
|
|
// arrays explicitly if they're unresolved, or else the cycles will be
|
|
|
|
// orphaned.
|
|
|
|
if (Elements)
|
2015-04-08 00:50:39 +08:00
|
|
|
trackIfUnresolved(Elements.get());
|
2014-12-18 08:46:16 +08:00
|
|
|
if (TParams)
|
2015-04-08 00:50:39 +08:00
|
|
|
trackIfUnresolved(TParams.get());
|
2014-12-18 08:46:16 +08:00
|
|
|
}
|