llvm-project/llvm/lib/Analysis/EHPersonalities.cpp

110 lines
4.6 KiB
C++
Raw Normal View History

//===- EHPersonalities.cpp - Compute EH-related information ---------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#include "llvm/Analysis/EHPersonalities.h"
#include "llvm/ADT/StringSwitch.h"
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
#include "llvm/IR/CFG.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/Function.h"
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
#include "llvm/IR/Instructions.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
/// See if the given exception handling personality function is one that we
/// understand. If so, return a description of it; otherwise return Unknown.
EHPersonality llvm::classifyEHPersonality(const Value *Pers) {
const Function *F =
Pers ? dyn_cast<Function>(Pers->stripPointerCasts()) : nullptr;
if (!F)
return EHPersonality::Unknown;
return StringSwitch<EHPersonality>(F->getName())
.Case("__gnat_eh_personality", EHPersonality::GNU_Ada)
.Case("__gxx_personality_v0", EHPersonality::GNU_CXX)
.Case("__gxx_personality_sj0", EHPersonality::GNU_CXX_SjLj)
.Case("__gcc_personality_v0", EHPersonality::GNU_C)
.Case("__gcc_personality_sj0", EHPersonality::GNU_C_SjLj)
.Case("__objc_personality_v0", EHPersonality::GNU_ObjC)
.Case("_except_handler3", EHPersonality::MSVC_X86SEH)
.Case("_except_handler4", EHPersonality::MSVC_X86SEH)
.Case("__C_specific_handler", EHPersonality::MSVC_Win64SEH)
.Case("__CxxFrameHandler3", EHPersonality::MSVC_CXX)
.Case("ProcessCLRException", EHPersonality::CoreCLR)
.Case("rust_eh_personality", EHPersonality::Rust)
.Default(EHPersonality::Unknown);
}
bool llvm::canSimplifyInvokeNoUnwind(const Function *F) {
EHPersonality Personality = classifyEHPersonality(F->getPersonalityFn());
// We can't simplify any invokes to nounwind functions if the personality
// function wants to catch asynch exceptions. The nounwind attribute only
// implies that the function does not throw synchronous exceptions.
return !isAsynchronousEHPersonality(Personality);
}
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
DenseMap<BasicBlock *, ColorVector> llvm::colorEHFunclets(Function &F) {
SmallVector<std::pair<BasicBlock *, BasicBlock *>, 16> Worklist;
BasicBlock *EntryBlock = &F.getEntryBlock();
DenseMap<BasicBlock *, ColorVector> BlockColors;
// Build up the color map, which maps each block to its set of 'colors'.
// For any block B the "colors" of B are the set of funclets F (possibly
// including a root "funclet" representing the main function) such that
// F will need to directly contain B or a copy of B (where the term "directly
// contain" is used to distinguish from being "transitively contained" in
// a nested funclet).
//
// Note: Despite not being a funclet in the truest sense, a catchswitch is
// considered to belong to its own funclet for the purposes of coloring.
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
DEBUG_WITH_TYPE("winehprepare-coloring", dbgs() << "\nColoring funclets for "
<< F.getName() << "\n");
Worklist.push_back({EntryBlock, EntryBlock});
while (!Worklist.empty()) {
BasicBlock *Visiting;
BasicBlock *Color;
std::tie(Visiting, Color) = Worklist.pop_back_val();
DEBUG_WITH_TYPE("winehprepare-coloring",
dbgs() << "Visiting " << Visiting->getName() << ", "
<< Color->getName() << "\n");
Instruction *VisitingHead = Visiting->getFirstNonPHI();
if (VisitingHead->isEHPad()) {
// Mark this funclet head as a member of itself.
Color = Visiting;
}
// Note that this is a member of the given color.
ColorVector &Colors = BlockColors[Visiting];
if (std::find(Colors.begin(), Colors.end(), Color) == Colors.end())
Colors.push_back(Color);
else
continue;
DEBUG_WITH_TYPE("winehprepare-coloring",
dbgs() << " Assigned color \'" << Color->getName()
<< "\' to block \'" << Visiting->getName()
<< "\'.\n");
BasicBlock *SuccColor = Color;
TerminatorInst *Terminator = Visiting->getTerminator();
if (auto *CatchRet = dyn_cast<CatchReturnInst>(Terminator)) {
Value *ParentPad = CatchRet->getCatchSwitchParentPad();
[IR] Reformulate LLVM's EH funclet IR While we have successfully implemented a funclet-oriented EH scheme on top of LLVM IR, our scheme has some notable deficiencies: - catchendpad and cleanupendpad are necessary in the current design but they are difficult to explain to others, even to seasoned LLVM experts. - catchendpad and cleanupendpad are optimization barriers. They cannot be split and force all potentially throwing call-sites to be invokes. This has a noticable effect on the quality of our code generation. - catchpad, while similar in some aspects to invoke, is fairly awkward. It is unsplittable, starts a funclet, and has control flow to other funclets. - The nesting relationship between funclets is currently a property of control flow edges. Because of this, we are forced to carefully analyze the flow graph to see if there might potentially exist illegal nesting among funclets. While we have logic to clone funclets when they are illegally nested, it would be nicer if we had a representation which forbade them upfront. Let's clean this up a bit by doing the following: - Instead, make catchpad more like cleanuppad and landingpad: no control flow, just a bunch of simple operands; catchpad would be splittable. - Introduce catchswitch, a control flow instruction designed to model the constraints of funclet oriented EH. - Make funclet scoping explicit by having funclet instructions consume the token produced by the funclet which contains them. - Remove catchendpad and cleanupendpad. Their presence can be inferred implicitly using coloring information. N.B. The state numbering code for the CLR has been updated but the veracity of it's output cannot be spoken for. An expert should take a look to make sure the results are reasonable. Reviewers: rnk, JosephTremoulet, andrew.w.kaylor Differential Revision: http://reviews.llvm.org/D15139 llvm-svn: 255422
2015-12-12 13:38:55 +08:00
if (isa<ConstantTokenNone>(ParentPad))
SuccColor = EntryBlock;
else
SuccColor = cast<Instruction>(ParentPad)->getParent();
}
for (BasicBlock *Succ : successors(Visiting))
Worklist.push_back({Succ, SuccColor});
}
return BlockColors;
}