maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/lib/Transforms/IPO/GlobalOpt.cpp

911 lines
35 KiB
C++
Raw Normal View History

* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
//===- GlobalOpt.cpp - Optimize Global Variables --------------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This pass transforms simple global variables that never have their address
// taken. If obviously true, it marks read/write globals as constant, deletes
// variables only stored to, etc.
//
//===----------------------------------------------------------------------===//
#define DEBUG_TYPE "globalopt"
#include "llvm/Transforms/IPO.h"
#include "llvm/Constants.h"
#include "llvm/DerivedTypes.h"
#include "llvm/Instructions.h"
#include "llvm/Module.h"
#include "llvm/Pass.h"
#include "llvm/Support/Debug.h"
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
#include "llvm/Target/TargetData.h"
#include "llvm/Transforms/Utils/Local.h"
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
#include "llvm/ADT/Statistic.h"
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
#include "llvm/ADT/StringExtras.h"
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
#include <set>
#include <algorithm>
using namespace llvm;
namespace {
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
Statistic<> NumMarked ("globalopt", "Number of globals marked constant");
Statistic<> NumSRA ("globalopt", "Number of aggregate globals broken "
"into scalars");
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
Statistic<> NumSubstitute("globalopt",
"Number of globals with initializers stored into them");
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
Statistic<> NumDeleted ("globalopt", "Number of globals deleted");
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
Statistic<> NumFnDeleted("globalopt", "Number of functions deleted");
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
Statistic<> NumGlobUses ("globalopt", "Number of global uses devirtualized");
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
struct GlobalOpt : public ModulePass {
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
virtual void getAnalysisUsage(AnalysisUsage &AU) const {
AU.addRequired<TargetData>();
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
bool runOnModule(Module &M);
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
private:
bool ProcessInternalGlobal(GlobalVariable *GV, Module::giterator &GVI);
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
};
RegisterOpt<GlobalOpt> X("globalopt", "Global Variable Optimizer");
}
ModulePass *llvm::createGlobalOptimizerPass() { return new GlobalOpt(); }
/// GlobalStatus - As we analyze each global, keep track of some information
/// about it. If we find out that the address of the global is taken, none of
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
/// this info will be accurate.
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
struct GlobalStatus {
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
/// isLoaded - True if the global is ever loaded. If the global isn't ever
/// loaded it can be deleted.
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
bool isLoaded;
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
/// StoredType - Keep track of what stores to the global look like.
///
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
enum StoredType {
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
/// NotStored - There is no store to this global. It can thus be marked
/// constant.
NotStored,
/// isInitializerStored - This global is stored to, but the only thing
/// stored is the constant it was initialized with. This is only tracked
/// for scalar globals.
isInitializerStored,
/// isStoredOnce - This global is stored to, but only its initializer and
/// one other value is ever stored to it. If this global isStoredOnce, we
/// track the value stored to it in StoredOnceValue below. This is only
/// tracked for scalar globals.
isStoredOnce,
/// isStored - This global is stored to by multiple values or something else
/// that we cannot track.
isStored
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
} StoredType;
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
/// StoredOnceValue - If only one value (besides the initializer constant) is
/// ever stored to this global, keep track of what value it is.
Value *StoredOnceValue;
/// isNotSuitableForSRA - Keep track of whether any SRA preventing users of
/// the global exist. Such users include GEP instruction with variable
/// indexes, and non-gep/load/store users like constant expr casts.
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
bool isNotSuitableForSRA;
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
GlobalStatus() : isLoaded(false), StoredType(NotStored), StoredOnceValue(0),
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
isNotSuitableForSRA(false) {}
};
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
/// ConstantIsDead - Return true if the specified constant is (transitively)
/// dead. The constant may be used by other constants (e.g. constant arrays and
/// constant exprs) as long as they are dead, but it cannot be used by anything
/// else.
static bool ConstantIsDead(Constant *C) {
if (isa<GlobalValue>(C)) return false;
for (Value::use_iterator UI = C->use_begin(), E = C->use_end(); UI != E; ++UI)
if (Constant *CU = dyn_cast<Constant>(*UI)) {
if (!ConstantIsDead(CU)) return false;
} else
return false;
return true;
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
/// AnalyzeGlobal - Look at all uses of the global and fill in the GlobalStatus
/// structure. If the global has its address taken, return true to indicate we
/// can't do anything with it.
///
static bool AnalyzeGlobal(Value *V, GlobalStatus &GS,
std::set<PHINode*> &PHIUsers) {
for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E; ++UI)
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(*UI)) {
if (AnalyzeGlobal(CE, GS, PHIUsers)) return true;
if (CE->getOpcode() != Instruction::GetElementPtr)
GS.isNotSuitableForSRA = true;
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
else if (!GS.isNotSuitableForSRA) {
// Check to see if this ConstantExpr GEP is SRA'able. In particular, we
// don't like < 3 operand CE's, and we don't like non-constant integer
// indices.
if (CE->getNumOperands() < 3 || !CE->getOperand(1)->isNullValue())
GS.isNotSuitableForSRA = true;
else {
for (unsigned i = 1, e = CE->getNumOperands(); i != e; ++i)
if (!isa<ConstantInt>(CE->getOperand(i))) {
GS.isNotSuitableForSRA = true;
break;
}
}
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
} else if (Instruction *I = dyn_cast<Instruction>(*UI)) {
if (isa<LoadInst>(I)) {
GS.isLoaded = true;
} else if (StoreInst *SI = dyn_cast<StoreInst>(I)) {
Fix a bug in the safety analysis routine llvm-svn: 16804
2004-10-07 14:01:25 +08:00
// Don't allow a store OF the address, only stores TO the address.
if (SI->getOperand(0) == V) return true;
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
// If this is a direct store to the global (i.e., the global is a scalar
// value, not an aggregate), keep more specific information about
// stores.
if (GS.StoredType != GlobalStatus::isStored)
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(SI->getOperand(1))){
if (SI->getOperand(0) == GV->getInitializer()) {
if (GS.StoredType < GlobalStatus::isInitializerStored)
GS.StoredType = GlobalStatus::isInitializerStored;
} else if (GS.StoredType < GlobalStatus::isStoredOnce) {
GS.StoredType = GlobalStatus::isStoredOnce;
GS.StoredOnceValue = SI->getOperand(0);
} else if (GS.StoredType == GlobalStatus::isStoredOnce &&
GS.StoredOnceValue == SI->getOperand(0)) {
// noop.
} else {
GS.StoredType = GlobalStatus::isStored;
}
} else {
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
GS.StoredType = GlobalStatus::isStored;
Improve comments, no functionality changes llvm-svn: 16814
2004-10-08 05:30:30 +08:00
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
} else if (I->getOpcode() == Instruction::GetElementPtr) {
if (AnalyzeGlobal(I, GS, PHIUsers)) return true;
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
// If the first two indices are constants, this can be SRA'd.
if (isa<GlobalVariable>(I->getOperand(0))) {
if (I->getNumOperands() < 3 || !isa<Constant>(I->getOperand(1)) ||
!cast<Constant>(I->getOperand(1))->isNullValue() ||
!isa<ConstantInt>(I->getOperand(2)))
GS.isNotSuitableForSRA = true;
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(I->getOperand(0))){
if (CE->getOpcode() != Instruction::GetElementPtr ||
CE->getNumOperands() < 3 || I->getNumOperands() < 2 ||
!isa<Constant>(I->getOperand(0)) ||
!cast<Constant>(I->getOperand(0))->isNullValue())
GS.isNotSuitableForSRA = true;
} else {
GS.isNotSuitableForSRA = true;
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
} else if (I->getOpcode() == Instruction::Select) {
if (AnalyzeGlobal(I, GS, PHIUsers)) return true;
GS.isNotSuitableForSRA = true;
} else if (PHINode *PN = dyn_cast<PHINode>(I)) {
// PHI nodes we can check just like select or GEP instructions, but we
// have to be careful about infinite recursion.
if (PHIUsers.insert(PN).second) // Not already visited.
if (AnalyzeGlobal(I, GS, PHIUsers)) return true;
GS.isNotSuitableForSRA = true;
} else if (isa<SetCondInst>(I)) {
GS.isNotSuitableForSRA = true;
} else {
return true; // Any other non-load instruction might take address!
}
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
} else if (Constant *C = dyn_cast<Constant>(*UI)) {
// We might have a dead and dangling constant hanging off of here.
if (!ConstantIsDead(C))
return true;
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
} else {
// Otherwise must be a global or some other user.
return true;
}
return false;
}
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
static Constant *getAggregateConstantElement(Constant *Agg, Constant *Idx) {
ConstantInt *CI = dyn_cast<ConstantInt>(Idx);
if (!CI) return 0;
uint64_t IdxV = CI->getRawValue();
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
if (ConstantStruct *CS = dyn_cast<ConstantStruct>(Agg)) {
if (IdxV < CS->getNumOperands()) return CS->getOperand(IdxV);
} else if (ConstantArray *CA = dyn_cast<ConstantArray>(Agg)) {
if (IdxV < CA->getNumOperands()) return CA->getOperand(IdxV);
} else if (ConstantPacked *CP = dyn_cast<ConstantPacked>(Agg)) {
if (IdxV < CP->getNumOperands()) return CP->getOperand(IdxV);
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
} else if (isa<ConstantAggregateZero>(Agg)) {
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
if (const StructType *STy = dyn_cast<StructType>(Agg->getType())) {
if (IdxV < STy->getNumElements())
return Constant::getNullValue(STy->getElementType(IdxV));
} else if (const SequentialType *STy =
dyn_cast<SequentialType>(Agg->getType())) {
return Constant::getNullValue(STy->getElementType());
}
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
} else if (isa<UndefValue>(Agg)) {
if (const StructType *STy = dyn_cast<StructType>(Agg->getType())) {
if (IdxV < STy->getNumElements())
return UndefValue::get(STy->getElementType(IdxV));
} else if (const SequentialType *STy =
dyn_cast<SequentialType>(Agg->getType())) {
return UndefValue::get(STy->getElementType());
}
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
}
return 0;
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
static Constant *TraverseGEPInitializer(User *GEP, Constant *Init) {
if (GEP->getNumOperands() == 1 ||
!isa<Constant>(GEP->getOperand(1)) ||
!cast<Constant>(GEP->getOperand(1))->isNullValue())
return 0;
for (unsigned i = 2, e = GEP->getNumOperands(); i != e; ++i) {
ConstantInt *Idx = dyn_cast<ConstantInt>(GEP->getOperand(i));
if (!Idx) return 0;
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
Init = getAggregateConstantElement(Init, Idx);
if (Init == 0) return 0;
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
}
return Init;
}
/// CleanupConstantGlobalUsers - We just marked GV constant. Loop over all
/// users of the global, cleaning up the obvious ones. This is largely just a
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
/// quick scan over the use list to clean up the easy and obvious cruft. This
/// returns true if it made a change.
static bool CleanupConstantGlobalUsers(Value *V, Constant *Init) {
bool Changed = false;
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E;) {
User *U = *UI++;
if (LoadInst *LI = dyn_cast<LoadInst>(U)) {
// Replace the load with the initializer.
LI->replaceAllUsesWith(Init);
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
LI->eraseFromParent();
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
Changed = true;
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
} else if (StoreInst *SI = dyn_cast<StoreInst>(U)) {
// Store must be unreachable or storing Init into the global.
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
SI->eraseFromParent();
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
Changed = true;
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(U)) {
if (CE->getOpcode() == Instruction::GetElementPtr) {
if (Constant *SubInit = TraverseGEPInitializer(CE, Init))
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
Changed |= CleanupConstantGlobalUsers(CE, SubInit);
if (CE->use_empty()) {
CE->destroyConstant();
Changed = true;
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
}
} else if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(U)) {
if (Constant *SubInit = TraverseGEPInitializer(GEP, Init))
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
Changed |= CleanupConstantGlobalUsers(GEP, SubInit);
Implement GlobalOpt/deadglobal-2.llx, deletion of globals that are only stored to, but are stored at variable indexes. This occurs at least in 176.gcc, but probably others, and we should handle it for completeness. llvm-svn: 16876
2004-10-11 00:47:33 +08:00
else {
// If this GEP has variable indexes, we should still be able to delete
// any stores through it.
for (Value::use_iterator GUI = GEP->use_begin(), E = GEP->use_end();
GUI != E;)
if (StoreInst *SI = dyn_cast<StoreInst>(*GUI++)) {
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
SI->eraseFromParent();
Implement GlobalOpt/deadglobal-2.llx, deletion of globals that are only stored to, but are stored at variable indexes. This occurs at least in 176.gcc, but probably others, and we should handle it for completeness. llvm-svn: 16876
2004-10-11 00:47:33 +08:00
Changed = true;
}
}
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
if (GEP->use_empty()) {
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GEP->eraseFromParent();
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
Changed = true;
}
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
} else if (Constant *C = dyn_cast<Constant>(U)) {
// If we have a chain of dead constantexprs or other things dangling from
// us, and if they are all dead, nuke them without remorse.
if (ConstantIsDead(C)) {
C->destroyConstant();
// This could have incalidated UI, start over from scratch.x
CleanupConstantGlobalUsers(V, Init);
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
return true;
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
}
}
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
return Changed;
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
}
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
/// SRAGlobal - Perform scalar replacement of aggregates on the specified global
/// variable. This opens the door for other optimizations by exposing the
/// behavior of the program in a more fine-grained way. We have determined that
/// this transformation is safe already. We return the first global variable we
/// insert so that the caller can reprocess it.
static GlobalVariable *SRAGlobal(GlobalVariable *GV) {
assert(GV->hasInternalLinkage() && !GV->isConstant());
Constant *Init = GV->getInitializer();
const Type *Ty = Init->getType();
std::vector<GlobalVariable*> NewGlobals;
Module::GlobalListType &Globals = GV->getParent()->getGlobalList();
if (const StructType *STy = dyn_cast<StructType>(Ty)) {
NewGlobals.reserve(STy->getNumElements());
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
Constant *In = getAggregateConstantElement(Init,
ConstantUInt::get(Type::UIntTy, i));
assert(In && "Couldn't get element of initializer?");
GlobalVariable *NGV = new GlobalVariable(STy->getElementType(i), false,
GlobalVariable::InternalLinkage,
In, GV->getName()+"."+utostr(i));
Globals.insert(GV, NGV);
NewGlobals.push_back(NGV);
}
} else if (const SequentialType *STy = dyn_cast<SequentialType>(Ty)) {
unsigned NumElements = 0;
if (const ArrayType *ATy = dyn_cast<ArrayType>(STy))
NumElements = ATy->getNumElements();
else if (const PackedType *PTy = dyn_cast<PackedType>(STy))
NumElements = PTy->getNumElements();
else
assert(0 && "Unknown aggregate sequential type!");
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
if (NumElements > 16 && GV->use_size() > 16) return 0; // It's not worth it.
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
NewGlobals.reserve(NumElements);
for (unsigned i = 0, e = NumElements; i != e; ++i) {
Constant *In = getAggregateConstantElement(Init,
ConstantUInt::get(Type::UIntTy, i));
assert(In && "Couldn't get element of initializer?");
GlobalVariable *NGV = new GlobalVariable(STy->getElementType(), false,
GlobalVariable::InternalLinkage,
In, GV->getName()+"."+utostr(i));
Globals.insert(GV, NGV);
NewGlobals.push_back(NGV);
}
}
if (NewGlobals.empty())
return 0;
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
DEBUG(std::cerr << "PERFORMING GLOBAL SRA ON: " << *GV);
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
Constant *NullInt = Constant::getNullValue(Type::IntTy);
// Loop over all of the uses of the global, replacing the constantexpr geps,
// with smaller constantexpr geps or direct references.
while (!GV->use_empty()) {
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
User *GEP = GV->use_back();
assert(((isa<ConstantExpr>(GEP) &&
cast<ConstantExpr>(GEP)->getOpcode()==Instruction::GetElementPtr)||
isa<GetElementPtrInst>(GEP)) && "NonGEP CE's are not SRAable!");
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
// Ignore the 1th operand, which has to be zero or else the program is quite
// broken (undefined). Get the 2nd operand, which is the structure or array
// index.
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
unsigned Val = cast<ConstantInt>(GEP->getOperand(2))->getRawValue();
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
if (Val >= NewGlobals.size()) Val = 0; // Out of bound array access.
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
Value *NewPtr = NewGlobals[Val];
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
// Form a shorter GEP if needed.
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
if (GEP->getNumOperands() > 3)
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(GEP)) {
std::vector<Constant*> Idxs;
Idxs.push_back(NullInt);
for (unsigned i = 3, e = CE->getNumOperands(); i != e; ++i)
Idxs.push_back(CE->getOperand(i));
NewPtr = ConstantExpr::getGetElementPtr(cast<Constant>(NewPtr), Idxs);
} else {
GetElementPtrInst *GEPI = cast<GetElementPtrInst>(GEP);
std::vector<Value*> Idxs;
Idxs.push_back(NullInt);
for (unsigned i = 3, e = GEPI->getNumOperands(); i != e; ++i)
Idxs.push_back(GEPI->getOperand(i));
NewPtr = new GetElementPtrInst(NewPtr, Idxs,
GEPI->getName()+"."+utostr(Val), GEPI);
}
GEP->replaceAllUsesWith(NewPtr);
if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(GEP))
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GEPI->eraseFromParent();
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
else
cast<ConstantExpr>(GEP)->destroyConstant();
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
}
We might as well delete the known-dead global sooner rather than later since we know it is dead. llvm-svn: 16855
2004-10-09 04:25:55 +08:00
// Delete the old global, now that it is dead.
Globals.erase(GV);
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
++NumSRA;
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
// Loop over the new globals array deleting any globals that are obviously
// dead. This can arise due to scalarization of a structure or an array that
// has elements that are dead.
unsigned FirstGlobal = 0;
for (unsigned i = 0, e = NewGlobals.size(); i != e; ++i)
if (NewGlobals[i]->use_empty()) {
Globals.erase(NewGlobals[i]);
if (FirstGlobal == i) ++FirstGlobal;
}
return FirstGlobal != NewGlobals.size() ? NewGlobals[FirstGlobal] : 0;
Implement SRA for global variables. This allows the other global variable optimizations to trigger much more often. This allows the elimination of several dozen more global variables in Programs/External. Note that we only do this for non-constant globals: constant globals will already be optimized out if the accesses to them permit it. This implements Transforms/GlobalOpt/globalsra.llx llvm-svn: 16842
2004-10-09 01:32:09 +08:00
}
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
/// AllUsesOfValueWillTrapIfNull - Return true if all users of the specified
/// value will trap if the value is dynamically null.
static bool AllUsesOfValueWillTrapIfNull(Value *V) {
for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E; ++UI)
if (isa<LoadInst>(*UI)) {
// Will trap.
} else if (StoreInst *SI = dyn_cast<StoreInst>(*UI)) {
if (SI->getOperand(0) == V) {
//std::cerr << "NONTRAPPING USE: " << **UI;
return false; // Storing the value.
}
} else if (CallInst *CI = dyn_cast<CallInst>(*UI)) {
if (CI->getOperand(0) != V) {
//std::cerr << "NONTRAPPING USE: " << **UI;
return false; // Not calling the ptr
}
} else if (InvokeInst *II = dyn_cast<InvokeInst>(*UI)) {
if (II->getOperand(0) != V) {
//std::cerr << "NONTRAPPING USE: " << **UI;
return false; // Not calling the ptr
}
} else if (CastInst *CI = dyn_cast<CastInst>(*UI)) {
if (!AllUsesOfValueWillTrapIfNull(CI)) return false;
} else if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(*UI)) {
if (!AllUsesOfValueWillTrapIfNull(GEPI)) return false;
*** empty log message *** llvm-svn: 17161
2004-10-22 14:43:28 +08:00
} else if (isa<SetCondInst>(*UI) &&
isa<ConstantPointerNull>(UI->getOperand(1))) {
// Ignore setcc X, null
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
} else {
//std::cerr << "NONTRAPPING USE: " << **UI;
return false;
}
return true;
}
/// AllUsesOfLoadedValueWillTrapIfNull - Return true if all uses of any loads
*** empty log message *** llvm-svn: 17161
2004-10-22 14:43:28 +08:00
/// from GV will trap if the loaded value is null. Note that this also permits
/// comparisons of the loaded value against null, as a special case.
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
static bool AllUsesOfLoadedValueWillTrapIfNull(GlobalVariable *GV) {
for (Value::use_iterator UI = GV->use_begin(), E = GV->use_end(); UI!=E; ++UI)
if (LoadInst *LI = dyn_cast<LoadInst>(*UI)) {
if (!AllUsesOfValueWillTrapIfNull(LI))
return false;
} else if (isa<StoreInst>(*UI)) {
// Ignore stores to the global.
} else {
// We don't know or understand this user, bail out.
//std::cerr << "UNKNOWN USER OF GLOBAL!: " << **UI;
return false;
}
return true;
}
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
static bool OptimizeAwayTrappingUsesOfValue(Value *V, Constant *NewV) {
bool Changed = false;
for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E; ) {
Instruction *I = cast<Instruction>(*UI++);
if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
LI->setOperand(0, NewV);
Changed = true;
} else if (StoreInst *SI = dyn_cast<StoreInst>(I)) {
if (SI->getOperand(1) == V) {
SI->setOperand(1, NewV);
Changed = true;
}
} else if (isa<CallInst>(I) || isa<InvokeInst>(I)) {
if (I->getOperand(0) == V) {
// Calling through the pointer! Turn into a direct call, but be careful
// that the pointer is not also being passed as an argument.
I->setOperand(0, NewV);
Changed = true;
bool PassedAsArg = false;
for (unsigned i = 1, e = I->getNumOperands(); i != e; ++i)
if (I->getOperand(i) == V) {
PassedAsArg = true;
I->setOperand(i, NewV);
}
if (PassedAsArg) {
// Being passed as an argument also. Be careful to not invalidate UI!
UI = V->use_begin();
}
}
} else if (CastInst *CI = dyn_cast<CastInst>(I)) {
Changed |= OptimizeAwayTrappingUsesOfValue(CI,
ConstantExpr::getCast(NewV, CI->getType()));
if (CI->use_empty()) {
Changed = true;
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
CI->eraseFromParent();
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
}
} else if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(I)) {
// Should handle GEP here.
std::vector<Constant*> Indices;
Indices.reserve(GEPI->getNumOperands()-1);
for (unsigned i = 1, e = GEPI->getNumOperands(); i != e; ++i)
if (Constant *C = dyn_cast<Constant>(GEPI->getOperand(i)))
Indices.push_back(C);
else
break;
if (Indices.size() == GEPI->getNumOperands()-1)
Changed |= OptimizeAwayTrappingUsesOfValue(GEPI,
ConstantExpr::getGetElementPtr(NewV, Indices));
if (GEPI->use_empty()) {
Changed = true;
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GEPI->eraseFromParent();
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
}
}
}
return Changed;
}
/// OptimizeAwayTrappingUsesOfLoads - The specified global has only one non-null
/// value stored into it. If there are uses of the loaded value that would trap
/// if the loaded value is dynamically null, then we know that they cannot be
/// reachable with a null optimize away the load.
static bool OptimizeAwayTrappingUsesOfLoads(GlobalVariable *GV, Constant *LV) {
std::vector<LoadInst*> Loads;
bool Changed = false;
// Replace all uses of loads with uses of uses of the stored value.
for (Value::use_iterator GUI = GV->use_begin(), E = GV->use_end();
GUI != E; ++GUI)
if (LoadInst *LI = dyn_cast<LoadInst>(*GUI)) {
Loads.push_back(LI);
Changed |= OptimizeAwayTrappingUsesOfValue(LI, LV);
} else {
assert(isa<StoreInst>(*GUI) && "Only expect load and stores!");
}
if (Changed) {
DEBUG(std::cerr << "OPTIMIZED LOADS FROM STORED ONCE POINTER: " << *GV);
++NumGlobUses;
}
// Delete all of the loads we can, keeping track of whether we nuked them all!
bool AllLoadsGone = true;
while (!Loads.empty()) {
LoadInst *L = Loads.back();
if (L->use_empty()) {
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
L->eraseFromParent();
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
Changed = true;
} else {
AllLoadsGone = false;
}
Loads.pop_back();
}
// If we nuked all of the loads, then none of the stores are needed either,
// nor is the global.
if (AllLoadsGone) {
DEBUG(std::cerr << " *** GLOBAL NOW DEAD!\n");
CleanupConstantGlobalUsers(GV, 0);
if (GV->use_empty()) {
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GV->eraseFromParent();
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
++NumDeleted;
}
Changed = true;
}
return Changed;
}
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
/// ConstantPropUsersOf - Walk the use list of V, constant folding all of the
/// instructions that are foldable.
static void ConstantPropUsersOf(Value *V) {
for (Value::use_iterator UI = V->use_begin(), E = V->use_end(); UI != E; )
if (Instruction *I = dyn_cast<Instruction>(*UI++))
if (Constant *NewC = ConstantFoldInstruction(I)) {
I->replaceAllUsesWith(NewC);
// Back up UI to avoid invalidating it!
bool AtBegin = false;
if (UI == V->use_begin())
AtBegin = true;
else
--UI;
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
I->eraseFromParent();
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
if (AtBegin)
UI = V->use_begin();
else
++UI;
}
}
/// OptimizeGlobalAddressOfMalloc - This function takes the specified global
/// variable, and transforms the program as if it always contained the result of
/// the specified malloc. Because it is always the result of the specified
/// malloc, there is no reason to actually DO the malloc. Instead, turn the
/// malloc into a global, and any laods of GV as uses of the new global.
static GlobalVariable *OptimizeGlobalAddressOfMalloc(GlobalVariable *GV,
MallocInst *MI) {
DEBUG(std::cerr << "PROMOTING MALLOC GLOBAL: " << *GV << " MALLOC = " <<*MI);
ConstantInt *NElements = cast<ConstantInt>(MI->getArraySize());
if (NElements->getRawValue() != 1) {
// If we have an array allocation, transform it to a single element
// allocation to make the code below simpler.
Type *NewTy = ArrayType::get(MI->getAllocatedType(),
NElements->getRawValue());
MallocInst *NewMI =
new MallocInst(NewTy, Constant::getNullValue(Type::UIntTy),
MI->getName(), MI);
std::vector<Value*> Indices;
Indices.push_back(Constant::getNullValue(Type::IntTy));
Indices.push_back(Indices[0]);
Value *NewGEP = new GetElementPtrInst(NewMI, Indices,
NewMI->getName()+".el0", MI);
MI->replaceAllUsesWith(NewGEP);
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
MI->eraseFromParent();
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
MI = NewMI;
}
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
// Create the new global variable. The contents of the malloc'd memory is
// undefined, so initialize with an undef value.
Constant *Init = UndefValue::get(MI->getAllocatedType());
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
GlobalVariable *NewGV = new GlobalVariable(MI->getAllocatedType(), false,
GlobalValue::InternalLinkage, Init,
GV->getName()+".body");
GV->getParent()->getGlobalList().insert(GV, NewGV);
// Anything that used the malloc now uses the global directly.
MI->replaceAllUsesWith(NewGV);
Constant *RepValue = NewGV;
if (NewGV->getType() != GV->getType()->getElementType())
RepValue = ConstantExpr::getCast(RepValue, GV->getType()->getElementType());
*** empty log message *** llvm-svn: 17161
2004-10-22 14:43:28 +08:00
// If there is a comparison against null, we will insert a global bool to
// keep track of whether the global was initialized yet or not.
GlobalVariable *InitBool = 0;
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
// Loop over all uses of GV, processing them in turn.
while (!GV->use_empty())
if (LoadInst *LI = dyn_cast<LoadInst>(GV->use_back())) {
*** empty log message *** llvm-svn: 17161
2004-10-22 14:43:28 +08:00
while (!LI->use_empty()) {
// FIXME: the iterator should expose a getUse() method.
Use &LoadUse = *(const iplist<Use>::iterator&)LI->use_begin();
if (!isa<SetCondInst>(LoadUse.getUser()))
LoadUse = RepValue;
else {
if (InitBool == 0) {
InitBool = new GlobalVariable(Type::BoolTy, false,
GlobalValue::InternalLinkage,
ConstantBool::False,
GV->getName()+".init");
GV->getParent()->getGlobalList().insert(GV, InitBool);
// The global is initialized when the malloc is run.
new StoreInst(ConstantBool::True, InitBool, MI);
}
// Replace the setcc X, 0 with a use of the bool value.
SetCondInst *SCI = cast<SetCondInst>(LoadUse.getUser());
Value *LV = new LoadInst(InitBool, InitBool->getName()+".val", SCI);
switch (SCI->getOpcode()) {
default: assert(0 && "Unknown opcode!");
case Instruction::SetLT:
LV = ConstantBool::False; // X < null -> always false
break;
case Instruction::SetEQ:
case Instruction::SetLE:
LV = BinaryOperator::createNot(LV, "notinit", SCI);
break;
case Instruction::SetNE:
case Instruction::SetGE:
case Instruction::SetGT:
break; // no change.
}
SCI->replaceAllUsesWith(LV);
SCI->eraseFromParent();
}
}
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
LI->eraseFromParent();
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
} else {
StoreInst *SI = cast<StoreInst>(GV->use_back());
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
SI->eraseFromParent();
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
}
*** empty log message *** llvm-svn: 17161
2004-10-22 14:43:28 +08:00
// Now the GV is dead, nuke it and the malloc.
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GV->eraseFromParent();
*** empty log message *** llvm-svn: 17161
2004-10-22 14:43:28 +08:00
MI->eraseFromParent();
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
// To further other optimizations, loop over all users of NewGV and try to
// constant prop them. This will promote GEP instructions with constant
// indices into GEP constant-exprs, which will allow global-opt to hack on it.
ConstantPropUsersOf(NewGV);
if (RepValue != NewGV)
ConstantPropUsersOf(RepValue);
return NewGV;
}
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
// OptimizeOnceStoredGlobal - Try to optimize globals based on the knowledge
// that only one value (besides its initializer) is ever stored to the global.
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
static bool OptimizeOnceStoredGlobal(GlobalVariable *GV, Value *StoredOnceVal,
Module::giterator &GVI, TargetData &TD) {
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
if (CastInst *CI = dyn_cast<CastInst>(StoredOnceVal))
StoredOnceVal = CI->getOperand(0);
else if (GetElementPtrInst *GEPI =dyn_cast<GetElementPtrInst>(StoredOnceVal)){
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
// "getelementptr Ptr, 0, 0, 0" is really just a cast.
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
bool IsJustACast = true;
for (unsigned i = 1, e = GEPI->getNumOperands(); i != e; ++i)
if (!isa<Constant>(GEPI->getOperand(i)) ||
!cast<Constant>(GEPI->getOperand(i))->isNullValue()) {
IsJustACast = false;
break;
}
if (IsJustACast)
StoredOnceVal = GEPI->getOperand(0);
}
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
// If we are dealing with a pointer global that is initialized to null and
// only has one (non-null) value stored into it, then we can optimize any
// users of the loaded value (often calls and loads) that would trap if the
// value was null.
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
if (isa<PointerType>(GV->getInitializer()->getType()) &&
GV->getInitializer()->isNullValue()) {
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
if (Constant *SOVC = dyn_cast<Constant>(StoredOnceVal)) {
if (GV->getInitializer()->getType() != SOVC->getType())
SOVC = ConstantExpr::getCast(SOVC, GV->getInitializer()->getType());
// Optimize away any trapping uses of the loaded value.
if (OptimizeAwayTrappingUsesOfLoads(GV, SOVC))
Fix 2004-10-10-CastStoreOnce.llx, by adjusting types back if we strip off a cast llvm-svn: 16878
2004-10-11 01:07:12 +08:00
return true;
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
} else if (MallocInst *MI = dyn_cast<MallocInst>(StoredOnceVal)) {
// If we have a global that is only initialized with a fixed size malloc,
// and if all users of the malloc trap, and if the malloc'd address is not
// put anywhere else, transform the program to use global memory instead
// of malloc'd memory. This eliminates dynamic allocation (good) and
// exposes the resultant global to further GlobalOpt (even better). Note
// that we restrict this transformation to only working on small
// allocations (2048 bytes currently), as we don't want to introduce a 16M
// global or something.
if (ConstantInt *NElements = dyn_cast<ConstantInt>(MI->getArraySize()))
if (MI->getAllocatedType()->isSized() &&
NElements->getRawValue()*
TD.getTypeSize(MI->getAllocatedType()) < 2048 &&
AllUsesOfLoadedValueWillTrapIfNull(GV)) {
// FIXME: do more correctness checking to make sure the result of the
// malloc isn't squirrelled away somewhere.
GVI = OptimizeGlobalAddressOfMalloc(GV, MI);
return true;
}
Just because we cannot completely eliminate all uses of a global, we can still optimize away all of the indirect calls and loads, etc from it. This turns code like this: if (G != 0) G(); into if (G != 0) ActualCallee(); This triggers a couple of times in gcc and libstdc++. llvm-svn: 16901
2004-10-11 07:14:11 +08:00
}
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
}
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
return false;
}
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
/// ProcessInternalGlobal - Analyze the specified global variable and optimize
/// it if possible. If we make a change, return true.
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
bool GlobalOpt::ProcessInternalGlobal(GlobalVariable *GV,
Module::giterator &GVI) {
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
std::set<PHINode*> PHIUsers;
GlobalStatus GS;
PHIUsers.clear();
GV->removeDeadConstantUsers();
if (GV->use_empty()) {
DEBUG(std::cerr << "GLOBAL DEAD: " << *GV);
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GV->eraseFromParent();
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
++NumDeleted;
return true;
}
if (!AnalyzeGlobal(GV, GS, PHIUsers)) {
// If the global is never loaded (but may be stored to), it is dead.
// Delete it now.
if (!GS.isLoaded) {
DEBUG(std::cerr << "GLOBAL NEVER LOADED: " << *GV);
Fix infinite loop due to iteration llvm-svn: 16864
2004-10-09 11:32:52 +08:00
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
// Delete any stores we can find to the global. We may not be able to
// make it completely dead though.
Avoid calling use_size() which could (in theory) be expensive if the global has a large number of users. Instead, just keep track of whether we're making changes as we do so. This patch has no functionlity changes. llvm-svn: 16874
2004-10-11 00:43:46 +08:00
bool Changed = CleanupConstantGlobalUsers(GV, GV->getInitializer());
Fix infinite loop due to iteration llvm-svn: 16864
2004-10-09 11:32:52 +08:00
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
// If the global is dead now, delete it.
if (GV->use_empty()) {
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GV->eraseFromParent();
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
++NumDeleted;
Fix infinite loop due to iteration llvm-svn: 16864
2004-10-09 11:32:52 +08:00
Changed = true;
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
}
Fix infinite loop due to iteration llvm-svn: 16864
2004-10-09 11:32:52 +08:00
return Changed;
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
} else if (GS.StoredType <= GlobalStatus::isInitializerStored) {
DEBUG(std::cerr << "MARKING CONSTANT: " << *GV);
GV->setConstant(true);
// Clean up any obviously simplifiable users now.
CleanupConstantGlobalUsers(GV, GV->getInitializer());
// If the global is dead now, just nuke it.
if (GV->use_empty()) {
DEBUG(std::cerr << " *** Marking constant allowed us to simplify "
"all users and delete global!\n");
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
GV->eraseFromParent();
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
++NumDeleted;
}
++NumMarked;
return true;
} else if (!GS.isNotSuitableForSRA &&
!GV->getInitializer()->getType()->isFirstClassType()) {
if (GlobalVariable *FirstNewGV = SRAGlobal(GV)) {
GVI = FirstNewGV; // Don't skip the newly produced globals!
return true;
}
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
} else if (GS.StoredType == GlobalStatus::isStoredOnce) {
Add support for the undef value. Implement a new optimization based on globals that are initialized with undef. When promoting malloc to a global, start out initialized to undef llvm-svn: 17042
2004-10-17 02:09:00 +08:00
// If the initial value for the global was an undef value, and if only one
// other value was stored into it, we can just change the initializer to
// be an undef value, then delete all stores to the global. This allows
// us to mark it constant.
if (isa<UndefValue>(GV->getInitializer()) &&
isa<Constant>(GS.StoredOnceValue)) {
// Change the initial value here.
GV->setInitializer(cast<Constant>(GS.StoredOnceValue));
// Clean up any obviously simplifiable users now.
CleanupConstantGlobalUsers(GV, GV->getInitializer());
if (GV->use_empty()) {
DEBUG(std::cerr << " *** Substituting initializer allowed us to "
"simplify all users and delete global!\n");
GV->eraseFromParent();
++NumDeleted;
} else {
GVI = GV;
}
++NumSubstitute;
return true;
}
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
// Try to optimize globals based on the knowledge that only one value
// (besides its initializer) is ever stored to the global.
This patch implements two things (sorry). First, it allows SRA of globals that have embedded arrays, implementing GlobalOpt/globalsra-partial.llx. This comes up infrequently, but does allow, for example, deleting several stores to dead parts of globals in dhrystone. Second, this implements GlobalOpt/malloc-promote-*.llx, which is the following nifty transformation: Basically if a global pointer is initialized with malloc, and we can tell that the program won't notice, we transform this: struct foo *FooPtr; ... FooPtr = malloc(sizeof(struct foo)); ... FooPtr->A FooPtr->B Into: struct foo FooPtrBody; ... FooPtrBody.A FooPtrBody.B This comes up occasionally, for example, the 'disp' global in 183.equake (where the xform speeds the CBE version of the program up from 56.16s to 52.40s (7%) on apoc), and the 'desired_accept', 'fixLRBT', 'macroArray', & 'key_queue' globals in 300.twolf (speeding it up from 22.29s to 21.55s (3.4%)). The nice thing about this xform is that it exposes the resulting global to global variable optimization and makes alias analysis easier in addition to eliminating a few loads. llvm-svn: 16916
2004-10-11 13:54:41 +08:00
if (OptimizeOnceStoredGlobal(GV, GS.StoredOnceValue, GVI,
getAnalysis<TargetData>()))
Eliminate global pointers that are only stored a single value and null if we know that all uses of the global will trap if the pointer contained is null. In this case, we forward substitute the stored value to any uses. This has the effect of devirtualizing trivial globals in trivial cases. For example, 164.gzip contains this: gzip.h:extern int (*read_buf) OF((char *buf, unsigned size)); bits.c: read_buf = file_read; deflate.c: lookahead = read_buf((char*)window, deflate.c: n = read_buf((char*)window+strstart+lookahead, more); Since read_buf has to point to file_read at every use, we just replace the calls through read_buf with a direct call to file_read. This occurs in several benchmarks, including 176.gcc and 164.gzip. Direct calls are good and stuff. llvm-svn: 16871
2004-10-10 05:48:45 +08:00
return true;
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
}
}
return false;
}
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
bool GlobalOpt::runOnModule(Module &M) {
bool Changed = false;
// As a prepass, delete functions that are trivially dead.
bool LocalChange = true;
while (LocalChange) {
LocalChange = false;
for (Module::iterator FI = M.begin(), E = M.end(); FI != E; ) {
Function *F = FI++;
F->removeDeadConstantUsers();
Fix a bug John tracked down in libstdc++ where we were incorrectly deleting weak functions. Thanks for finding this John! llvm-svn: 16997
2004-10-15 03:53:50 +08:00
if (F->use_empty() && (F->hasInternalLinkage() ||
F->hasLinkOnceLinkage())) {
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
M.getFunctionList().erase(F);
LocalChange = true;
++NumFnDeleted;
}
}
Changed |= LocalChange;
}
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
LocalChange = true;
while (LocalChange) {
LocalChange = false;
for (Module::giterator GVI = M.gbegin(), E = M.gend(); GVI != E;) {
GlobalVariable *GV = GVI++;
if (!GV->isConstant() && GV->hasInternalLinkage() &&
GV->hasInitializer())
LocalChange |= ProcessInternalGlobal(GV, GVI);
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
}
* Pull out the meat of runOnModule into another function for clarity. * Do not lead dangling dead constants prevent optimization * Iterate global optimization while we're making progress. These changes allow us to be more aggressive, handling cases like GlobalOpt/iterate.llx without a problem (turning it into 'ret int 0'). llvm-svn: 16857
2004-10-09 04:59:28 +08:00
Changed |= LocalChange;
* Rename pass to globalopt, since we do more than just constify * Instead of handling dead functions specially, just nuke them. * Be more aggressive about cleaning up after constification, in particular, handle getelementptr instructions and constantexprs. * Be a little bit more structured about how we process globals. *** Delete globals that are only stored to, and never read. These are clearly not useful, so they should go. This implements deadglobal.llx This last one triggers quite a few times. In particular, 2208 in the external tests, 1865 of which are in 252.eon. This shrinks eon from 1995094 to 1732341 bytes of bytecode. llvm-svn: 16802
2004-10-07 12:16:33 +08:00
}
return Changed;
}