llvm-project/llvm/unittests/IR/PassManagerTest.cpp

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

954 lines
36 KiB
C++
Raw Normal View History

//===- llvm/unittest/IR/PassManager.cpp - PassManager tests ---------------===//
//
// 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
//
//===----------------------------------------------------------------------===//
#include "llvm/IR/PassManager.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/AsmParser/Parser.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Module.h"
Add PassManagerImpl.h to hide implementation details ClangBuildAnalyzer results show that a lot of time is spent instantiating AnalysisManager::getResultImpl across the code base: **** Templates that took longest to instantiate: 50445 ms: llvm::AnalysisManager<llvm::Function>::getResultImpl (412 times, avg 122 ms) 47797 ms: llvm::AnalysisManager<llvm::Function>::getResult<llvm::TargetLibraryAnalysis> (389 times, avg 122 ms) 46894 ms: std::tie<const unsigned long long, const bool> (2452 times, avg 19 ms) 43851 ms: llvm::BumpPtrAllocatorImpl<llvm::MallocAllocator, 4096, 4096>::Allocate (3228 times, avg 13 ms) 33911 ms: std::tie<const unsigned int, const unsigned int, const unsigned int, const unsigned int> (897 times, avg 37 ms) 33854 ms: std::tie<const unsigned long long, const unsigned long long> (1897 times, avg 17 ms) 27886 ms: std::basic_string<char, std::char_traits<char>, std::allocator<char> >::basic_string (11156 times, avg 2 ms) I mentioned this result to @chandlerc, and he suggested this direction. AnalysisManager is already explicitly instantiated, and getResultImpl doesn't need to be inlined. Move the definition to an Impl header, and include that header in files that explicitly instantiate AnalysisManager. There are only four (real) IR units: - function - module - loop - cgscc Looking at a specific transform (ArgumentPromotion.cpp), here are three compilations before & after this change: BEFORE: $ for i in $(seq 3) ; do ./ccit.bat ; done peak memory: 258.15MB real: 0m6.297s peak memory: 257.54MB real: 0m5.906s peak memory: 257.47MB real: 0m6.219s AFTER: $ for i in $(seq 3) ; do ./ccit.bat ; done peak memory: 235.35MB real: 0m5.454s peak memory: 234.72MB real: 0m5.235s peak memory: 234.39MB real: 0m5.469s The 20MB of memory saved seems real, and the time improvement seems like it is there. Reviewed By: MaskRay Differential Revision: https://reviews.llvm.org/D73817
2020-02-01 08:05:32 +08:00
#include "llvm/IR/PassManagerImpl.h"
#include "llvm/Passes/StandardInstrumentations.h"
#include "llvm/Support/SourceMgr.h"
#include "llvm/Transforms/Scalar/SimplifyCFG.h"
#include "gtest/gtest.h"
using namespace llvm;
namespace {
class TestFunctionAnalysis : public AnalysisInfoMixin<TestFunctionAnalysis> {
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
public:
struct Result {
Result(int Count) : InstructionCount(Count) {}
int InstructionCount;
2020-01-16 06:30:21 +08:00
bool invalidate(Function &, const PreservedAnalyses &PA,
FunctionAnalysisManager::Invalidator &) {
// Check whether the analysis or all analyses on functions have been
// preserved.
auto PAC = PA.getChecker<TestFunctionAnalysis>();
return !(PAC.preserved() || PAC.preservedSet<AllAnalysesOn<Function>>());
}
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
};
TestFunctionAnalysis(int &Runs) : Runs(Runs) {}
/// Run the analysis pass over the function and return a result.
Result run(Function &F, FunctionAnalysisManager &AM) {
++Runs;
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
int Count = 0;
for (Function::iterator BBI = F.begin(), BBE = F.end(); BBI != BBE; ++BBI)
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
for (BasicBlock::iterator II = BBI->begin(), IE = BBI->end(); II != IE;
++II)
++Count;
return Result(Count);
}
private:
friend AnalysisInfoMixin<TestFunctionAnalysis>;
[PM] Change the static object whose address is used to uniquely identify analyses to have a common type which is enforced rather than using a char object and a `void *` type when used as an identifier. This has a number of advantages. First, it at least helps some of the confusion raised in Justin Lebar's code review of why `void *` was being used everywhere by having a stronger type that connects to documentation about this. However, perhaps more importantly, it addresses a serious issue where the alignment of these pointer-like identifiers was unknown. This made it hard to use them in pointer-like data structures. We were already dodging this in dangerous ways to create the "all analyses" entry. In a subsequent patch I attempted to use these with TinyPtrVector and things fell apart in a very bad way. And it isn't just a compile time or type system issue. Worse than that, the actual alignment of these pointer-like opaque identifiers wasn't guaranteed to be a useful alignment as they were just characters. This change introduces a type to use as the "key" object whose address forms the opaque identifier. This both forces the objects to have proper alignment, and provides type checking that we get it right everywhere. It also makes the types somewhat less mysterious than `void *`. We could go one step further and introduce a truly opaque pointer-like type to return from the `ID()` static function rather than returning `AnalysisKey *`, but that didn't seem to be a clear win so this is just the initial change to get to a reliably typed and aligned object serving is a key for all the analyses. Thanks to Richard Smith and Justin Lebar for helping pick plausible names and avoid making this refactoring many times. =] And thanks to Sean for the super fast review! While here, I've tried to move away from the "PassID" nomenclature entirely as it wasn't really helping and is overloaded with old pass manager constructs. Now we have IDs for analyses, and key objects whose address can be used as IDs. Where possible and clear I've shortened this to just "ID". In a few places I kept "AnalysisID" to make it clear what was being identified. Differential Revision: https://reviews.llvm.org/D27031 llvm-svn: 287783
2016-11-24 01:53:26 +08:00
static AnalysisKey Key;
int &Runs;
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
};
[PM] Change the static object whose address is used to uniquely identify analyses to have a common type which is enforced rather than using a char object and a `void *` type when used as an identifier. This has a number of advantages. First, it at least helps some of the confusion raised in Justin Lebar's code review of why `void *` was being used everywhere by having a stronger type that connects to documentation about this. However, perhaps more importantly, it addresses a serious issue where the alignment of these pointer-like identifiers was unknown. This made it hard to use them in pointer-like data structures. We were already dodging this in dangerous ways to create the "all analyses" entry. In a subsequent patch I attempted to use these with TinyPtrVector and things fell apart in a very bad way. And it isn't just a compile time or type system issue. Worse than that, the actual alignment of these pointer-like opaque identifiers wasn't guaranteed to be a useful alignment as they were just characters. This change introduces a type to use as the "key" object whose address forms the opaque identifier. This both forces the objects to have proper alignment, and provides type checking that we get it right everywhere. It also makes the types somewhat less mysterious than `void *`. We could go one step further and introduce a truly opaque pointer-like type to return from the `ID()` static function rather than returning `AnalysisKey *`, but that didn't seem to be a clear win so this is just the initial change to get to a reliably typed and aligned object serving is a key for all the analyses. Thanks to Richard Smith and Justin Lebar for helping pick plausible names and avoid making this refactoring many times. =] And thanks to Sean for the super fast review! While here, I've tried to move away from the "PassID" nomenclature entirely as it wasn't really helping and is overloaded with old pass manager constructs. Now we have IDs for analyses, and key objects whose address can be used as IDs. Where possible and clear I've shortened this to just "ID". In a few places I kept "AnalysisID" to make it clear what was being identified. Differential Revision: https://reviews.llvm.org/D27031 llvm-svn: 287783
2016-11-24 01:53:26 +08:00
AnalysisKey TestFunctionAnalysis::Key;
class TestModuleAnalysis : public AnalysisInfoMixin<TestModuleAnalysis> {
public:
struct Result {
Result(int Count) : FunctionCount(Count) {}
int FunctionCount;
2020-01-16 06:30:21 +08:00
bool invalidate(Module &, const PreservedAnalyses &PA,
ModuleAnalysisManager::Invalidator &) {
// Check whether the analysis or all analyses on modules have been
// preserved.
auto PAC = PA.getChecker<TestModuleAnalysis>();
return !(PAC.preserved() || PAC.preservedSet<AllAnalysesOn<Module>>());
}
};
TestModuleAnalysis(int &Runs) : Runs(Runs) {}
Result run(Module &M, ModuleAnalysisManager &AM) {
++Runs;
int Count = 0;
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
++Count;
return Result(Count);
}
private:
friend AnalysisInfoMixin<TestModuleAnalysis>;
[PM] Change the static object whose address is used to uniquely identify analyses to have a common type which is enforced rather than using a char object and a `void *` type when used as an identifier. This has a number of advantages. First, it at least helps some of the confusion raised in Justin Lebar's code review of why `void *` was being used everywhere by having a stronger type that connects to documentation about this. However, perhaps more importantly, it addresses a serious issue where the alignment of these pointer-like identifiers was unknown. This made it hard to use them in pointer-like data structures. We were already dodging this in dangerous ways to create the "all analyses" entry. In a subsequent patch I attempted to use these with TinyPtrVector and things fell apart in a very bad way. And it isn't just a compile time or type system issue. Worse than that, the actual alignment of these pointer-like opaque identifiers wasn't guaranteed to be a useful alignment as they were just characters. This change introduces a type to use as the "key" object whose address forms the opaque identifier. This both forces the objects to have proper alignment, and provides type checking that we get it right everywhere. It also makes the types somewhat less mysterious than `void *`. We could go one step further and introduce a truly opaque pointer-like type to return from the `ID()` static function rather than returning `AnalysisKey *`, but that didn't seem to be a clear win so this is just the initial change to get to a reliably typed and aligned object serving is a key for all the analyses. Thanks to Richard Smith and Justin Lebar for helping pick plausible names and avoid making this refactoring many times. =] And thanks to Sean for the super fast review! While here, I've tried to move away from the "PassID" nomenclature entirely as it wasn't really helping and is overloaded with old pass manager constructs. Now we have IDs for analyses, and key objects whose address can be used as IDs. Where possible and clear I've shortened this to just "ID". In a few places I kept "AnalysisID" to make it clear what was being identified. Differential Revision: https://reviews.llvm.org/D27031 llvm-svn: 287783
2016-11-24 01:53:26 +08:00
static AnalysisKey Key;
int &Runs;
};
[PM] Change the static object whose address is used to uniquely identify analyses to have a common type which is enforced rather than using a char object and a `void *` type when used as an identifier. This has a number of advantages. First, it at least helps some of the confusion raised in Justin Lebar's code review of why `void *` was being used everywhere by having a stronger type that connects to documentation about this. However, perhaps more importantly, it addresses a serious issue where the alignment of these pointer-like identifiers was unknown. This made it hard to use them in pointer-like data structures. We were already dodging this in dangerous ways to create the "all analyses" entry. In a subsequent patch I attempted to use these with TinyPtrVector and things fell apart in a very bad way. And it isn't just a compile time or type system issue. Worse than that, the actual alignment of these pointer-like opaque identifiers wasn't guaranteed to be a useful alignment as they were just characters. This change introduces a type to use as the "key" object whose address forms the opaque identifier. This both forces the objects to have proper alignment, and provides type checking that we get it right everywhere. It also makes the types somewhat less mysterious than `void *`. We could go one step further and introduce a truly opaque pointer-like type to return from the `ID()` static function rather than returning `AnalysisKey *`, but that didn't seem to be a clear win so this is just the initial change to get to a reliably typed and aligned object serving is a key for all the analyses. Thanks to Richard Smith and Justin Lebar for helping pick plausible names and avoid making this refactoring many times. =] And thanks to Sean for the super fast review! While here, I've tried to move away from the "PassID" nomenclature entirely as it wasn't really helping and is overloaded with old pass manager constructs. Now we have IDs for analyses, and key objects whose address can be used as IDs. Where possible and clear I've shortened this to just "ID". In a few places I kept "AnalysisID" to make it clear what was being identified. Differential Revision: https://reviews.llvm.org/D27031 llvm-svn: 287783
2016-11-24 01:53:26 +08:00
AnalysisKey TestModuleAnalysis::Key;
struct TestModulePass : PassInfoMixin<TestModulePass> {
TestModulePass(int &RunCount) : RunCount(RunCount) {}
PreservedAnalyses run(Module &M, ModuleAnalysisManager &) {
++RunCount;
[PM] Add the preservation system to the new pass manager. This adds a new set-like type which represents a set of preserved analysis passes. The set is managed via the opaque PassT::ID() void*s. The expected convenience templates for interacting with specific passes are provided. It also supports a symbolic "all" state which is represented by an invalid pointer in the set. This state is nicely saturating as it comes up often. Finally, it supports intersection which is used when finding the set of preserved passes after N different transforms. The pass API is then changed to return the preserved set rather than a bool. This is much more self-documenting than the previous system. Returning "none" is a conservatively correct solution just like returning "true" from todays passes and not marking any passes as preserved. Passes can also be dynamically preserved or not throughout the run of the pass, and whatever gets returned is the binding state. Finally, preserving "all" the passes is allowed for no-op transforms that simply can't harm such things. Finally, the analysis managers are changed to instead of blindly invalidating all of the analyses, invalidate those which were not preserved. This should rig up all of the basic preservation functionality. This also correctly combines the preservation moving up from one IR-layer to the another and the preservation aggregation across N pass runs. Still to go is incrementally correct invalidation and preservation across IR layers incrementally during N pass runs. That will wait until we have a device for even exposing analyses across IR layers. While the core of this change is obvious, I'm not happy with the current testing, so will improve it to cover at least some of the invalidation that I can test easily in a subsequent commit. llvm-svn: 195241
2013-11-20 19:31:50 +08:00
return PreservedAnalyses::none();
}
int &RunCount;
};
struct TestPreservingModulePass : PassInfoMixin<TestPreservingModulePass> {
PreservedAnalyses run(Module &M, ModuleAnalysisManager &) {
return PreservedAnalyses::all();
}
};
struct TestFunctionPass : PassInfoMixin<TestFunctionPass> {
TestFunctionPass(int &RunCount, int &AnalyzedInstrCount,
int &AnalyzedFunctionCount, ModuleAnalysisManager &MAM,
bool OnlyUseCachedResults = false)
: RunCount(RunCount), AnalyzedInstrCount(AnalyzedInstrCount),
AnalyzedFunctionCount(AnalyzedFunctionCount), MAM(MAM),
OnlyUseCachedResults(OnlyUseCachedResults) {}
PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM) {
++RunCount;
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
// Getting a cached result that isn't stateless through the proxy will
// trigger an assert:
// auto &ModuleProxy = AM.getResult<ModuleAnalysisManagerFunctionProxy>(F);
// Use MAM, for the purposes of this unittest.
if (TestModuleAnalysis::Result *TMA =
MAM.getCachedResult<TestModuleAnalysis>(*F.getParent())) {
AnalyzedFunctionCount += TMA->FunctionCount;
}
if (OnlyUseCachedResults) {
// Hack to force the use of the cached interface.
if (TestFunctionAnalysis::Result *AR =
AM.getCachedResult<TestFunctionAnalysis>(F))
AnalyzedInstrCount += AR->InstructionCount;
} else {
// Typical path just runs the analysis as needed.
TestFunctionAnalysis::Result &AR = AM.getResult<TestFunctionAnalysis>(F);
AnalyzedInstrCount += AR.InstructionCount;
}
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
return PreservedAnalyses::all();
}
int &RunCount;
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
int &AnalyzedInstrCount;
int &AnalyzedFunctionCount;
ModuleAnalysisManager &MAM;
bool OnlyUseCachedResults;
};
// A test function pass that invalidates all function analyses for a function
// with a specific name.
struct TestInvalidationFunctionPass
: PassInfoMixin<TestInvalidationFunctionPass> {
TestInvalidationFunctionPass(StringRef FunctionName) : Name(FunctionName) {}
PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM) {
return F.getName() == Name ? PreservedAnalyses::none()
: PreservedAnalyses::all();
}
StringRef Name;
};
std::unique_ptr<Module> parseIR(LLVMContext &Context, const char *IR) {
SMDiagnostic Err;
return parseAssemblyString(IR, Err, Context);
}
class PassManagerTest : public ::testing::Test {
protected:
LLVMContext Context;
std::unique_ptr<Module> M;
public:
PassManagerTest()
: M(parseIR(Context, "define void @f() {\n"
"entry:\n"
" call void @g()\n"
" call void @h()\n"
" ret void\n"
"}\n"
"define void @g() {\n"
" ret void\n"
"}\n"
"define void @h() {\n"
" ret void\n"
"}\n")) {}
};
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
TEST(PreservedAnalysesTest, Basic) {
PreservedAnalyses PA1 = PreservedAnalyses();
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
{
auto PAC = PA1.getChecker<TestFunctionAnalysis>();
EXPECT_FALSE(PAC.preserved());
EXPECT_FALSE(PAC.preservedSet<AllAnalysesOn<Function>>());
}
{
auto PAC = PA1.getChecker<TestModuleAnalysis>();
EXPECT_FALSE(PAC.preserved());
EXPECT_FALSE(PAC.preservedSet<AllAnalysesOn<Module>>());
}
auto PA2 = PreservedAnalyses::none();
{
auto PAC = PA2.getChecker<TestFunctionAnalysis>();
EXPECT_FALSE(PAC.preserved());
EXPECT_FALSE(PAC.preservedSet<AllAnalysesOn<Function>>());
}
auto PA3 = PreservedAnalyses::all();
{
auto PAC = PA3.getChecker<TestFunctionAnalysis>();
EXPECT_TRUE(PAC.preserved());
EXPECT_TRUE(PAC.preservedSet<AllAnalysesOn<Function>>());
}
PreservedAnalyses PA4 = PA1;
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
{
auto PAC = PA4.getChecker<TestFunctionAnalysis>();
EXPECT_FALSE(PAC.preserved());
EXPECT_FALSE(PAC.preservedSet<AllAnalysesOn<Function>>());
}
PA4 = PA3;
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
{
auto PAC = PA4.getChecker<TestFunctionAnalysis>();
EXPECT_TRUE(PAC.preserved());
EXPECT_TRUE(PAC.preservedSet<AllAnalysesOn<Function>>());
}
PA4 = std::move(PA2);
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
{
auto PAC = PA4.getChecker<TestFunctionAnalysis>();
EXPECT_FALSE(PAC.preserved());
EXPECT_FALSE(PAC.preservedSet<AllAnalysesOn<Function>>());
}
auto PA5 = PreservedAnalyses::allInSet<AllAnalysesOn<Function>>();
{
auto PAC = PA5.getChecker<TestFunctionAnalysis>();
EXPECT_FALSE(PAC.preserved());
EXPECT_TRUE(PAC.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(PAC.preservedSet<AllAnalysesOn<Module>>());
}
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
}
TEST(PreservedAnalysesTest, Preserve) {
auto PA = PreservedAnalyses::none();
PA.preserve<TestFunctionAnalysis>();
EXPECT_TRUE(PA.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(PA.getChecker<TestModuleAnalysis>().preserved());
PA.preserve<TestModuleAnalysis>();
EXPECT_TRUE(PA.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_TRUE(PA.getChecker<TestModuleAnalysis>().preserved());
// Redundant calls are fine.
PA.preserve<TestFunctionAnalysis>();
EXPECT_TRUE(PA.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_TRUE(PA.getChecker<TestModuleAnalysis>().preserved());
}
TEST(PreservedAnalysesTest, PreserveSets) {
auto PA = PreservedAnalyses::none();
PA.preserveSet<AllAnalysesOn<Function>>();
EXPECT_TRUE(PA.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(PA.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
PA.preserveSet<AllAnalysesOn<Module>>();
EXPECT_TRUE(PA.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_TRUE(PA.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Mixing is fine.
PA.preserve<TestFunctionAnalysis>();
EXPECT_TRUE(PA.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_TRUE(PA.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Redundant calls are fine.
PA.preserveSet<AllAnalysesOn<Module>>();
EXPECT_TRUE(PA.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_TRUE(PA.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
}
TEST(PreservedAnalysisTest, Intersect) {
// Setup the initial sets.
auto PA1 = PreservedAnalyses::none();
PA1.preserve<TestFunctionAnalysis>();
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
PA1.preserveSet<AllAnalysesOn<Module>>();
auto PA2 = PreservedAnalyses::none();
PA2.preserve<TestFunctionAnalysis>();
PA2.preserveSet<AllAnalysesOn<Function>>();
PA2.preserve<TestModuleAnalysis>();
PA2.preserveSet<AllAnalysesOn<Module>>();
auto PA3 = PreservedAnalyses::none();
PA3.preserve<TestModuleAnalysis>();
PA3.preserveSet<AllAnalysesOn<Function>>();
// Self intersection is a no-op.
auto Intersected = PA1;
Intersected.intersect(PA1);
EXPECT_TRUE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_TRUE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Intersecting with all is a no-op.
Intersected.intersect(PreservedAnalyses::all());
EXPECT_TRUE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_TRUE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Intersecting a narrow set with a more broad set is the narrow set.
Intersected.intersect(PA2);
EXPECT_TRUE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_TRUE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Intersecting a broad set with a more narrow set is the narrow set.
Intersected = PA2;
Intersected.intersect(PA1);
EXPECT_TRUE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_TRUE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Intersecting with empty clears.
Intersected.intersect(PreservedAnalyses::none());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Intersecting non-overlapping clears.
Intersected = PA1;
Intersected.intersect(PA3);
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Intersecting with moves works in when there is storage on both sides.
Intersected = PA1;
auto Tmp = PA2;
Intersected.intersect(std::move(Tmp));
EXPECT_TRUE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_TRUE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// Intersecting with move works for incoming all and existing all.
auto Tmp2 = PreservedAnalyses::all();
Intersected.intersect(std::move(Tmp2));
EXPECT_TRUE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_TRUE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
Intersected = PreservedAnalyses::all();
auto Tmp3 = PA1;
Intersected.intersect(std::move(Tmp3));
EXPECT_TRUE(Intersected.getChecker<TestFunctionAnalysis>().preserved());
EXPECT_FALSE(Intersected.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(Intersected.getChecker<TestModuleAnalysis>().preserved());
EXPECT_TRUE(Intersected.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
}
TEST(PreservedAnalysisTest, Abandon) {
auto PA = PreservedAnalyses::none();
// We can abandon things after they are preserved.
PA.preserve<TestFunctionAnalysis>();
PA.abandon<TestFunctionAnalysis>();
EXPECT_FALSE(PA.getChecker<TestFunctionAnalysis>().preserved());
// Repeated is fine, and abandoning if they were never preserved is fine.
PA.abandon<TestFunctionAnalysis>();
EXPECT_FALSE(PA.getChecker<TestFunctionAnalysis>().preserved());
PA.abandon<TestModuleAnalysis>();
EXPECT_FALSE(PA.getChecker<TestModuleAnalysis>().preserved());
// Even if the sets are preserved, the abandoned analyses' checker won't
// return true for those sets.
PA.preserveSet<AllAnalysesOn<Function>>();
PA.preserveSet<AllAnalysesOn<Module>>();
EXPECT_FALSE(PA.getChecker<TestFunctionAnalysis>()
.preservedSet<AllAnalysesOn<Function>>());
EXPECT_FALSE(PA.getChecker<TestModuleAnalysis>()
.preservedSet<AllAnalysesOn<Module>>());
// But an arbitrary (opaque) analysis will still observe the sets as
// preserved. This also checks that we can use an explicit ID rather than
// a type.
AnalysisKey FakeKey, *FakeID = &FakeKey;
EXPECT_TRUE(PA.getChecker(FakeID).preservedSet<AllAnalysesOn<Function>>());
EXPECT_TRUE(PA.getChecker(FakeID).preservedSet<AllAnalysesOn<Module>>());
}
TEST_F(PassManagerTest, Basic) {
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
FunctionAnalysisManager FAM(/*DebugLogging*/ true);
int FunctionAnalysisRuns = 0;
FAM.registerPass([&] { return TestFunctionAnalysis(FunctionAnalysisRuns); });
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
ModuleAnalysisManager MAM(/*DebugLogging*/ true);
int ModuleAnalysisRuns = 0;
MAM.registerPass([&] { return TestModuleAnalysis(ModuleAnalysisRuns); });
MAM.registerPass([&] { return FunctionAnalysisManagerModuleProxy(FAM); });
FAM.registerPass([&] { return ModuleAnalysisManagerFunctionProxy(MAM); });
[New PM] Introducing PassInstrumentation framework Pass Execution Instrumentation interface enables customizable instrumentation of pass execution, as per "RFC: Pass Execution Instrumentation interface" posted 06/07/2018 on llvm-dev@ The intent is to provide a common machinery to implement all the pass-execution-debugging features like print-before/after, opt-bisect, time-passes etc. Here we get a basic implementation consisting of: * PassInstrumentationCallbacks class that handles registration of callbacks and access to them. * PassInstrumentation class that handles instrumentation-point interfaces that call into PassInstrumentationCallbacks. * Callbacks accept StringRef which is just a name of the Pass right now. There were some ideas to pass an opaque wrapper for the pointer to pass instance, however it appears that pointer does not actually identify the instance (adaptors and managers might have the same address with the pass they govern). Hence it was decided to go simple for now and then later decide on what the proper mental model of identifying a "pass in a phase of pipeline" is. * Callbacks accept llvm::Any serving as a wrapper for const IRUnit*, to remove direct dependencies on different IRUnits (e.g. Analyses). * PassInstrumentationAnalysis analysis is explicitly requested from PassManager through usual AnalysisManager::getResult. All pass managers were updated to run that to get PassInstrumentation object for instrumentation calls. * Using tuples/index_sequence getAnalysisResult helper to extract generic AnalysisManager's extra args out of a generic PassManager's extra args. This is the only way I was able to explicitly run getResult for PassInstrumentationAnalysis out of a generic code like PassManager::run or RepeatedPass::run. TODO: Upon lengthy discussions we agreed to accept this as an initial implementation and then get rid of getAnalysisResult by improving RepeatedPass implementation. * PassBuilder takes PassInstrumentationCallbacks object to pass it further into PassInstrumentationAnalysis. Callbacks registration should be performed directly through PassInstrumentationCallbacks. * new-pm tests updated to account for PassInstrumentationAnalysis being run * Added PassInstrumentation tests to PassBuilderCallbacks unit tests. Other unit tests updated with registration of the now-required PassInstrumentationAnalysis. Made getName helper to return std::string (instead of StringRef initially) to fix asan builtbot failures on CGSCC tests. Reviewers: chandlerc, philip.pfaffe Differential Revision: https://reviews.llvm.org/D47858 llvm-svn: 342664
2018-09-21 01:08:45 +08:00
MAM.registerPass([&] { return PassInstrumentationAnalysis(); });
FAM.registerPass([&] { return PassInstrumentationAnalysis(); });
[PM] Switch analysis managers to be threaded through the run methods rather than the constructors of passes. This simplifies the APIs of passes significantly and removes an error prone pattern where the *same* manager had to be given to every different layer. With the new API the analysis managers themselves will have to be cross connected with proxy analyses that allow a pass at one layer to query for the analysis manager of another layer. The proxy will both expose a handle to the other layer's manager and it will provide the invalidation hooks to ensure things remain consistent across layers. Finally, the outer-most analysis manager has to be passed to the run method of the outer-most pass manager. The rest of the propagation is automatic. I've used SFINAE again to allow passes to completely disregard the analysis manager if they don't need or want to care. This helps keep simple things simple for users of the new pass manager. Also, the system specifically supports passing a null pointer into the outer-most run method if your pass pipeline neither needs nor wants to deal with analyses. I find this of dubious utility as while some *passes* don't care about analysis, I'm not sure there are any real-world users of the pass manager itself that need to avoid even creating an analysis manager. But it is easy to support, so there we go. Finally I renamed the module proxy for the function analysis manager to the more verbose but less confusing name of FunctionAnalysisManagerModuleProxy. I hate this name, but I have no idea what else to name these things. I'm expecting in the fullness of time to potentially have the complete cross product of types at the proxy layer: {Module,SCC,Function,Loop,Region}AnalysisManager{Module,SCC,Function,Loop,Region}Proxy (except for XAnalysisManagerXProxy which doesn't make any sense) This should make it somewhat easier to do the next phases which is to build the upward proxy and get its invalidation correct, as well as to make the invalidation within the Module -> Function mapping pass be more fine grained so as to invalidate fewer fuction analyses. After all of the proxy analyses are done and the invalidation working, I'll finally be able to start working on the next two fun fronts: how to adapt an existing pass to work in both the legacy pass world and the new one, and building the SCC, Loop, and Region counterparts. Fun times! llvm-svn: 195400
2013-11-22 08:43:29 +08:00
ModulePassManager MPM;
Introduce an AnalysisManager which is like a pass manager but with a lot more smarts in it. This is where most of the interesting logic that used to live in the implicit-scheduling-hackery of the old pass manager will live. Like the previous commits, note that this is a very early prototype! I expect substantial changes before this is ready to use. The core of the design is the following: - We have an AnalysisManager which can be used across a series of passes over a module. - The code setting up a pass pipeline registers the analyses available with the manager. - Individual transform passes can check than an analysis manager provides the analyses they require in order to fail-fast. - There is *no* implicit registration or scheduling. - Analysis passes are different from other passes: they produce an analysis result that is cached and made available via the analysis manager. - Cached results are invalidated automatically by the pass managers. - When a transform pass requests an analysis result, either the analysis is run to produce the result or a cached result is provided. There are a few aspects of this design that I *know* will change in subsequent commits: - Currently there is no "preservation" system, that needs to be added. - All of the analysis management should move up to the analysis library. - The analysis management needs to support at least SCC passes. Maybe loop passes. Living in the analysis library will facilitate this. - Need support for analyses which are *both* module and function passes. - Need support for pro-actively running module analyses to have cached results within a function pass manager. - Need a clear design for "immutable" passes. - Need support for requesting cached results when available and not re-running the pass even if that would be necessary. - Need more thorough testing of all of this infrastructure. There are other aspects that I view as open questions I'm hoping to resolve as I iterate a bit on the infrastructure, and especially as I start writing actual passes against this. - Should we have separate management layers for function, module, and SCC analyses? I think "yes", but I'm not yet ready to switch the code. Adding SCC support will likely resolve this definitively. - How should the 'require' functionality work? Should *that* be the only way to request results to ensure that passes always require things? - How should preservation work? - Probably some other things I'm forgetting. =] Look forward to more patches in shorter order now that this is in place. llvm-svn: 194538
2013-11-13 09:12:08 +08:00
// Count the runs over a Function.
int FunctionPassRunCount1 = 0;
int AnalyzedInstrCount1 = 0;
int AnalyzedFunctionCount1 = 0;
{
// Pointless scoped copy to test move assignment.
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
ModulePassManager NestedMPM(/*DebugLogging*/ true);
FunctionPassManager FPM;
{
// Pointless scope to test move assignment.
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
FunctionPassManager NestedFPM(/*DebugLogging*/ true);
NestedFPM.addPass(TestFunctionPass(FunctionPassRunCount1,
AnalyzedInstrCount1,
AnalyzedFunctionCount1, MAM));
FPM = std::move(NestedFPM);
}
NestedMPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
MPM = std::move(NestedMPM);
}
// Count the runs over a module.
int ModulePassRunCount = 0;
MPM.addPass(TestModulePass(ModulePassRunCount));
// Count the runs over a Function in a separate manager.
int FunctionPassRunCount2 = 0;
int AnalyzedInstrCount2 = 0;
int AnalyzedFunctionCount2 = 0;
{
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
FunctionPassManager FPM(/*DebugLogging*/ true);
FPM.addPass(TestFunctionPass(FunctionPassRunCount2, AnalyzedInstrCount2,
AnalyzedFunctionCount2, MAM));
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
}
// A third function pass manager but with only preserving intervening passes
// and with a function pass that invalidates exactly one analysis.
MPM.addPass(TestPreservingModulePass());
int FunctionPassRunCount3 = 0;
int AnalyzedInstrCount3 = 0;
int AnalyzedFunctionCount3 = 0;
{
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
FunctionPassManager FPM(/*DebugLogging*/ true);
FPM.addPass(TestFunctionPass(FunctionPassRunCount3, AnalyzedInstrCount3,
AnalyzedFunctionCount3, MAM));
FPM.addPass(TestInvalidationFunctionPass("f"));
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
}
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
// A fourth function pass manager but with only preserving intervening
// passes but triggering the module analysis.
MPM.addPass(RequireAnalysisPass<TestModuleAnalysis, Module>());
int FunctionPassRunCount4 = 0;
int AnalyzedInstrCount4 = 0;
int AnalyzedFunctionCount4 = 0;
{
FunctionPassManager FPM;
FPM.addPass(TestFunctionPass(FunctionPassRunCount4, AnalyzedInstrCount4,
AnalyzedFunctionCount4, MAM));
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
}
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
// A fifth function pass manager which invalidates one function first but
// uses only cached results.
int FunctionPassRunCount5 = 0;
int AnalyzedInstrCount5 = 0;
int AnalyzedFunctionCount5 = 0;
{
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
FunctionPassManager FPM(/*DebugLogging*/ true);
FPM.addPass(TestInvalidationFunctionPass("f"));
FPM.addPass(TestFunctionPass(FunctionPassRunCount5, AnalyzedInstrCount5,
AnalyzedFunctionCount5, MAM,
/*OnlyUseCachedResults=*/true));
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
}
MPM.run(*M, MAM);
// Validate module pass counters.
EXPECT_EQ(1, ModulePassRunCount);
// Validate all function pass counter sets are the same.
EXPECT_EQ(3, FunctionPassRunCount1);
EXPECT_EQ(5, AnalyzedInstrCount1);
EXPECT_EQ(0, AnalyzedFunctionCount1);
EXPECT_EQ(3, FunctionPassRunCount2);
EXPECT_EQ(5, AnalyzedInstrCount2);
EXPECT_EQ(0, AnalyzedFunctionCount2);
EXPECT_EQ(3, FunctionPassRunCount3);
EXPECT_EQ(5, AnalyzedInstrCount3);
EXPECT_EQ(0, AnalyzedFunctionCount3);
EXPECT_EQ(3, FunctionPassRunCount4);
EXPECT_EQ(5, AnalyzedInstrCount4);
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
EXPECT_EQ(9, AnalyzedFunctionCount4);
EXPECT_EQ(3, FunctionPassRunCount5);
EXPECT_EQ(2, AnalyzedInstrCount5); // Only 'g' and 'h' were cached.
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
EXPECT_EQ(9, AnalyzedFunctionCount5);
// Validate the analysis counters:
// first run over 3 functions, then module pass invalidates
// second run over 3 functions, nothing invalidates
// third run over 0 functions, but 1 function invalidated
// fourth run over 1 function
[PM] Support invalidation of inner analysis managers from a pass over the outer IR unit. Summary: This never really got implemented, and was very hard to test before a lot of the refactoring changes to make things more robust. But now we can test it thoroughly and cleanly, especially at the CGSCC level. The core idea is that when an inner analysis manager proxy receives the invalidation event for the outer IR unit, it needs to walk the inner IR units and propagate it to the inner analysis manager for each of those units. For example, each function in the SCC needs to get an invalidation event when the SCC gets one. The function / module interaction is somewhat boring here. This really becomes interesting in the face of analysis-backed IR units. This patch effectively handles all of the CGSCC layer's needs -- both invalidating SCC analysis and invalidating function analysis when an SCC gets invalidated. However, this second aspect doesn't really handle the LoopAnalysisManager well at this point. That one will need some change of design in order to fully integrate, because unlike the call graph, the entire function behind a LoopAnalysis's results can vanish out from under us, and we won't even have a cached API to access. I'd like to try to separate solving the loop problems into a subsequent patch though in order to keep this more focused so I've adapted them to the API and updated the tests that immediately fail, but I've not added the level of testing and validation at that layer that I have at the CGSCC layer. An important aspect of this change is that the proxy for the FunctionAnalysisManager at the SCC pass layer doesn't work like the other proxies for an inner IR unit as it doesn't directly manage the FunctionAnalysisManager and invalidation or clearing of it. This would create an ever worsening problem of dual ownership of this responsibility, split between the module-level FAM proxy and this SCC-level FAM proxy. Instead, this patch changes the SCC-level FAM proxy to work in terms of the module-level proxy and defer to it to handle much of the updates. It only does SCC-specific invalidation. This will become more important in subsequent patches that support more complex invalidaiton scenarios. Reviewers: jlebar Subscribers: mehdi_amini, mcrosier, mzolotukhin, llvm-commits Differential Revision: https://reviews.llvm.org/D27197 llvm-svn: 289317
2016-12-10 14:34:44 +08:00
// fifth run invalidates 1 function first, but runs over 0 functions
EXPECT_EQ(7, FunctionAnalysisRuns);
EXPECT_EQ(1, ModuleAnalysisRuns);
}
// A customized pass manager that passes extra arguments through the
// infrastructure.
typedef AnalysisManager<Function, int> CustomizedAnalysisManager;
typedef PassManager<Function, CustomizedAnalysisManager, int, int &>
CustomizedPassManager;
class CustomizedAnalysis : public AnalysisInfoMixin<CustomizedAnalysis> {
public:
struct Result {
Result(int I) : I(I) {}
int I;
};
Result run(Function &F, CustomizedAnalysisManager &AM, int I) {
return Result(I);
}
private:
friend AnalysisInfoMixin<CustomizedAnalysis>;
[PM] Change the static object whose address is used to uniquely identify analyses to have a common type which is enforced rather than using a char object and a `void *` type when used as an identifier. This has a number of advantages. First, it at least helps some of the confusion raised in Justin Lebar's code review of why `void *` was being used everywhere by having a stronger type that connects to documentation about this. However, perhaps more importantly, it addresses a serious issue where the alignment of these pointer-like identifiers was unknown. This made it hard to use them in pointer-like data structures. We were already dodging this in dangerous ways to create the "all analyses" entry. In a subsequent patch I attempted to use these with TinyPtrVector and things fell apart in a very bad way. And it isn't just a compile time or type system issue. Worse than that, the actual alignment of these pointer-like opaque identifiers wasn't guaranteed to be a useful alignment as they were just characters. This change introduces a type to use as the "key" object whose address forms the opaque identifier. This both forces the objects to have proper alignment, and provides type checking that we get it right everywhere. It also makes the types somewhat less mysterious than `void *`. We could go one step further and introduce a truly opaque pointer-like type to return from the `ID()` static function rather than returning `AnalysisKey *`, but that didn't seem to be a clear win so this is just the initial change to get to a reliably typed and aligned object serving is a key for all the analyses. Thanks to Richard Smith and Justin Lebar for helping pick plausible names and avoid making this refactoring many times. =] And thanks to Sean for the super fast review! While here, I've tried to move away from the "PassID" nomenclature entirely as it wasn't really helping and is overloaded with old pass manager constructs. Now we have IDs for analyses, and key objects whose address can be used as IDs. Where possible and clear I've shortened this to just "ID". In a few places I kept "AnalysisID" to make it clear what was being identified. Differential Revision: https://reviews.llvm.org/D27031 llvm-svn: 287783
2016-11-24 01:53:26 +08:00
static AnalysisKey Key;
};
[PM] Change the static object whose address is used to uniquely identify analyses to have a common type which is enforced rather than using a char object and a `void *` type when used as an identifier. This has a number of advantages. First, it at least helps some of the confusion raised in Justin Lebar's code review of why `void *` was being used everywhere by having a stronger type that connects to documentation about this. However, perhaps more importantly, it addresses a serious issue where the alignment of these pointer-like identifiers was unknown. This made it hard to use them in pointer-like data structures. We were already dodging this in dangerous ways to create the "all analyses" entry. In a subsequent patch I attempted to use these with TinyPtrVector and things fell apart in a very bad way. And it isn't just a compile time or type system issue. Worse than that, the actual alignment of these pointer-like opaque identifiers wasn't guaranteed to be a useful alignment as they were just characters. This change introduces a type to use as the "key" object whose address forms the opaque identifier. This both forces the objects to have proper alignment, and provides type checking that we get it right everywhere. It also makes the types somewhat less mysterious than `void *`. We could go one step further and introduce a truly opaque pointer-like type to return from the `ID()` static function rather than returning `AnalysisKey *`, but that didn't seem to be a clear win so this is just the initial change to get to a reliably typed and aligned object serving is a key for all the analyses. Thanks to Richard Smith and Justin Lebar for helping pick plausible names and avoid making this refactoring many times. =] And thanks to Sean for the super fast review! While here, I've tried to move away from the "PassID" nomenclature entirely as it wasn't really helping and is overloaded with old pass manager constructs. Now we have IDs for analyses, and key objects whose address can be used as IDs. Where possible and clear I've shortened this to just "ID". In a few places I kept "AnalysisID" to make it clear what was being identified. Differential Revision: https://reviews.llvm.org/D27031 llvm-svn: 287783
2016-11-24 01:53:26 +08:00
AnalysisKey CustomizedAnalysis::Key;
struct CustomizedPass : PassInfoMixin<CustomizedPass> {
std::function<void(CustomizedAnalysis::Result &, int &)> Callback;
template <typename CallbackT>
CustomizedPass(CallbackT Callback) : Callback(Callback) {}
PreservedAnalyses run(Function &F, CustomizedAnalysisManager &AM, int I,
int &O) {
Callback(AM.getResult<CustomizedAnalysis>(F, I), O);
return PreservedAnalyses::none();
}
};
TEST_F(PassManagerTest, CustomizedPassManagerArgs) {
CustomizedAnalysisManager AM;
AM.registerPass([&] { return CustomizedAnalysis(); });
[New PM] Introducing PassInstrumentation framework Pass Execution Instrumentation interface enables customizable instrumentation of pass execution, as per "RFC: Pass Execution Instrumentation interface" posted 06/07/2018 on llvm-dev@ The intent is to provide a common machinery to implement all the pass-execution-debugging features like print-before/after, opt-bisect, time-passes etc. Here we get a basic implementation consisting of: * PassInstrumentationCallbacks class that handles registration of callbacks and access to them. * PassInstrumentation class that handles instrumentation-point interfaces that call into PassInstrumentationCallbacks. * Callbacks accept StringRef which is just a name of the Pass right now. There were some ideas to pass an opaque wrapper for the pointer to pass instance, however it appears that pointer does not actually identify the instance (adaptors and managers might have the same address with the pass they govern). Hence it was decided to go simple for now and then later decide on what the proper mental model of identifying a "pass in a phase of pipeline" is. * Callbacks accept llvm::Any serving as a wrapper for const IRUnit*, to remove direct dependencies on different IRUnits (e.g. Analyses). * PassInstrumentationAnalysis analysis is explicitly requested from PassManager through usual AnalysisManager::getResult. All pass managers were updated to run that to get PassInstrumentation object for instrumentation calls. * Using tuples/index_sequence getAnalysisResult helper to extract generic AnalysisManager's extra args out of a generic PassManager's extra args. This is the only way I was able to explicitly run getResult for PassInstrumentationAnalysis out of a generic code like PassManager::run or RepeatedPass::run. TODO: Upon lengthy discussions we agreed to accept this as an initial implementation and then get rid of getAnalysisResult by improving RepeatedPass implementation. * PassBuilder takes PassInstrumentationCallbacks object to pass it further into PassInstrumentationAnalysis. Callbacks registration should be performed directly through PassInstrumentationCallbacks. * new-pm tests updated to account for PassInstrumentationAnalysis being run * Added PassInstrumentation tests to PassBuilderCallbacks unit tests. Other unit tests updated with registration of the now-required PassInstrumentationAnalysis. Made getName helper to return std::string (instead of StringRef initially) to fix asan builtbot failures on CGSCC tests. Reviewers: chandlerc, philip.pfaffe Differential Revision: https://reviews.llvm.org/D47858 llvm-svn: 342664
2018-09-21 01:08:45 +08:00
PassInstrumentationCallbacks PIC;
AM.registerPass([&] { return PassInstrumentationAnalysis(&PIC); });
CustomizedPassManager PM;
// Add an instance of the customized pass that just accumulates the input
// after it is round-tripped through the analysis.
int Result = 0;
PM.addPass(
CustomizedPass([](CustomizedAnalysis::Result &R, int &O) { O += R.I; }));
// Run this over every function with the input of 42.
for (Function &F : *M)
PM.run(F, AM, 42, Result);
// And ensure that we accumulated the correct result.
EXPECT_EQ(42 * (int)M->size(), Result);
}
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
/// A test analysis pass which caches in its result another analysis pass and
/// uses it to serve queries. This requires the result to invalidate itself
/// when its dependency is invalidated.
struct TestIndirectFunctionAnalysis
: public AnalysisInfoMixin<TestIndirectFunctionAnalysis> {
struct Result {
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
Result(TestFunctionAnalysis::Result &FDep, TestModuleAnalysis::Result &MDep)
: FDep(FDep), MDep(MDep) {}
TestFunctionAnalysis::Result &FDep;
TestModuleAnalysis::Result &MDep;
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
bool invalidate(Function &F, const PreservedAnalyses &PA,
FunctionAnalysisManager::Invalidator &Inv) {
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
auto PAC = PA.getChecker<TestIndirectFunctionAnalysis>();
return !(PAC.preserved() ||
PAC.preservedSet<AllAnalysesOn<Function>>()) ||
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
Inv.invalidate<TestFunctionAnalysis>(F, PA);
}
};
TestIndirectFunctionAnalysis(int &Runs, ModuleAnalysisManager &MAM)
: Runs(Runs), MAM(MAM) {}
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
/// Run the analysis pass over the function and return a result.
Result run(Function &F, FunctionAnalysisManager &AM) {
++Runs;
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
auto &FDep = AM.getResult<TestFunctionAnalysis>(F);
auto &MAMProxy = AM.getResult<ModuleAnalysisManagerFunctionProxy>(F);
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
// For the test, we insist that the module analysis starts off in the
// cache. Getting a cached result that isn't stateless trigger an assert.
// Use MAM, for the purposes of this unittest.
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
auto &MDep = *MAM.getCachedResult<TestModuleAnalysis>(*F.getParent());
// And register the dependency as module analysis dependencies have to be
// pre-registered on the proxy.
MAMProxy.registerOuterAnalysisInvalidation<TestModuleAnalysis,
TestIndirectFunctionAnalysis>();
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
return Result(FDep, MDep);
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
}
private:
friend AnalysisInfoMixin<TestIndirectFunctionAnalysis>;
static AnalysisKey Key;
int &Runs;
ModuleAnalysisManager &MAM;
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
};
AnalysisKey TestIndirectFunctionAnalysis::Key;
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
/// A test analysis pass which chaches in its result the result from the above
/// indirect analysis pass.
///
/// This allows us to ensure that whenever an analysis pass is invalidated due
/// to dependencies (especially dependencies across IR units that trigger
/// asynchronous invalidation) we correctly detect that this may in turn cause
/// other analysis to be invalidated.
struct TestDoublyIndirectFunctionAnalysis
: public AnalysisInfoMixin<TestDoublyIndirectFunctionAnalysis> {
struct Result {
Result(TestIndirectFunctionAnalysis::Result &IDep) : IDep(IDep) {}
TestIndirectFunctionAnalysis::Result &IDep;
bool invalidate(Function &F, const PreservedAnalyses &PA,
FunctionAnalysisManager::Invalidator &Inv) {
auto PAC = PA.getChecker<TestDoublyIndirectFunctionAnalysis>();
return !(PAC.preserved() ||
PAC.preservedSet<AllAnalysesOn<Function>>()) ||
Inv.invalidate<TestIndirectFunctionAnalysis>(F, PA);
}
};
TestDoublyIndirectFunctionAnalysis(int &Runs) : Runs(Runs) {}
/// Run the analysis pass over the function and return a result.
Result run(Function &F, FunctionAnalysisManager &AM) {
++Runs;
auto &IDep = AM.getResult<TestIndirectFunctionAnalysis>(F);
return Result(IDep);
}
private:
friend AnalysisInfoMixin<TestDoublyIndirectFunctionAnalysis>;
static AnalysisKey Key;
int &Runs;
};
AnalysisKey TestDoublyIndirectFunctionAnalysis::Key;
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
struct LambdaPass : public PassInfoMixin<LambdaPass> {
using FuncT = std::function<PreservedAnalyses(Function &, FunctionAnalysisManager &)>;
LambdaPass(FuncT Func) : Func(std::move(Func)) {}
PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM) {
return Func(F, AM);
}
FuncT Func;
};
TEST_F(PassManagerTest, IndirectAnalysisInvalidation) {
FunctionAnalysisManager FAM(/*DebugLogging*/ true);
ModuleAnalysisManager MAM(/*DebugLogging*/ true);
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
int FunctionAnalysisRuns = 0, ModuleAnalysisRuns = 0,
IndirectAnalysisRuns = 0, DoublyIndirectAnalysisRuns = 0;
FAM.registerPass([&] { return TestFunctionAnalysis(FunctionAnalysisRuns); });
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
FAM.registerPass(
[&] { return TestIndirectFunctionAnalysis(IndirectAnalysisRuns, MAM); });
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
FAM.registerPass([&] {
return TestDoublyIndirectFunctionAnalysis(DoublyIndirectAnalysisRuns);
});
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
MAM.registerPass([&] { return TestModuleAnalysis(ModuleAnalysisRuns); });
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
MAM.registerPass([&] { return FunctionAnalysisManagerModuleProxy(FAM); });
FAM.registerPass([&] { return ModuleAnalysisManagerFunctionProxy(MAM); });
[New PM] Introducing PassInstrumentation framework Pass Execution Instrumentation interface enables customizable instrumentation of pass execution, as per "RFC: Pass Execution Instrumentation interface" posted 06/07/2018 on llvm-dev@ The intent is to provide a common machinery to implement all the pass-execution-debugging features like print-before/after, opt-bisect, time-passes etc. Here we get a basic implementation consisting of: * PassInstrumentationCallbacks class that handles registration of callbacks and access to them. * PassInstrumentation class that handles instrumentation-point interfaces that call into PassInstrumentationCallbacks. * Callbacks accept StringRef which is just a name of the Pass right now. There were some ideas to pass an opaque wrapper for the pointer to pass instance, however it appears that pointer does not actually identify the instance (adaptors and managers might have the same address with the pass they govern). Hence it was decided to go simple for now and then later decide on what the proper mental model of identifying a "pass in a phase of pipeline" is. * Callbacks accept llvm::Any serving as a wrapper for const IRUnit*, to remove direct dependencies on different IRUnits (e.g. Analyses). * PassInstrumentationAnalysis analysis is explicitly requested from PassManager through usual AnalysisManager::getResult. All pass managers were updated to run that to get PassInstrumentation object for instrumentation calls. * Using tuples/index_sequence getAnalysisResult helper to extract generic AnalysisManager's extra args out of a generic PassManager's extra args. This is the only way I was able to explicitly run getResult for PassInstrumentationAnalysis out of a generic code like PassManager::run or RepeatedPass::run. TODO: Upon lengthy discussions we agreed to accept this as an initial implementation and then get rid of getAnalysisResult by improving RepeatedPass implementation. * PassBuilder takes PassInstrumentationCallbacks object to pass it further into PassInstrumentationAnalysis. Callbacks registration should be performed directly through PassInstrumentationCallbacks. * new-pm tests updated to account for PassInstrumentationAnalysis being run * Added PassInstrumentation tests to PassBuilderCallbacks unit tests. Other unit tests updated with registration of the now-required PassInstrumentationAnalysis. Made getName helper to return std::string (instead of StringRef initially) to fix asan builtbot failures on CGSCC tests. Reviewers: chandlerc, philip.pfaffe Differential Revision: https://reviews.llvm.org/D47858 llvm-svn: 342664
2018-09-21 01:08:45 +08:00
PassInstrumentationCallbacks PIC;
MAM.registerPass([&] { return PassInstrumentationAnalysis(&PIC); });
FAM.registerPass([&] { return PassInstrumentationAnalysis(&PIC); });
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
int InstrCount = 0, FunctionCount = 0;
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
ModulePassManager MPM(/*DebugLogging*/ true);
FunctionPassManager FPM(/*DebugLogging*/ true);
// First just use the analysis to get the instruction count, and preserve
// everything.
FPM.addPass(LambdaPass([&](Function &F, FunctionAnalysisManager &AM) {
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
auto &DoublyIndirectResult =
AM.getResult<TestDoublyIndirectFunctionAnalysis>(F);
auto &IndirectResult = DoublyIndirectResult.IDep;
InstrCount += IndirectResult.FDep.InstructionCount;
FunctionCount += IndirectResult.MDep.FunctionCount;
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
return PreservedAnalyses::all();
}));
// Next, invalidate
// - both analyses for "f",
// - just the underlying (indirect) analysis for "g", and
// - just the direct analysis for "h".
FPM.addPass(LambdaPass([&](Function &F, FunctionAnalysisManager &AM) {
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
auto &DoublyIndirectResult =
AM.getResult<TestDoublyIndirectFunctionAnalysis>(F);
auto &IndirectResult = DoublyIndirectResult.IDep;
InstrCount += IndirectResult.FDep.InstructionCount;
FunctionCount += IndirectResult.MDep.FunctionCount;
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
auto PA = PreservedAnalyses::none();
if (F.getName() == "g")
PA.preserve<TestFunctionAnalysis>();
else if (F.getName() == "h")
PA.preserve<TestIndirectFunctionAnalysis>();
return PA;
}));
// Finally, use the analysis again on each function, forcing re-computation
// for all of them.
FPM.addPass(LambdaPass([&](Function &F, FunctionAnalysisManager &AM) {
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
auto &DoublyIndirectResult =
AM.getResult<TestDoublyIndirectFunctionAnalysis>(F);
auto &IndirectResult = DoublyIndirectResult.IDep;
InstrCount += IndirectResult.FDep.InstructionCount;
FunctionCount += IndirectResult.MDep.FunctionCount;
return PreservedAnalyses::all();
}));
// Create a second function pass manager. This will cause the module-level
// invalidation to occur, which will force yet another invalidation of the
// indirect function-level analysis as the module analysis it depends on gets
// invalidated.
FunctionPassManager FPM2(/*DebugLogging*/ true);
FPM2.addPass(LambdaPass([&](Function &F, FunctionAnalysisManager &AM) {
auto &DoublyIndirectResult =
AM.getResult<TestDoublyIndirectFunctionAnalysis>(F);
auto &IndirectResult = DoublyIndirectResult.IDep;
InstrCount += IndirectResult.FDep.InstructionCount;
FunctionCount += IndirectResult.MDep.FunctionCount;
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
return PreservedAnalyses::all();
}));
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
// Add a requires pass to populate the module analysis and then our function
// pass pipeline.
MPM.addPass(RequireAnalysisPass<TestModuleAnalysis, Module>());
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
// Now require the module analysis again (it will have been invalidated once)
// and then use it again from a function pass manager.
MPM.addPass(RequireAnalysisPass<TestModuleAnalysis, Module>());
MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM2)));
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
MPM.run(*M, MAM);
// There are generally two possible runs for each of the three functions. But
// for one function, we only invalidate the indirect analysis so the base one
// only gets run five times.
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
EXPECT_EQ(5, FunctionAnalysisRuns);
// The module analysis pass should be run twice here.
EXPECT_EQ(2, ModuleAnalysisRuns);
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
// The indirect analysis is invalidated for each function (either directly or
// indirectly) and run twice for each.
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
EXPECT_EQ(9, IndirectAnalysisRuns);
EXPECT_EQ(9, DoublyIndirectAnalysisRuns);
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
// There are five instructions in the module and we add the count four
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
// times.
[PM] Introduce the facilities for registering cross-IR-unit dependencies that require deferred invalidation. This handles the other real-world invalidation scenario that we have cases of: a function analysis which caches references to a module analysis. We currently do this in the AA aggregation layer and might well do this in other places as well. Since this is relative rare, the technique is somewhat more cumbersome. Analyses need to register themselves when accessing the outer analysis manager's proxy. This proxy is already necessarily present to allow access to the outer IR unit's analyses. By registering here we can track and trigger invalidation when that outer analysis goes away. To make this work we need to enhance the PreservedAnalyses infrastructure to support a (slightly) more explicit model for "sets" of analyses, and allow abandoning a single specific analyses even when a set covering that analysis is preserved. That allows us to describe the scenario of preserving all Function analyses *except* for the one where deferred invalidation has triggered. We also need to teach the invalidator API to support direct ID calls instead of always going through a template to dispatch so that we can just record the ID mapping. I've introduced testing of all of this both for simple module<->function cases as well as for more complex cases involving a CGSCC layer. Much like the previous patch I've not tried to fully update the loop pass management layer because that layer is due to be heavily reworked to use similar techniques to the CGSCC to handle updates. As that happens, we'll have a better testing basis for adding support like this. Many thanks to both Justin and Sean for the extensive reviews on this to help bring the API design and documentation into a better state. Differential Revision: https://reviews.llvm.org/D27198 llvm-svn: 290594
2016-12-27 16:40:39 +08:00
EXPECT_EQ(5 * 4, InstrCount);
// There are three functions and we count them four times for each of the
// three functions.
EXPECT_EQ(3 * 4 * 3, FunctionCount);
[PM] Extend the explicit 'invalidate' method API on analysis results to accept an Invalidator that allows them to invalidate themselves if their dependencies are in turn invalidated. Rather than recording the dependency graph ahead of time when analysis get results from other analyses, this simply lets each result trigger the immediate invalidation of any analyses they actually depend on. They do this in a way that has three nice properties: 1) They don't have to handle transitive dependencies because the infrastructure will recurse for them. 2) The invalidate methods are still called only once. We just dynamically discover the necessary topological ordering, everything is memoized nicely. 3) The infrastructure still provides a default implementation and can access it so that only analyses which have dependencies need to do anything custom. To make this work at all, the invalidation logic also has to defer the deletion of the result objects themselves so that they can remain alive until we have collected the complete set of results to invalidate. A unittest is added here that has exactly the dependency pattern we are concerned with. It hit the use-after-free described by Sean in much detail in the long thread about analysis invalidation before this change, and even in an intermediate form of this change where we failed to defer the deletion of the result objects. There is an important problem with doing dependency invalidation that *isn't* solved here: we don't *enforce* that results correctly invalidate all the analyses whose results they depend on. I actually looked at what it would take to do that, and it isn't as hard as I had thought but the complexity it introduces seems very likely to outweigh the benefit. The technique would be to provide a base class for an analysis result that would be populated with other results, and automatically provide the invalidate method which immediately does the correct thing. This approach has some nice pros IMO: - Handles the case we care about and nothing else: only *results* that depend on other analyses trigger extra invalidation. - Localized to the result rather than centralized in the analysis manager. - Ties the storage of the reference to another result to the triggering of the invalidation of that analysis. - Still supports extending invalidation in customized ways. But the down sides here are: - Very heavy-weight meta-programming is needed to provide this base class. - Requires a pretty awful API for accessing the dependencies. Ultimately, I fear it will not pull its weight. But we can re-evaluate this at any point if we start discovering consistent problems where the invalidation and dependencies get out of sync. It will fit as a clean layer on top of the facilities in this patch that we can add if and when we need it. Note that I'm not really thrilled with the names for these APIs... The name "Invalidator" seems ok but not great. The method name "invalidate" also. In review some improvements were suggested, but they really need *other* uses of these terms to be updated as well so I'm going to do that in a follow-up commit. I'm working on the actual fixes to various analyses that need to use these, but I want to try to get tests for each of them so we don't regress. And those changes are seperable and obvious so once this goes in I should be able to roll them out throughout LLVM. Many thanks to Sean, Justin, and others for help reviewing here. Differential Revision: https://reviews.llvm.org/D23738 llvm-svn: 288077
2016-11-29 06:04:31 +08:00
}
// Run SimplifyCFGPass that makes CFG changes and reports PreservedAnalyses
// without CFGAnalyses. So the CFGChecker does not complain.
TEST_F(PassManagerTest, FunctionPassCFGChecker) {
LLVMContext Context;
// SimplifyCFG changes this function to
// define void @foo {next: ret void}
auto M = parseIR(Context, "define void @foo() {\n"
" br label %next\n"
"next:\n"
" br label %exit\n"
"exit:\n"
" ret void\n"
"}\n");
auto *F = M->getFunction("foo");
FunctionAnalysisManager FAM(/*DebugLogging*/ true);
FunctionPassManager FPM(/*DebugLogging*/ true);
PassInstrumentationCallbacks PIC;
StandardInstrumentations SI(/*DebugLogging*/ true);
SI.registerCallbacks(PIC, &FAM);
FAM.registerPass([&] { return PassInstrumentationAnalysis(&PIC); });
FAM.registerPass([&] { return DominatorTreeAnalysis(); });
FAM.registerPass([&] { return AssumptionAnalysis(); });
FAM.registerPass([&] { return TargetIRAnalysis(); });
FPM.addPass(SimplifyCFGPass());
FPM.run(*F, FAM);
}
// FunctionPass that manually invalidates analyses and always returns
// PreservedAnalyses::all().
struct TestSimplifyCFGInvalidatingAnalysisPass
: PassInfoMixin<TestSimplifyCFGInvalidatingAnalysisPass> {
PreservedAnalyses run(Function &F, FunctionAnalysisManager &FAM) {
// Run SimplifyCFG and if it changes CFG then invalidate the CFG analysis.
// This allows to return PreserveAnalysis::all().
PreservedAnalyses PA = CFGSimplifier.run(F, FAM);
FAM.invalidate(F, PA);
return PreservedAnalyses::all();
}
SimplifyCFGPass CFGSimplifier;
};
// Run TestSimplifyCFGInvalidatingAnalysisPass which changes CFG by running
// SimplifyCFGPass then manually invalidates analyses and always returns
// PreservedAnalyses::all(). CFGChecker does not complain because it resets
// its saved CFG snapshot when the analyses are invalidated manually.
TEST_F(PassManagerTest, FunctionPassCFGCheckerInvalidateAnalysis) {
LLVMContext Context;
// SimplifyCFG changes this function to
// define void @foo {next: ret void}
auto M = parseIR(Context, "define void @foo() {\n"
" br label %next\n"
"next:\n"
" br label %exit\n"
"exit:\n"
" ret void\n"
"}\n");
auto *F = M->getFunction("foo");
FunctionAnalysisManager FAM(/*DebugLogging*/ true);
FunctionPassManager FPM(/*DebugLogging*/ true);
PassInstrumentationCallbacks PIC;
StandardInstrumentations SI(/*DebugLogging*/ true);
SI.registerCallbacks(PIC, &FAM);
FAM.registerPass([&] { return PassInstrumentationAnalysis(&PIC); });
FAM.registerPass([&] { return DominatorTreeAnalysis(); });
FAM.registerPass([&] { return AssumptionAnalysis(); });
FAM.registerPass([&] { return TargetIRAnalysis(); });
FPM.addPass(TestSimplifyCFGInvalidatingAnalysisPass());
FPM.run(*F, FAM);
}
// Wrap a FunctionPassManager running SimplifyCFG pass with another
// FunctionPassManager.
struct TestSimplifyCFGWrapperPass : PassInfoMixin<TestSimplifyCFGWrapperPass> {
TestSimplifyCFGWrapperPass(FunctionPassManager &InnerPM) : InnerPM(InnerPM) {}
PreservedAnalyses run(Function &F, FunctionAnalysisManager &FAM) {
// Here we simulate exactly what FunctionPassManager::run() does but
// instead of running all passes from InnerPM.Passes we run them in bulk
// by calling InnerPM.run().
PreservedAnalyses PA = PreservedAnalyses::all();
PassInstrumentation PI = FAM.getResult<PassInstrumentationAnalysis>(F);
if (!PI.runBeforePass<Function>(InnerPM, F))
return PreservedAnalyses::all();
PreservedAnalyses PassPA = InnerPM.run(F, FAM);
PI.runAfterPass(InnerPM, F, PassPA);
FAM.invalidate(F, PassPA);
PA.intersect(PassPA);
PA.preserveSet<AllAnalysesOn<Function>>();
return PA;
}
FunctionPassManager &InnerPM;
};
// Run TestSimplifyCFGWrapperPass which simulates behavior of
// FunctionPassManager::run() except that it runs all passes at once by calling
// an inner pass manager's passes with PassManager::run(). This is how one pass
// manager is expected to wrap another pass manager.
// SimplifyCFGPass, which is called by the inner pass manager, changes the CFG.
// The CFGChecker's AfterPassCallback, run right after SimplifyCFGPass, does not
// complain because CFGAnalyses is not in the PreservedAnalises set returned by
// SimplifyCFGPass. Then the CFG analysis is invalidated by the analysis manager
// according to the PreservedAnalises set. Further calls to CFGChecker's
// AfterPassCallback see that all analyses for the current function are
// preserved but there is no CFG snapshot available (i.e.
// AM.getCachedResult<PreservedCFGCheckerAnalysis>(F) returns nullptr).
TEST_F(PassManagerTest, FunctionPassCFGCheckerWrapped) {
LLVMContext Context;
// SimplifyCFG changes this function to
// define void @foo {next: ret void}
auto M = parseIR(Context, "define void @foo() {\n"
" br label %next\n"
"next:\n"
" br label %exit\n"
"exit:\n"
" ret void\n"
"}\n");
auto *F = M->getFunction("foo");
FunctionAnalysisManager FAM(/*DebugLogging*/ true);
FunctionPassManager FPM(/*DebugLogging*/ true);
PassInstrumentationCallbacks PIC;
StandardInstrumentations SI(/*DebugLogging*/ true);
SI.registerCallbacks(PIC, &FAM);
FAM.registerPass([&] { return PassInstrumentationAnalysis(&PIC); });
FAM.registerPass([&] { return DominatorTreeAnalysis(); });
FAM.registerPass([&] { return AssumptionAnalysis(); });
FAM.registerPass([&] { return TargetIRAnalysis(); });
FunctionPassManager InnerFPM(/*DebugLogging*/ true);
InnerFPM.addPass(SimplifyCFGPass());
FPM.addPass(TestSimplifyCFGWrapperPass(InnerFPM));
FPM.run(*F, FAM);
}
}