llvm-project/llvm/lib/Bitcode/Writer/ValueEnumerator.h

304 lines
10 KiB
C
Raw Normal View History

//===- Bitcode/Writer/ValueEnumerator.h - Number values ---------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This class gives values and types Unique ID's.
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_LIB_BITCODE_WRITER_VALUEENUMERATOR_H
#define LLVM_LIB_BITCODE_WRITER_VALUEENUMERATOR_H
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/UniqueVector.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/UseListOrder.h"
#include <cassert>
#include <cstdint>
#include <utility>
#include <vector>
namespace llvm {
class BasicBlock;
class Comdat;
class Function;
class Instruction;
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
class LocalAsMetadata;
class MDNode;
class Metadata;
class Module;
class NamedMDNode;
class raw_ostream;
class Type;
class Value;
class ValueSymbolTable;
class ValueEnumerator {
public:
using TypeList = std::vector<Type *>;
// For each value, we remember its Value* and occurrence frequency.
using ValueList = std::vector<std::pair<const Value *, unsigned>>;
/// Attribute groups as encoded in bitcode are almost AttributeSets, but they
/// include the AttributeList index, so we have to track that in our map.
using IndexAndAttrSet = std::pair<unsigned, AttributeSet>;
UseListOrderStack UseListOrders;
private:
using TypeMapType = DenseMap<Type *, unsigned>;
TypeMapType TypeMap;
TypeList Types;
using ValueMapType = DenseMap<const Value *, unsigned>;
ValueMapType ValueMap;
ValueList Values;
using ComdatSetType = UniqueVector<const Comdat *>;
ComdatSetType Comdats;
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
std::vector<const Metadata *> MDs;
2016-04-02 23:22:57 +08:00
std::vector<const Metadata *> FunctionMDs;
/// Index of information about a piece of metadata.
struct MDIndex {
unsigned F = 0; ///< The ID of the function for this metadata, if any.
unsigned ID = 0; ///< The implicit ID of this metadata in bitcode.
MDIndex() = default;
explicit MDIndex(unsigned F) : F(F) {}
/// Check if this has a function tag, and it's different from NewF.
bool hasDifferentFunction(unsigned NewF) const { return F && F != NewF; }
/// Fetch the MD this references out of the given metadata array.
const Metadata *get(ArrayRef<const Metadata *> MDs) const {
assert(ID && "Expected non-zero ID");
assert(ID <= MDs.size() && "Expected valid ID");
return MDs[ID - 1];
}
};
using MetadataMapType = DenseMap<const Metadata *, MDIndex>;
MetadataMapType MetadataMap;
2016-04-02 23:22:57 +08:00
/// Range of metadata IDs, as a half-open range.
struct MDRange {
unsigned First = 0;
unsigned Last = 0;
/// Number of strings in the prefix of the metadata range.
unsigned NumStrings = 0;
MDRange() = default;
2016-04-02 23:22:57 +08:00
explicit MDRange(unsigned First) : First(First) {}
};
SmallDenseMap<unsigned, MDRange, 1> FunctionMDInfo;
bool ShouldPreserveUseListOrder;
using AttributeGroupMapType = DenseMap<IndexAndAttrSet, unsigned>;
AttributeGroupMapType AttributeGroupMap;
std::vector<IndexAndAttrSet> AttributeGroups;
using AttributeListMapType = DenseMap<AttributeList, unsigned>;
AttributeListMapType AttributeListMap;
std::vector<AttributeList> AttributeLists;
/// GlobalBasicBlockIDs - This map memoizes the basic block ID's referenced by
/// the "getGlobalBasicBlockID" method.
mutable DenseMap<const BasicBlock*, unsigned> GlobalBasicBlockIDs;
using InstructionMapType = DenseMap<const Instruction *, unsigned>;
InstructionMapType InstructionMap;
unsigned InstructionCount;
/// BasicBlocks - This contains all the basic blocks for the currently
/// incorporated function. Their reverse mapping is stored in ValueMap.
std::vector<const BasicBlock*> BasicBlocks;
/// When a function is incorporated, this is the size of the Values list
/// before incorporation.
unsigned NumModuleValues;
/// When a function is incorporated, this is the size of the Metadatas list
/// before incorporation.
unsigned NumModuleMDs = 0;
unsigned NumMDStrings = 0;
unsigned FirstFuncConstantID;
unsigned FirstInstID;
public:
ValueEnumerator(const Module &M, bool ShouldPreserveUseListOrder);
ValueEnumerator(const ValueEnumerator &) = delete;
ValueEnumerator &operator=(const ValueEnumerator &) = delete;
void dump() const;
void print(raw_ostream &OS, const ValueMapType &Map, const char *Name) const;
IR: Split Metadata from Value Split `Metadata` away from the `Value` class hierarchy, as part of PR21532. Assembly and bitcode changes are in the wings, but this is the bulk of the change for the IR C++ API. I have a follow-up patch prepared for `clang`. If this breaks other sub-projects, I apologize in advance :(. Help me compile it on Darwin I'll try to fix it. FWIW, the errors should be easy to fix, so it may be simpler to just fix it yourself. This breaks the build for all metadata-related code that's out-of-tree. Rest assured the transition is mechanical and the compiler should catch almost all of the problems. Here's a quick guide for updating your code: - `Metadata` is the root of a class hierarchy with three main classes: `MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from the `Value` class hierarchy. It is typeless -- i.e., instances do *not* have a `Type`. - `MDNode`'s operands are all `Metadata *` (instead of `Value *`). - `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively. If you're referring solely to resolved `MDNode`s -- post graph construction -- just use `MDNode*`. - `MDNode` (and the rest of `Metadata`) have only limited support for `replaceAllUsesWith()`. As long as an `MDNode` is pointing at a forward declaration -- the result of `MDNode::getTemporary()` -- it maintains a side map of its uses and can RAUW itself. Once the forward declarations are fully resolved RAUW support is dropped on the ground. This means that uniquing collisions on changing operands cause nodes to become "distinct". (This already happened fairly commonly, whenever an operand went to null.) If you're constructing complex (non self-reference) `MDNode` cycles, you need to call `MDNode::resolveCycles()` on each node (or on a top-level node that somehow references all of the nodes). Also, don't do that. Metadata cycles (and the RAUW machinery needed to construct them) are expensive. - An `MDNode` can only refer to a `Constant` through a bridge called `ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`). As a side effect, accessing an operand of an `MDNode` that is known to be, e.g., `ConstantInt`, takes three steps: first, cast from `Metadata` to `ConstantAsMetadata`; second, extract the `Constant`; third, cast down to `ConstantInt`. The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have metadata schema owners transition away from using `Constant`s when the type isn't important (and they don't care about referring to `GlobalValue`s). In the meantime, I've added transitional API to the `mdconst` namespace that matches semantics with the old code, in order to avoid adding the error-prone three-step equivalent to every call site. If your old code was: MDNode *N = foo(); bar(isa <ConstantInt>(N->getOperand(0))); baz(cast <ConstantInt>(N->getOperand(1))); bak(cast_or_null <ConstantInt>(N->getOperand(2))); bat(dyn_cast <ConstantInt>(N->getOperand(3))); bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4))); you can trivially match its semantics with: MDNode *N = foo(); bar(mdconst::hasa <ConstantInt>(N->getOperand(0))); baz(mdconst::extract <ConstantInt>(N->getOperand(1))); bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2))); bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3))); bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4))); and when you transition your metadata schema to `MDInt`: MDNode *N = foo(); bar(isa <MDInt>(N->getOperand(0))); baz(cast <MDInt>(N->getOperand(1))); bak(cast_or_null <MDInt>(N->getOperand(2))); bat(dyn_cast <MDInt>(N->getOperand(3))); bay(dyn_cast_or_null<MDInt>(N->getOperand(4))); - A `CallInst` -- specifically, intrinsic instructions -- can refer to metadata through a bridge called `MetadataAsValue`. This is a subclass of `Value` where `getType()->isMetadataTy()`. `MetadataAsValue` is the *only* class that can legally refer to a `LocalAsMetadata`, which is a bridged form of non-`Constant` values like `Argument` and `Instruction`. It can also refer to any other `Metadata` subclass. (I'll break all your testcases in a follow-up commit, when I propagate this change to assembly.) llvm-svn: 223802
2014-12-10 02:38:53 +08:00
void print(raw_ostream &OS, const MetadataMapType &Map,
const char *Name) const;
unsigned getValueID(const Value *V) const;
unsigned getMetadataID(const Metadata *MD) const {
auto ID = getMetadataOrNullID(MD);
assert(ID != 0 && "Metadata not in slotcalculator!");
return ID - 1;
}
unsigned getMetadataOrNullID(const Metadata *MD) const {
2016-04-02 23:22:57 +08:00
return MetadataMap.lookup(MD).ID;
}
unsigned numMDs() const { return MDs.size(); }
bool shouldPreserveUseListOrder() const { return ShouldPreserveUseListOrder; }
unsigned getTypeID(Type *T) const {
TypeMapType::const_iterator I = TypeMap.find(T);
assert(I != TypeMap.end() && "Type not in ValueEnumerator!");
return I->second-1;
}
unsigned getInstructionID(const Instruction *I) const;
void setInstructionID(const Instruction *I);
unsigned getAttributeListID(AttributeList PAL) const {
if (PAL.isEmpty()) return 0; // Null maps to zero.
AttributeListMapType::const_iterator I = AttributeListMap.find(PAL);
assert(I != AttributeListMap.end() && "Attribute not in ValueEnumerator!");
return I->second;
}
unsigned getAttributeGroupID(IndexAndAttrSet Group) const {
if (!Group.second.hasAttributes())
return 0; // Null maps to zero.
AttributeGroupMapType::const_iterator I = AttributeGroupMap.find(Group);
assert(I != AttributeGroupMap.end() && "Attribute not in ValueEnumerator!");
return I->second;
}
/// getFunctionConstantRange - Return the range of values that corresponds to
/// function-local constants.
void getFunctionConstantRange(unsigned &Start, unsigned &End) const {
Start = FirstFuncConstantID;
End = FirstInstID;
}
const ValueList &getValues() const { return Values; }
/// Check whether the current block has any metadata to emit.
bool hasMDs() const { return NumModuleMDs < MDs.size(); }
/// Get the MDString metadata for this block.
ArrayRef<const Metadata *> getMDStrings() const {
return makeArrayRef(MDs).slice(NumModuleMDs, NumMDStrings);
}
/// Get the non-MDString metadata for this block.
ArrayRef<const Metadata *> getNonMDStrings() const {
return makeArrayRef(MDs).slice(NumModuleMDs).slice(NumMDStrings);
}
const TypeList &getTypes() const { return Types; }
const std::vector<const BasicBlock*> &getBasicBlocks() const {
return BasicBlocks;
}
const std::vector<AttributeList> &getAttributeLists() const { return AttributeLists; }
const std::vector<IndexAndAttrSet> &getAttributeGroups() const {
return AttributeGroups;
}
const ComdatSetType &getComdats() const { return Comdats; }
unsigned getComdatID(const Comdat *C) const;
/// getGlobalBasicBlockID - This returns the function-specific ID for the
/// specified basic block. This is relatively expensive information, so it
/// should only be used by rare constructs such as address-of-label.
unsigned getGlobalBasicBlockID(const BasicBlock *BB) const;
/// incorporateFunction/purgeFunction - If you'd like to deal with a function,
/// use these two methods to get its data into the ValueEnumerator!
void incorporateFunction(const Function &F);
void purgeFunction();
uint64_t computeBitsRequiredForTypeIndicies() const;
private:
void OptimizeConstants(unsigned CstStart, unsigned CstEnd);
2016-04-02 23:22:57 +08:00
/// Reorder the reachable metadata.
///
/// This is not just an optimization, but is mandatory for emitting MDString
/// correctly.
void organizeMetadata();
2016-04-02 23:22:57 +08:00
/// Drop the function tag from the transitive operands of the given node.
void dropFunctionFromMetadata(MetadataMapType::value_type &FirstMD);
2016-04-02 23:22:57 +08:00
/// Incorporate the function metadata.
///
/// This should be called before enumerating LocalAsMetadata for the
/// function.
void incorporateFunctionMetadata(const Function &F);
ValueMapper/Enumerator: Clean up code in post-order traversals, NFC Re-layer the functions in the new (i.e., newly correct) post-order traversals in ValueEnumerator (r266947) and ValueMapper (r266949). Instead of adding a node to the worklist in a helper function and returning a flag to say what happened, return the node itself. This makes the code way cleaner: the worklist is local to the main function, there is no flag for an early loop exit (since we can cleanly bury the loop), and it's perfectly clear when pointers into the worklist might be invalidated. I'm fixing both algorithms in the same commit to avoid repeating the commit message; if you take the time to understand one the other should be easy. The diff itself isn't entirely obvious since the traversals have some noise (i.e., things to do), but here's the high-level change: auto helper = [&WL](T *Op) { auto helper = [](T **&I, T **E) { => while (I != E) { if (shouldVisit(Op)) { T *Op = *I++; WL.push(Op, Op->begin()); if (shouldVisit(Op)) { return true; return Op; } } return false; return nullptr; }; }; => WL.push(S, S->begin()); WL.push(S, S->begin()); while (!empty()) { while (!empty()) { auto *N = WL.top().N; auto *N = WL.top().N; auto *&I = WL.top().I; auto *&I = WL.top().I; bool DidChange = false; while (I != N->end()) if (helper(*I++)) { => if (T *Op = helper(I, N->end()) { DidChange = true; WL.push(Op, Op->begin()); break; continue; } } if (DidChange) continue; POT.push(WL.pop()); => POT.push(WL.pop()); } } Thanks to Mehdi for helping me find a better way to layer this. llvm-svn: 267099
2016-04-22 10:33:06 +08:00
/// Enumerate a single instance of metadata with the given function tag.
///
/// If \c MD has already been enumerated, check that \c F matches its
/// function tag. If not, call \a dropFunctionFromMetadata().
///
/// Otherwise, mark \c MD as visited. Assign it an ID, or just return it if
/// it's an \a MDNode.
const MDNode *enumerateMetadataImpl(unsigned F, const Metadata *MD);
2016-04-02 23:22:57 +08:00
IR: Allow metadata attachments on declarations, and fix lazy loaded metadata issue with globals. This change is motivated by an upcoming change to the metadata representation used for CFI. The indirect function call checker needs type information for external function declarations in order to correctly generate jump table entries for such declarations. We currently associate such type information with declarations using a global metadata node, but I plan [1] to move all such metadata to global object attachments. In bitcode, metadata attachments for function declarations appear in the global metadata block. This seems reasonable to me because I expect metadata attachments on declarations to be uncommon. In the long term I'd also expect this to be the case for CFI, because we'd want to use some specialized bitcode format for this metadata that could be read as part of the ThinLTO thin-link phase, which would mean that it would not appear in the global metadata block. To solve the lazy loaded metadata issue I was seeing with D20147, I use the same bitcode representation for metadata attachments for global variables as I do for function declarations. Since there's a use case for metadata attachments in the global metadata block, we might as well use that representation for global variables as well, at least until we have a mechanism for lazy loading global variables. In the assembly format, the metadata attachments appear after the "declare" keyword in order to avoid a parsing ambiguity. [1] http://lists.llvm.org/pipermail/llvm-dev/2016-June/100462.html Differential Revision: http://reviews.llvm.org/D21052 llvm-svn: 273336
2016-06-22 07:42:48 +08:00
unsigned getMetadataFunctionID(const Function *F) const;
BitcodeWriter: Emit uniqued subgraphs after all distinct nodes Since forward references for uniqued node operands are expensive (and those for distinct node operands are cheap due to DistinctMDOperandPlaceholder), minimize forward references in uniqued node operands. Moreover, guarantee that when a cycle is broken by a distinct node, none of the uniqued nodes have any forward references. In ValueEnumerator::EnumerateMetadata, enumerate uniqued node subgraphs first, delaying distinct nodes until all uniqued nodes have been handled. This guarantees that uniqued nodes only have forward references when there is a uniquing cycle (since r267276 changed ValueEnumerator::organizeMetadata to partition distinct nodes in front of uniqued nodes as a post-pass). Note that a single uniqued subgraph can hit multiple distinct nodes at its leaves. Ideally these would themselves be emitted in post-order, but this commit doesn't attempt that; I think it requires an extra pass through the edges, which I'm not convinced is worth it (since DistinctMDOperandPlaceholder makes forward references quite cheap between distinct nodes). I've added two testcases: - test/Bitcode/mdnodes-distinct-in-post-order.ll is just like test/Bitcode/mdnodes-in-post-order.ll, except with distinct nodes instead of uniqued ones. This confirms that, in the absence of uniqued nodes, distinct nodes are still emitted in post-order. - test/Bitcode/mdnodes-distinct-nodes-break-cycles.ll is the minimal example where a naive post-order traversal would cause one uniqued node to forward-reference another. IOW, it's the motivating test. llvm-svn: 267278
2016-04-23 12:59:22 +08:00
/// Enumerate reachable metadata in (almost) post-order.
///
/// Enumerate all the metadata reachable from MD. We want to minimize the
/// cost of reading bitcode records, and so the primary consideration is that
/// operands of uniqued nodes are resolved before the nodes are read. This
/// avoids re-uniquing them on the context and factors away RAUW support.
///
/// This algorithm guarantees that subgraphs of uniqued nodes are in
/// post-order. Distinct subgraphs reachable only from a single uniqued node
/// will be in post-order.
///
/// \note The relative order of a distinct and uniqued node is irrelevant.
/// \a organizeMetadata() will later partition distinct nodes ahead of
/// uniqued ones.
///{
IR: Allow metadata attachments on declarations, and fix lazy loaded metadata issue with globals. This change is motivated by an upcoming change to the metadata representation used for CFI. The indirect function call checker needs type information for external function declarations in order to correctly generate jump table entries for such declarations. We currently associate such type information with declarations using a global metadata node, but I plan [1] to move all such metadata to global object attachments. In bitcode, metadata attachments for function declarations appear in the global metadata block. This seems reasonable to me because I expect metadata attachments on declarations to be uncommon. In the long term I'd also expect this to be the case for CFI, because we'd want to use some specialized bitcode format for this metadata that could be read as part of the ThinLTO thin-link phase, which would mean that it would not appear in the global metadata block. To solve the lazy loaded metadata issue I was seeing with D20147, I use the same bitcode representation for metadata attachments for global variables as I do for function declarations. Since there's a use case for metadata attachments in the global metadata block, we might as well use that representation for global variables as well, at least until we have a mechanism for lazy loading global variables. In the assembly format, the metadata attachments appear after the "declare" keyword in order to avoid a parsing ambiguity. [1] http://lists.llvm.org/pipermail/llvm-dev/2016-June/100462.html Differential Revision: http://reviews.llvm.org/D21052 llvm-svn: 273336
2016-06-22 07:42:48 +08:00
void EnumerateMetadata(const Function *F, const Metadata *MD);
2016-04-02 23:22:57 +08:00
void EnumerateMetadata(unsigned F, const Metadata *MD);
BitcodeWriter: Emit uniqued subgraphs after all distinct nodes Since forward references for uniqued node operands are expensive (and those for distinct node operands are cheap due to DistinctMDOperandPlaceholder), minimize forward references in uniqued node operands. Moreover, guarantee that when a cycle is broken by a distinct node, none of the uniqued nodes have any forward references. In ValueEnumerator::EnumerateMetadata, enumerate uniqued node subgraphs first, delaying distinct nodes until all uniqued nodes have been handled. This guarantees that uniqued nodes only have forward references when there is a uniquing cycle (since r267276 changed ValueEnumerator::organizeMetadata to partition distinct nodes in front of uniqued nodes as a post-pass). Note that a single uniqued subgraph can hit multiple distinct nodes at its leaves. Ideally these would themselves be emitted in post-order, but this commit doesn't attempt that; I think it requires an extra pass through the edges, which I'm not convinced is worth it (since DistinctMDOperandPlaceholder makes forward references quite cheap between distinct nodes). I've added two testcases: - test/Bitcode/mdnodes-distinct-in-post-order.ll is just like test/Bitcode/mdnodes-in-post-order.ll, except with distinct nodes instead of uniqued ones. This confirms that, in the absence of uniqued nodes, distinct nodes are still emitted in post-order. - test/Bitcode/mdnodes-distinct-nodes-break-cycles.ll is the minimal example where a naive post-order traversal would cause one uniqued node to forward-reference another. IOW, it's the motivating test. llvm-svn: 267278
2016-04-23 12:59:22 +08:00
///}
2016-04-02 23:22:57 +08:00
void EnumerateFunctionLocalMetadata(const Function &F,
const LocalAsMetadata *Local);
void EnumerateFunctionLocalMetadata(unsigned F, const LocalAsMetadata *Local);
void EnumerateNamedMDNode(const NamedMDNode *NMD);
void EnumerateValue(const Value *V);
void EnumerateType(Type *T);
void EnumerateOperandType(const Value *V);
void EnumerateAttributes(AttributeList PAL);
void EnumerateValueSymbolTable(const ValueSymbolTable &ST);
void EnumerateNamedMetadata(const Module &M);
};
} // end namespace llvm
#endif // LLVM_LIB_BITCODE_WRITER_VALUEENUMERATOR_H