2007-03-31 04:06:36 +00:00
|
|
|
//===- CodeGenPrepare.cpp - Prepare a function for code generation --------===//
|
|
|
|
//
|
|
|
|
// The LLVM Compiler Infrastructure
|
|
|
|
//
|
2007-12-29 20:36:04 +00:00
|
|
|
// This file is distributed under the University of Illinois Open Source
|
|
|
|
// License. See LICENSE.TXT for details.
|
2007-03-31 04:06:36 +00:00
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
//
|
|
|
|
// This pass munges the code in the input function to better prepare it for
|
2008-05-08 17:46:35 +00:00
|
|
|
// SelectionDAG-based code generation. This works around limitations in it's
|
|
|
|
// basic-block-at-a-time approach. It should eventually be removed.
|
2007-03-31 04:06:36 +00:00
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
|
|
|
#define DEBUG_TYPE "codegenprepare"
|
|
|
|
#include "llvm/Transforms/Scalar.h"
|
|
|
|
#include "llvm/Constants.h"
|
|
|
|
#include "llvm/DerivedTypes.h"
|
|
|
|
#include "llvm/Function.h"
|
2008-02-26 02:42:37 +00:00
|
|
|
#include "llvm/InlineAsm.h"
|
2007-03-31 04:06:36 +00:00
|
|
|
#include "llvm/Instructions.h"
|
2009-03-26 01:15:07 +00:00
|
|
|
#include "llvm/IntrinsicInst.h"
|
2009-07-03 00:17:18 +00:00
|
|
|
#include "llvm/LLVMContext.h"
|
2007-03-31 04:06:36 +00:00
|
|
|
#include "llvm/Pass.h"
|
|
|
|
#include "llvm/Target/TargetData.h"
|
|
|
|
#include "llvm/Target/TargetLowering.h"
|
2009-02-20 18:24:38 +00:00
|
|
|
#include "llvm/Transforms/Utils/AddrModeMatcher.h"
|
2007-03-31 04:06:36 +00:00
|
|
|
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
#include "llvm/Transforms/Utils/Local.h"
|
|
|
|
#include "llvm/ADT/DenseMap.h"
|
2007-03-31 04:06:36 +00:00
|
|
|
#include "llvm/ADT/SmallSet.h"
|
2009-02-13 17:45:12 +00:00
|
|
|
#include "llvm/Assembly/Writer.h"
|
2008-02-26 02:42:37 +00:00
|
|
|
#include "llvm/Support/CallSite.h"
|
2008-12-19 18:03:11 +00:00
|
|
|
#include "llvm/Support/CommandLine.h"
|
2007-04-02 01:35:34 +00:00
|
|
|
#include "llvm/Support/Compiler.h"
|
2007-12-05 23:58:20 +00:00
|
|
|
#include "llvm/Support/Debug.h"
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
#include "llvm/Support/GetElementPtrTypeIterator.h"
|
2008-11-25 04:42:10 +00:00
|
|
|
#include "llvm/Support/PatternMatch.h"
|
2009-07-25 01:13:51 +00:00
|
|
|
#include "llvm/Support/raw_ostream.h"
|
2007-03-31 04:06:36 +00:00
|
|
|
using namespace llvm;
|
2008-11-25 04:42:10 +00:00
|
|
|
using namespace llvm::PatternMatch;
|
2007-03-31 04:06:36 +00:00
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
static cl::opt<bool> FactorCommonPreds("split-critical-paths-tweak",
|
|
|
|
cl::init(false), cl::Hidden);
|
|
|
|
|
2008-09-24 05:32:41 +00:00
|
|
|
namespace {
|
2007-03-31 04:06:36 +00:00
|
|
|
class VISIBILITY_HIDDEN CodeGenPrepare : public FunctionPass {
|
|
|
|
/// TLI - Keep a pointer of a TargetLowering to consult for determining
|
|
|
|
/// transformation profitability.
|
|
|
|
const TargetLowering *TLI;
|
2008-12-19 18:03:11 +00:00
|
|
|
|
|
|
|
/// BackEdges - Keep a set of all the loop back edges.
|
|
|
|
///
|
2009-05-04 18:40:41 +00:00
|
|
|
SmallSet<std::pair<const BasicBlock*, const BasicBlock*>, 8> BackEdges;
|
2007-03-31 04:06:36 +00:00
|
|
|
public:
|
2007-05-06 13:37:16 +00:00
|
|
|
static char ID; // Pass identification, replacement for typeid
|
2007-08-01 15:32:29 +00:00
|
|
|
explicit CodeGenPrepare(const TargetLowering *tli = 0)
|
2008-09-04 17:05:41 +00:00
|
|
|
: FunctionPass(&ID), TLI(tli) {}
|
2007-03-31 04:06:36 +00:00
|
|
|
bool runOnFunction(Function &F);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
private:
|
2007-04-02 01:35:34 +00:00
|
|
|
bool EliminateMostlyEmptyBlocks(Function &F);
|
|
|
|
bool CanMergeBlocks(const BasicBlock *BB, const BasicBlock *DestBB) const;
|
|
|
|
void EliminateMostlyEmptyBlock(BasicBlock *BB);
|
2007-03-31 04:06:36 +00:00
|
|
|
bool OptimizeBlock(BasicBlock &BB);
|
2008-11-25 07:09:13 +00:00
|
|
|
bool OptimizeMemoryInst(Instruction *I, Value *Addr, const Type *AccessTy,
|
|
|
|
DenseMap<Value*,Value*> &SunkAddrs);
|
2008-02-26 02:42:37 +00:00
|
|
|
bool OptimizeInlineAsmInst(Instruction *I, CallSite CS,
|
|
|
|
DenseMap<Value*,Value*> &SunkAddrs);
|
2007-12-05 23:58:20 +00:00
|
|
|
bool OptimizeExtUses(Instruction *I);
|
2009-05-04 18:40:41 +00:00
|
|
|
void findLoopBackEdges(const Function &F);
|
2007-03-31 04:06:36 +00:00
|
|
|
};
|
|
|
|
}
|
2007-05-01 21:15:47 +00:00
|
|
|
|
2007-05-03 01:11:54 +00:00
|
|
|
char CodeGenPrepare::ID = 0;
|
2007-03-31 04:06:36 +00:00
|
|
|
static RegisterPass<CodeGenPrepare> X("codegenprepare",
|
|
|
|
"Optimize for code generation");
|
|
|
|
|
|
|
|
FunctionPass *llvm::createCodeGenPreparePass(const TargetLowering *TLI) {
|
|
|
|
return new CodeGenPrepare(TLI);
|
|
|
|
}
|
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
/// findLoopBackEdges - Do a DFS walk to find loop back edges.
|
|
|
|
///
|
2009-05-04 18:40:41 +00:00
|
|
|
void CodeGenPrepare::findLoopBackEdges(const Function &F) {
|
|
|
|
SmallVector<std::pair<const BasicBlock*,const BasicBlock*>, 32> Edges;
|
|
|
|
FindFunctionBackedges(F, Edges);
|
|
|
|
|
|
|
|
BackEdges.insert(Edges.begin(), Edges.end());
|
2008-12-19 18:03:11 +00:00
|
|
|
}
|
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
|
|
|
|
bool CodeGenPrepare::runOnFunction(Function &F) {
|
|
|
|
bool EverMadeChange = false;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// First pass, eliminate blocks that contain only PHI nodes and an
|
|
|
|
// unconditional branch.
|
|
|
|
EverMadeChange |= EliminateMostlyEmptyBlocks(F);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2009-01-05 21:17:27 +00:00
|
|
|
// Now find loop back edges.
|
|
|
|
findLoopBackEdges(F);
|
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
bool MadeChange = true;
|
2007-03-31 04:06:36 +00:00
|
|
|
while (MadeChange) {
|
|
|
|
MadeChange = false;
|
|
|
|
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
|
|
|
|
MadeChange |= OptimizeBlock(*BB);
|
|
|
|
EverMadeChange |= MadeChange;
|
|
|
|
}
|
|
|
|
return EverMadeChange;
|
|
|
|
}
|
|
|
|
|
2009-03-27 01:13:37 +00:00
|
|
|
/// EliminateMostlyEmptyBlocks - eliminate blocks that contain only PHI nodes,
|
|
|
|
/// debug info directives, and an unconditional branch. Passes before isel
|
|
|
|
/// (e.g. LSR/loopsimplify) often split edges in ways that are non-optimal for
|
|
|
|
/// isel. Start by eliminating these blocks so we can split them the way we
|
|
|
|
/// want them.
|
2007-04-02 01:35:34 +00:00
|
|
|
bool CodeGenPrepare::EliminateMostlyEmptyBlocks(Function &F) {
|
|
|
|
bool MadeChange = false;
|
|
|
|
// Note that this intentionally skips the entry block.
|
|
|
|
for (Function::iterator I = ++F.begin(), E = F.end(); I != E; ) {
|
|
|
|
BasicBlock *BB = I++;
|
|
|
|
|
|
|
|
// If this block doesn't end with an uncond branch, ignore it.
|
|
|
|
BranchInst *BI = dyn_cast<BranchInst>(BB->getTerminator());
|
|
|
|
if (!BI || !BI->isUnconditional())
|
|
|
|
continue;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2009-03-27 01:13:37 +00:00
|
|
|
// If the instruction before the branch (skipping debug info) isn't a phi
|
|
|
|
// node, then other stuff is happening here.
|
2007-04-02 01:35:34 +00:00
|
|
|
BasicBlock::iterator BBI = BI;
|
|
|
|
if (BBI != BB->begin()) {
|
|
|
|
--BBI;
|
2009-03-27 01:13:37 +00:00
|
|
|
while (isa<DbgInfoIntrinsic>(BBI)) {
|
|
|
|
if (BBI == BB->begin())
|
|
|
|
break;
|
|
|
|
--BBI;
|
|
|
|
}
|
|
|
|
if (!isa<DbgInfoIntrinsic>(BBI) && !isa<PHINode>(BBI))
|
|
|
|
continue;
|
2007-04-02 01:35:34 +00:00
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// Do not break infinite loops.
|
|
|
|
BasicBlock *DestBB = BI->getSuccessor(0);
|
|
|
|
if (DestBB == BB)
|
|
|
|
continue;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
if (!CanMergeBlocks(BB, DestBB))
|
|
|
|
continue;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
EliminateMostlyEmptyBlock(BB);
|
|
|
|
MadeChange = true;
|
|
|
|
}
|
|
|
|
return MadeChange;
|
|
|
|
}
|
|
|
|
|
|
|
|
/// CanMergeBlocks - Return true if we can merge BB into DestBB if there is a
|
|
|
|
/// single uncond branch between them, and BB contains no other non-phi
|
|
|
|
/// instructions.
|
|
|
|
bool CodeGenPrepare::CanMergeBlocks(const BasicBlock *BB,
|
|
|
|
const BasicBlock *DestBB) const {
|
|
|
|
// We only want to eliminate blocks whose phi nodes are used by phi nodes in
|
|
|
|
// the successor. If there are more complex condition (e.g. preheaders),
|
|
|
|
// don't mess around with them.
|
|
|
|
BasicBlock::const_iterator BBI = BB->begin();
|
|
|
|
while (const PHINode *PN = dyn_cast<PHINode>(BBI++)) {
|
|
|
|
for (Value::use_const_iterator UI = PN->use_begin(), E = PN->use_end();
|
|
|
|
UI != E; ++UI) {
|
|
|
|
const Instruction *User = cast<Instruction>(*UI);
|
|
|
|
if (User->getParent() != DestBB || !isa<PHINode>(User))
|
|
|
|
return false;
|
2008-09-24 05:32:41 +00:00
|
|
|
// If User is inside DestBB block and it is a PHINode then check
|
|
|
|
// incoming value. If incoming value is not from BB then this is
|
2007-04-25 00:37:04 +00:00
|
|
|
// a complex condition (e.g. preheaders) we want to avoid here.
|
|
|
|
if (User->getParent() == DestBB) {
|
|
|
|
if (const PHINode *UPN = dyn_cast<PHINode>(User))
|
|
|
|
for (unsigned I = 0, E = UPN->getNumIncomingValues(); I != E; ++I) {
|
|
|
|
Instruction *Insn = dyn_cast<Instruction>(UPN->getIncomingValue(I));
|
|
|
|
if (Insn && Insn->getParent() == BB &&
|
|
|
|
Insn->getParent() != UPN->getIncomingBlock(I))
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
}
|
2007-04-02 01:35:34 +00:00
|
|
|
}
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// If BB and DestBB contain any common predecessors, then the phi nodes in BB
|
|
|
|
// and DestBB may have conflicting incoming values for the block. If so, we
|
|
|
|
// can't merge the block.
|
|
|
|
const PHINode *DestBBPN = dyn_cast<PHINode>(DestBB->begin());
|
|
|
|
if (!DestBBPN) return true; // no conflict.
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// Collect the preds of BB.
|
2007-11-06 22:07:40 +00:00
|
|
|
SmallPtrSet<const BasicBlock*, 16> BBPreds;
|
2007-04-02 01:35:34 +00:00
|
|
|
if (const PHINode *BBPN = dyn_cast<PHINode>(BB->begin())) {
|
|
|
|
// It is faster to get preds from a PHI than with pred_iterator.
|
|
|
|
for (unsigned i = 0, e = BBPN->getNumIncomingValues(); i != e; ++i)
|
|
|
|
BBPreds.insert(BBPN->getIncomingBlock(i));
|
|
|
|
} else {
|
|
|
|
BBPreds.insert(pred_begin(BB), pred_end(BB));
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// Walk the preds of DestBB.
|
|
|
|
for (unsigned i = 0, e = DestBBPN->getNumIncomingValues(); i != e; ++i) {
|
|
|
|
BasicBlock *Pred = DestBBPN->getIncomingBlock(i);
|
|
|
|
if (BBPreds.count(Pred)) { // Common predecessor?
|
|
|
|
BBI = DestBB->begin();
|
|
|
|
while (const PHINode *PN = dyn_cast<PHINode>(BBI++)) {
|
|
|
|
const Value *V1 = PN->getIncomingValueForBlock(Pred);
|
|
|
|
const Value *V2 = PN->getIncomingValueForBlock(BB);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// If V2 is a phi node in BB, look up what the mapped value will be.
|
|
|
|
if (const PHINode *V2PN = dyn_cast<PHINode>(V2))
|
|
|
|
if (V2PN->getParent() == BB)
|
|
|
|
V2 = V2PN->getIncomingValueForBlock(Pred);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// If there is a conflict, bail out.
|
|
|
|
if (V1 != V2) return false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/// EliminateMostlyEmptyBlock - Eliminate a basic block that have only phi's and
|
|
|
|
/// an unconditional branch in it.
|
|
|
|
void CodeGenPrepare::EliminateMostlyEmptyBlock(BasicBlock *BB) {
|
|
|
|
BranchInst *BI = cast<BranchInst>(BB->getTerminator());
|
|
|
|
BasicBlock *DestBB = BI->getSuccessor(0);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2009-08-23 04:37:46 +00:00
|
|
|
DEBUG(errs() << "MERGING MOSTLY EMPTY BLOCKS - BEFORE:\n" << *BB << *DestBB);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// If the destination block has a single pred, then this is a trivial edge,
|
|
|
|
// just collapse it.
|
2008-11-27 19:29:14 +00:00
|
|
|
if (BasicBlock *SinglePred = DestBB->getSinglePredecessor()) {
|
2008-11-28 19:54:49 +00:00
|
|
|
if (SinglePred != DestBB) {
|
|
|
|
// Remember if SinglePred was the entry block of the function. If so, we
|
|
|
|
// will need to move BB back to the entry position.
|
|
|
|
bool isEntry = SinglePred == &SinglePred->getParent()->getEntryBlock();
|
|
|
|
MergeBasicBlockIntoOnlyPred(DestBB);
|
|
|
|
|
|
|
|
if (isEntry && BB != &BB->getParent()->getEntryBlock())
|
|
|
|
BB->moveBefore(&BB->getParent()->getEntryBlock());
|
|
|
|
|
2009-08-23 04:37:46 +00:00
|
|
|
DEBUG(errs() << "AFTER:\n" << *DestBB << "\n\n\n");
|
2008-11-28 19:54:49 +00:00
|
|
|
return;
|
|
|
|
}
|
2007-04-02 01:35:34 +00:00
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// Otherwise, we have multiple predecessors of BB. Update the PHIs in DestBB
|
|
|
|
// to handle the new incoming edges it is about to have.
|
|
|
|
PHINode *PN;
|
|
|
|
for (BasicBlock::iterator BBI = DestBB->begin();
|
|
|
|
(PN = dyn_cast<PHINode>(BBI)); ++BBI) {
|
|
|
|
// Remove the incoming value for BB, and remember it.
|
|
|
|
Value *InVal = PN->removeIncomingValue(BB, false);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// Two options: either the InVal is a phi node defined in BB or it is some
|
|
|
|
// value that dominates BB.
|
|
|
|
PHINode *InValPhi = dyn_cast<PHINode>(InVal);
|
|
|
|
if (InValPhi && InValPhi->getParent() == BB) {
|
|
|
|
// Add all of the input values of the input PHI as inputs of this phi.
|
|
|
|
for (unsigned i = 0, e = InValPhi->getNumIncomingValues(); i != e; ++i)
|
|
|
|
PN->addIncoming(InValPhi->getIncomingValue(i),
|
|
|
|
InValPhi->getIncomingBlock(i));
|
|
|
|
} else {
|
|
|
|
// Otherwise, add one instance of the dominating value for each edge that
|
|
|
|
// we will be adding.
|
|
|
|
if (PHINode *BBPN = dyn_cast<PHINode>(BB->begin())) {
|
|
|
|
for (unsigned i = 0, e = BBPN->getNumIncomingValues(); i != e; ++i)
|
|
|
|
PN->addIncoming(InVal, BBPN->getIncomingBlock(i));
|
|
|
|
} else {
|
|
|
|
for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI)
|
|
|
|
PN->addIncoming(InVal, *PI);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-04-02 01:35:34 +00:00
|
|
|
// The PHIs are now updated, change everything that refers to BB to use
|
|
|
|
// DestBB and remove BB.
|
|
|
|
BB->replaceAllUsesWith(DestBB);
|
|
|
|
BB->eraseFromParent();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2009-08-23 04:37:46 +00:00
|
|
|
DEBUG(errs() << "AFTER:\n" << *DestBB << "\n\n\n");
|
2007-04-02 01:35:34 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-12-24 19:32:55 +00:00
|
|
|
/// SplitEdgeNicely - Split the critical edge from TI to its specified
|
2007-03-31 04:06:36 +00:00
|
|
|
/// successor if it will improve codegen. We only do this if the successor has
|
|
|
|
/// phi nodes (otherwise critical edges are ok). If there is already another
|
|
|
|
/// predecessor of the succ that is empty (and thus has no phi nodes), use it
|
|
|
|
/// instead of introducing a new block.
|
2008-12-19 18:03:11 +00:00
|
|
|
static void SplitEdgeNicely(TerminatorInst *TI, unsigned SuccNum,
|
2009-05-04 18:40:41 +00:00
|
|
|
SmallSet<std::pair<const BasicBlock*,
|
|
|
|
const BasicBlock*>, 8> &BackEdges,
|
2008-12-19 18:03:11 +00:00
|
|
|
Pass *P) {
|
2007-03-31 04:06:36 +00:00
|
|
|
BasicBlock *TIBB = TI->getParent();
|
|
|
|
BasicBlock *Dest = TI->getSuccessor(SuccNum);
|
|
|
|
assert(isa<PHINode>(Dest->begin()) &&
|
|
|
|
"This should only be called if Dest has a PHI!");
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2009-03-13 22:59:14 +00:00
|
|
|
// Do not split edges to EH landing pads.
|
|
|
|
if (InvokeInst *Invoke = dyn_cast<InvokeInst>(TI)) {
|
|
|
|
if (Invoke->getSuccessor(1) == Dest)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2007-12-24 19:32:55 +00:00
|
|
|
// As a hack, never split backedges of loops. Even though the copy for any
|
|
|
|
// PHIs inserted on the backedge would be dead for exits from the loop, we
|
|
|
|
// assume that the cost of *splitting* the backedge would be too high.
|
2008-12-19 18:03:11 +00:00
|
|
|
if (BackEdges.count(std::make_pair(TIBB, Dest)))
|
2007-12-24 19:32:55 +00:00
|
|
|
return;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
if (!FactorCommonPreds) {
|
|
|
|
/// TIPHIValues - This array is lazily computed to determine the values of
|
|
|
|
/// PHIs in Dest that TI would provide.
|
|
|
|
SmallVector<Value*, 32> TIPHIValues;
|
|
|
|
|
|
|
|
// Check to see if Dest has any blocks that can be used as a split edge for
|
|
|
|
// this terminator.
|
|
|
|
for (pred_iterator PI = pred_begin(Dest), E = pred_end(Dest); PI != E; ++PI) {
|
|
|
|
BasicBlock *Pred = *PI;
|
|
|
|
// To be usable, the pred has to end with an uncond branch to the dest.
|
|
|
|
BranchInst *PredBr = dyn_cast<BranchInst>(Pred->getTerminator());
|
2009-03-26 01:15:07 +00:00
|
|
|
if (!PredBr || !PredBr->isUnconditional())
|
|
|
|
continue;
|
|
|
|
// Must be empty other than the branch and debug info.
|
|
|
|
BasicBlock::iterator I = Pred->begin();
|
|
|
|
while (isa<DbgInfoIntrinsic>(I))
|
|
|
|
I++;
|
|
|
|
if (dyn_cast<Instruction>(I) != PredBr)
|
|
|
|
continue;
|
|
|
|
// Cannot be the entry block; its label does not get emitted.
|
|
|
|
if (Pred == &(Dest->getParent()->getEntryBlock()))
|
2008-12-19 18:03:11 +00:00
|
|
|
continue;
|
|
|
|
|
|
|
|
// Finally, since we know that Dest has phi nodes in it, we have to make
|
2009-03-26 01:15:07 +00:00
|
|
|
// sure that jumping to Pred will have the same effect as going to Dest in
|
2008-12-19 18:03:11 +00:00
|
|
|
// terms of PHI values.
|
|
|
|
PHINode *PN;
|
|
|
|
unsigned PHINo = 0;
|
|
|
|
bool FoundMatch = true;
|
|
|
|
for (BasicBlock::iterator I = Dest->begin();
|
|
|
|
(PN = dyn_cast<PHINode>(I)); ++I, ++PHINo) {
|
|
|
|
if (PHINo == TIPHIValues.size())
|
|
|
|
TIPHIValues.push_back(PN->getIncomingValueForBlock(TIBB));
|
|
|
|
|
|
|
|
// If the PHI entry doesn't work, we can't use this pred.
|
|
|
|
if (TIPHIValues[PHINo] != PN->getIncomingValueForBlock(Pred)) {
|
|
|
|
FoundMatch = false;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
// If we found a workable predecessor, change TI to branch to Succ.
|
|
|
|
if (FoundMatch) {
|
|
|
|
Dest->removePredecessor(TIBB);
|
|
|
|
TI->setSuccessor(SuccNum, Pred);
|
|
|
|
return;
|
2007-03-31 04:06:36 +00:00
|
|
|
}
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
SplitCriticalEdge(TI, SuccNum, P, true);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
PHINode *PN;
|
|
|
|
SmallVector<Value*, 8> TIPHIValues;
|
|
|
|
for (BasicBlock::iterator I = Dest->begin();
|
|
|
|
(PN = dyn_cast<PHINode>(I)); ++I)
|
|
|
|
TIPHIValues.push_back(PN->getIncomingValueForBlock(TIBB));
|
|
|
|
|
|
|
|
SmallVector<BasicBlock*, 8> IdenticalPreds;
|
|
|
|
for (pred_iterator PI = pred_begin(Dest), E = pred_end(Dest); PI != E; ++PI) {
|
|
|
|
BasicBlock *Pred = *PI;
|
|
|
|
if (BackEdges.count(std::make_pair(Pred, Dest)))
|
|
|
|
continue;
|
|
|
|
if (PI == TIBB)
|
|
|
|
IdenticalPreds.push_back(Pred);
|
|
|
|
else {
|
|
|
|
bool Identical = true;
|
|
|
|
unsigned PHINo = 0;
|
|
|
|
for (BasicBlock::iterator I = Dest->begin();
|
|
|
|
(PN = dyn_cast<PHINode>(I)); ++I, ++PHINo)
|
|
|
|
if (TIPHIValues[PHINo] != PN->getIncomingValueForBlock(Pred)) {
|
|
|
|
Identical = false;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (Identical)
|
|
|
|
IdenticalPreds.push_back(Pred);
|
2007-03-31 04:06:36 +00:00
|
|
|
}
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
assert(!IdenticalPreds.empty());
|
|
|
|
SplitBlockPredecessors(Dest, &IdenticalPreds[0], IdenticalPreds.size(),
|
|
|
|
".critedge", P);
|
2007-03-31 04:06:36 +00:00
|
|
|
}
|
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
/// OptimizeNoopCopyExpression - If the specified cast instruction is a noop
|
2009-06-14 23:30:43 +00:00
|
|
|
/// copy (e.g. it's casting from one pointer type to another, i32->i8 on PPC),
|
|
|
|
/// sink it into user blocks to reduce the number of virtual
|
2007-06-12 16:50:17 +00:00
|
|
|
/// registers that must be created and coalesced.
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
///
|
|
|
|
/// Return true if any changes are made.
|
2008-11-24 22:44:16 +00:00
|
|
|
///
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
static bool OptimizeNoopCopyExpression(CastInst *CI, const TargetLowering &TLI){
|
2008-09-24 05:32:41 +00:00
|
|
|
// If this is a noop copy,
|
2009-08-10 22:56:29 +00:00
|
|
|
EVT SrcVT = TLI.getValueType(CI->getOperand(0)->getType());
|
|
|
|
EVT DstVT = TLI.getValueType(CI->getType());
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// This is an fp<->int conversion?
|
2008-06-06 12:08:01 +00:00
|
|
|
if (SrcVT.isInteger() != DstVT.isInteger())
|
2007-03-31 04:06:36 +00:00
|
|
|
return false;
|
2008-06-08 20:54:56 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// If this is an extension, it will be a zero or sign extension, which
|
|
|
|
// isn't a noop.
|
2008-06-08 20:54:56 +00:00
|
|
|
if (SrcVT.bitsLT(DstVT)) return false;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// If these values will be promoted, find out what they will be promoted
|
|
|
|
// to. This helps us consider truncates on PPC as noop copies when they
|
|
|
|
// are.
|
2009-08-12 00:36:31 +00:00
|
|
|
if (TLI.getTypeAction(CI->getContext(), SrcVT) == TargetLowering::Promote)
|
|
|
|
SrcVT = TLI.getTypeToTransformTo(CI->getContext(), SrcVT);
|
|
|
|
if (TLI.getTypeAction(CI->getContext(), DstVT) == TargetLowering::Promote)
|
|
|
|
DstVT = TLI.getTypeToTransformTo(CI->getContext(), DstVT);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// If, after promotion, these are the same types, this is a noop copy.
|
|
|
|
if (SrcVT != DstVT)
|
2007-03-31 04:06:36 +00:00
|
|
|
return false;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
BasicBlock *DefBB = CI->getParent();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
/// InsertedCasts - Only insert a cast in each block once.
|
2007-06-12 16:50:17 +00:00
|
|
|
DenseMap<BasicBlock*, CastInst*> InsertedCasts;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
bool MadeChange = false;
|
2008-09-24 05:32:41 +00:00
|
|
|
for (Value::use_iterator UI = CI->use_begin(), E = CI->use_end();
|
2007-03-31 04:06:36 +00:00
|
|
|
UI != E; ) {
|
|
|
|
Use &TheUse = UI.getUse();
|
|
|
|
Instruction *User = cast<Instruction>(*UI);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
// Figure out which BB this cast is used in. For PHI's this is the
|
|
|
|
// appropriate predecessor block.
|
|
|
|
BasicBlock *UserBB = User->getParent();
|
|
|
|
if (PHINode *PN = dyn_cast<PHINode>(User)) {
|
2009-01-23 19:40:15 +00:00
|
|
|
UserBB = PN->getIncomingBlock(UI);
|
2007-03-31 04:06:36 +00:00
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
// Preincrement use iterator so we don't invalidate it.
|
|
|
|
++UI;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
// If this user is in the same block as the cast, don't change the cast.
|
|
|
|
if (UserBB == DefBB) continue;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
// If we have already inserted a cast into this block, use it.
|
|
|
|
CastInst *&InsertedCast = InsertedCasts[UserBB];
|
|
|
|
|
|
|
|
if (!InsertedCast) {
|
2008-05-23 21:05:58 +00:00
|
|
|
BasicBlock::iterator InsertPt = UserBB->getFirstNonPHI();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
|
|
|
InsertedCast =
|
|
|
|
CastInst::Create(CI->getOpcode(), CI->getOperand(0), CI->getType(), "",
|
2007-03-31 04:06:36 +00:00
|
|
|
InsertPt);
|
|
|
|
MadeChange = true;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
// Replace a use of the cast with a use of the new cast.
|
2007-03-31 04:06:36 +00:00
|
|
|
TheUse = InsertedCast;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
// If we removed all uses, nuke the cast.
|
2008-01-20 16:51:46 +00:00
|
|
|
if (CI->use_empty()) {
|
2007-03-31 04:06:36 +00:00
|
|
|
CI->eraseFromParent();
|
2008-01-20 16:51:46 +00:00
|
|
|
MadeChange = true;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
return MadeChange;
|
|
|
|
}
|
|
|
|
|
2008-09-24 05:32:41 +00:00
|
|
|
/// OptimizeCmpExpression - sink the given CmpInst into user blocks to reduce
|
2007-06-12 16:50:17 +00:00
|
|
|
/// the number of virtual registers that must be created and coalesced. This is
|
2007-08-02 16:53:43 +00:00
|
|
|
/// a clear win except on targets with multiple condition code registers
|
|
|
|
/// (PowerPC), where it might lose; some adjustment may be wanted there.
|
2007-06-12 16:50:17 +00:00
|
|
|
///
|
|
|
|
/// Return true if any changes are made.
|
2008-11-24 22:44:16 +00:00
|
|
|
static bool OptimizeCmpExpression(CmpInst *CI) {
|
2007-06-12 16:50:17 +00:00
|
|
|
BasicBlock *DefBB = CI->getParent();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
/// InsertedCmp - Only insert a cmp in each block once.
|
|
|
|
DenseMap<BasicBlock*, CmpInst*> InsertedCmps;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
bool MadeChange = false;
|
2008-09-24 05:32:41 +00:00
|
|
|
for (Value::use_iterator UI = CI->use_begin(), E = CI->use_end();
|
2007-06-12 16:50:17 +00:00
|
|
|
UI != E; ) {
|
|
|
|
Use &TheUse = UI.getUse();
|
|
|
|
Instruction *User = cast<Instruction>(*UI);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
// Preincrement use iterator so we don't invalidate it.
|
|
|
|
++UI;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
// Don't bother for PHI nodes.
|
|
|
|
if (isa<PHINode>(User))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
// Figure out which BB this cmp is used in.
|
|
|
|
BasicBlock *UserBB = User->getParent();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
// If this user is in the same block as the cmp, don't change the cmp.
|
|
|
|
if (UserBB == DefBB) continue;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
// If we have already inserted a cmp into this block, use it.
|
|
|
|
CmpInst *&InsertedCmp = InsertedCmps[UserBB];
|
|
|
|
|
|
|
|
if (!InsertedCmp) {
|
2008-05-23 21:05:58 +00:00
|
|
|
BasicBlock::iterator InsertPt = UserBB->getFirstNonPHI();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
|
|
|
InsertedCmp =
|
2009-08-25 23:17:54 +00:00
|
|
|
CmpInst::Create(CI->getOpcode(),
|
2009-07-09 23:48:35 +00:00
|
|
|
CI->getPredicate(), CI->getOperand(0),
|
2007-06-12 16:50:17 +00:00
|
|
|
CI->getOperand(1), "", InsertPt);
|
|
|
|
MadeChange = true;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
// Replace a use of the cmp with a use of the new cmp.
|
|
|
|
TheUse = InsertedCmp;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
// If we removed all uses, nuke the cmp.
|
|
|
|
if (CI->use_empty())
|
|
|
|
CI->eraseFromParent();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-06-12 16:50:17 +00:00
|
|
|
return MadeChange;
|
|
|
|
}
|
|
|
|
|
2008-11-25 07:09:13 +00:00
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
// Memory Optimization
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
/// IsNonLocalValue - Return true if the specified values are defined in a
|
|
|
|
/// different basic block than BB.
|
|
|
|
static bool IsNonLocalValue(Value *V, BasicBlock *BB) {
|
|
|
|
if (Instruction *I = dyn_cast<Instruction>(V))
|
|
|
|
return I->getParent() != BB;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2008-11-25 07:09:13 +00:00
|
|
|
/// OptimizeMemoryInst - Load and Store Instructions have often have
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
/// addressing modes that can do significant amounts of computation. As such,
|
|
|
|
/// instruction selection will try to get the load or store to do as much
|
|
|
|
/// computation as possible for the program. The problem is that isel can only
|
|
|
|
/// see within a single block. As such, we sink as much legal addressing mode
|
|
|
|
/// stuff into the block as possible.
|
2008-11-25 07:09:13 +00:00
|
|
|
///
|
|
|
|
/// This method is used to optimize both load/store and inline asms with memory
|
|
|
|
/// operands.
|
2008-11-26 03:20:37 +00:00
|
|
|
bool CodeGenPrepare::OptimizeMemoryInst(Instruction *MemoryInst, Value *Addr,
|
2008-11-25 07:09:13 +00:00
|
|
|
const Type *AccessTy,
|
|
|
|
DenseMap<Value*,Value*> &SunkAddrs) {
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// Figure out what addressing mode will be built up for this operation.
|
|
|
|
SmallVector<Instruction*, 16> AddrModeInsts;
|
2008-11-26 03:20:37 +00:00
|
|
|
ExtAddrMode AddrMode = AddressingModeMatcher::Match(Addr, AccessTy,MemoryInst,
|
|
|
|
AddrModeInsts, *TLI);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// Check to see if any of the instructions supersumed by this addr mode are
|
|
|
|
// non-local to I's BB.
|
|
|
|
bool AnyNonLocal = false;
|
|
|
|
for (unsigned i = 0, e = AddrModeInsts.size(); i != e; ++i) {
|
2008-11-26 03:20:37 +00:00
|
|
|
if (IsNonLocalValue(AddrModeInsts[i], MemoryInst->getParent())) {
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
AnyNonLocal = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// If all the instructions matched are already in this BB, don't do anything.
|
|
|
|
if (!AnyNonLocal) {
|
2009-07-25 01:13:51 +00:00
|
|
|
DEBUG(errs() << "CGP: Found local addrmode: " << AddrMode << "\n");
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
return false;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// Insert this computation right after this user. Since our caller is
|
|
|
|
// scanning from the top of the BB to the bottom, reuse of the expr are
|
|
|
|
// guaranteed to happen later.
|
2008-11-26 03:20:37 +00:00
|
|
|
BasicBlock::iterator InsertPt = MemoryInst;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// Now that we determined the addressing expression we want to use and know
|
|
|
|
// that we have to sink it into this block. Check to see if we have already
|
|
|
|
// done this for some other load/store instr in this block. If so, reuse the
|
|
|
|
// computation.
|
|
|
|
Value *&SunkAddr = SunkAddrs[Addr];
|
|
|
|
if (SunkAddr) {
|
2009-07-25 01:13:51 +00:00
|
|
|
DEBUG(errs() << "CGP: Reusing nonlocal addrmode: " << AddrMode << " for "
|
|
|
|
<< *MemoryInst);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
if (SunkAddr->getType() != Addr->getType())
|
|
|
|
SunkAddr = new BitCastInst(SunkAddr, Addr->getType(), "tmp", InsertPt);
|
|
|
|
} else {
|
2009-07-25 01:13:51 +00:00
|
|
|
DEBUG(errs() << "CGP: SINKING nonlocal addrmode: " << AddrMode << " for "
|
|
|
|
<< *MemoryInst);
|
2009-08-13 21:58:54 +00:00
|
|
|
const Type *IntPtrTy =
|
|
|
|
TLI->getTargetData()->getIntPtrType(AccessTy->getContext());
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
Value *Result = 0;
|
|
|
|
// Start with the scale value.
|
|
|
|
if (AddrMode.Scale) {
|
|
|
|
Value *V = AddrMode.ScaledReg;
|
|
|
|
if (V->getType() == IntPtrTy) {
|
|
|
|
// done.
|
|
|
|
} else if (isa<PointerType>(V->getType())) {
|
|
|
|
V = new PtrToIntInst(V, IntPtrTy, "sunkaddr", InsertPt);
|
|
|
|
} else if (cast<IntegerType>(IntPtrTy)->getBitWidth() <
|
|
|
|
cast<IntegerType>(V->getType())->getBitWidth()) {
|
|
|
|
V = new TruncInst(V, IntPtrTy, "sunkaddr", InsertPt);
|
|
|
|
} else {
|
|
|
|
V = new SExtInst(V, IntPtrTy, "sunkaddr", InsertPt);
|
|
|
|
}
|
|
|
|
if (AddrMode.Scale != 1)
|
2009-07-24 23:12:02 +00:00
|
|
|
V = BinaryOperator::CreateMul(V, ConstantInt::get(IntPtrTy,
|
2009-07-03 00:17:18 +00:00
|
|
|
AddrMode.Scale),
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
"sunkaddr", InsertPt);
|
|
|
|
Result = V;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Add in the base register.
|
|
|
|
if (AddrMode.BaseReg) {
|
|
|
|
Value *V = AddrMode.BaseReg;
|
2009-06-02 21:29:13 +00:00
|
|
|
if (isa<PointerType>(V->getType()))
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
V = new PtrToIntInst(V, IntPtrTy, "sunkaddr", InsertPt);
|
2009-06-02 21:29:13 +00:00
|
|
|
if (V->getType() != IntPtrTy)
|
|
|
|
V = CastInst::CreateIntegerCast(V, IntPtrTy, /*isSigned=*/true,
|
|
|
|
"sunkaddr", InsertPt);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
if (Result)
|
2008-05-16 19:29:10 +00:00
|
|
|
Result = BinaryOperator::CreateAdd(Result, V, "sunkaddr", InsertPt);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
else
|
|
|
|
Result = V;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// Add in the BaseGV if present.
|
|
|
|
if (AddrMode.BaseGV) {
|
|
|
|
Value *V = new PtrToIntInst(AddrMode.BaseGV, IntPtrTy, "sunkaddr",
|
|
|
|
InsertPt);
|
|
|
|
if (Result)
|
2008-05-16 19:29:10 +00:00
|
|
|
Result = BinaryOperator::CreateAdd(Result, V, "sunkaddr", InsertPt);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
else
|
|
|
|
Result = V;
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// Add in the Base Offset if present.
|
|
|
|
if (AddrMode.BaseOffs) {
|
2009-07-24 23:12:02 +00:00
|
|
|
Value *V = ConstantInt::get(IntPtrTy, AddrMode.BaseOffs);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
if (Result)
|
2008-05-16 19:29:10 +00:00
|
|
|
Result = BinaryOperator::CreateAdd(Result, V, "sunkaddr", InsertPt);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
else
|
|
|
|
Result = V;
|
|
|
|
}
|
2007-03-31 04:06:36 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
if (Result == 0)
|
2009-07-31 20:28:14 +00:00
|
|
|
SunkAddr = Constant::getNullValue(Addr->getType());
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
else
|
|
|
|
SunkAddr = new IntToPtrInst(Result, Addr->getType(), "sunkaddr",InsertPt);
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2008-11-26 03:20:37 +00:00
|
|
|
MemoryInst->replaceUsesOfWith(Addr, SunkAddr);
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
if (Addr->use_empty())
|
2008-11-27 22:57:53 +00:00
|
|
|
RecursivelyDeleteTriviallyDeadInstructions(Addr);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
return true;
|
|
|
|
}
|
2007-03-31 04:06:36 +00:00
|
|
|
|
2008-02-26 02:42:37 +00:00
|
|
|
/// OptimizeInlineAsmInst - If there are any memory operands, use
|
2008-11-25 07:09:13 +00:00
|
|
|
/// OptimizeMemoryInst to sink their address computing into the block when
|
2008-02-26 02:42:37 +00:00
|
|
|
/// possible / profitable.
|
|
|
|
bool CodeGenPrepare::OptimizeInlineAsmInst(Instruction *I, CallSite CS,
|
|
|
|
DenseMap<Value*,Value*> &SunkAddrs) {
|
|
|
|
bool MadeChange = false;
|
|
|
|
InlineAsm *IA = cast<InlineAsm>(CS.getCalledValue());
|
|
|
|
|
|
|
|
// Do a prepass over the constraints, canonicalizing them, and building up the
|
|
|
|
// ConstraintOperands list.
|
|
|
|
std::vector<InlineAsm::ConstraintInfo>
|
|
|
|
ConstraintInfos = IA->ParseConstraints();
|
|
|
|
|
|
|
|
/// ConstraintOperands - Information about all of the constraints.
|
|
|
|
std::vector<TargetLowering::AsmOperandInfo> ConstraintOperands;
|
|
|
|
unsigned ArgNo = 0; // ArgNo - The argument of the CallInst.
|
|
|
|
for (unsigned i = 0, e = ConstraintInfos.size(); i != e; ++i) {
|
|
|
|
ConstraintOperands.
|
|
|
|
push_back(TargetLowering::AsmOperandInfo(ConstraintInfos[i]));
|
|
|
|
TargetLowering::AsmOperandInfo &OpInfo = ConstraintOperands.back();
|
|
|
|
|
|
|
|
// Compute the value type for each operand.
|
|
|
|
switch (OpInfo.Type) {
|
|
|
|
case InlineAsm::isOutput:
|
|
|
|
if (OpInfo.isIndirect)
|
|
|
|
OpInfo.CallOperandVal = CS.getArgument(ArgNo++);
|
|
|
|
break;
|
|
|
|
case InlineAsm::isInput:
|
|
|
|
OpInfo.CallOperandVal = CS.getArgument(ArgNo++);
|
|
|
|
break;
|
|
|
|
case InlineAsm::isClobber:
|
|
|
|
// Nothing to do.
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Compute the constraint code and ConstraintType to use.
|
2008-09-24 06:48:55 +00:00
|
|
|
TLI->ComputeConstraintToUse(OpInfo, SDValue(),
|
|
|
|
OpInfo.ConstraintType == TargetLowering::C_Memory);
|
2008-02-26 02:42:37 +00:00
|
|
|
|
2008-02-26 18:37:49 +00:00
|
|
|
if (OpInfo.ConstraintType == TargetLowering::C_Memory &&
|
|
|
|
OpInfo.isIndirect) {
|
2008-02-26 02:42:37 +00:00
|
|
|
Value *OpVal = OpInfo.CallOperandVal;
|
2008-11-25 07:09:13 +00:00
|
|
|
MadeChange |= OptimizeMemoryInst(I, OpVal, OpVal->getType(), SunkAddrs);
|
2008-02-26 02:42:37 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return MadeChange;
|
|
|
|
}
|
|
|
|
|
2007-12-05 23:58:20 +00:00
|
|
|
bool CodeGenPrepare::OptimizeExtUses(Instruction *I) {
|
|
|
|
BasicBlock *DefBB = I->getParent();
|
|
|
|
|
|
|
|
// If both result of the {s|z}xt and its source are live out, rewrite all
|
|
|
|
// other uses of the source with result of extension.
|
|
|
|
Value *Src = I->getOperand(0);
|
|
|
|
if (Src->hasOneUse())
|
|
|
|
return false;
|
|
|
|
|
2007-12-13 07:50:36 +00:00
|
|
|
// Only do this xform if truncating is free.
|
2008-02-26 19:13:21 +00:00
|
|
|
if (TLI && !TLI->isTruncateFree(I->getType(), Src->getType()))
|
2007-12-13 03:32:53 +00:00
|
|
|
return false;
|
|
|
|
|
2007-12-12 00:51:06 +00:00
|
|
|
// Only safe to perform the optimization if the source is also defined in
|
2007-12-12 02:53:41 +00:00
|
|
|
// this block.
|
|
|
|
if (!isa<Instruction>(Src) || DefBB != cast<Instruction>(Src)->getParent())
|
2007-12-12 00:51:06 +00:00
|
|
|
return false;
|
|
|
|
|
2007-12-05 23:58:20 +00:00
|
|
|
bool DefIsLiveOut = false;
|
2008-09-24 05:32:41 +00:00
|
|
|
for (Value::use_iterator UI = I->use_begin(), E = I->use_end();
|
2007-12-05 23:58:20 +00:00
|
|
|
UI != E; ++UI) {
|
|
|
|
Instruction *User = cast<Instruction>(*UI);
|
|
|
|
|
|
|
|
// Figure out which BB this ext is used in.
|
|
|
|
BasicBlock *UserBB = User->getParent();
|
|
|
|
if (UserBB == DefBB) continue;
|
|
|
|
DefIsLiveOut = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (!DefIsLiveOut)
|
|
|
|
return false;
|
|
|
|
|
2007-12-12 02:53:41 +00:00
|
|
|
// Make sure non of the uses are PHI nodes.
|
2008-09-24 05:32:41 +00:00
|
|
|
for (Value::use_iterator UI = Src->use_begin(), E = Src->use_end();
|
2007-12-12 02:53:41 +00:00
|
|
|
UI != E; ++UI) {
|
|
|
|
Instruction *User = cast<Instruction>(*UI);
|
2007-12-13 03:32:53 +00:00
|
|
|
BasicBlock *UserBB = User->getParent();
|
|
|
|
if (UserBB == DefBB) continue;
|
|
|
|
// Be conservative. We don't want this xform to end up introducing
|
|
|
|
// reloads just before load / store instructions.
|
|
|
|
if (isa<PHINode>(User) || isa<LoadInst>(User) || isa<StoreInst>(User))
|
2007-12-12 02:53:41 +00:00
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2007-12-05 23:58:20 +00:00
|
|
|
// InsertedTruncs - Only insert one trunc in each block once.
|
|
|
|
DenseMap<BasicBlock*, Instruction*> InsertedTruncs;
|
|
|
|
|
|
|
|
bool MadeChange = false;
|
2008-09-24 05:32:41 +00:00
|
|
|
for (Value::use_iterator UI = Src->use_begin(), E = Src->use_end();
|
2007-12-05 23:58:20 +00:00
|
|
|
UI != E; ++UI) {
|
|
|
|
Use &TheUse = UI.getUse();
|
|
|
|
Instruction *User = cast<Instruction>(*UI);
|
|
|
|
|
|
|
|
// Figure out which BB this ext is used in.
|
|
|
|
BasicBlock *UserBB = User->getParent();
|
|
|
|
if (UserBB == DefBB) continue;
|
|
|
|
|
|
|
|
// Both src and def are live in this block. Rewrite the use.
|
|
|
|
Instruction *&InsertedTrunc = InsertedTruncs[UserBB];
|
|
|
|
|
|
|
|
if (!InsertedTrunc) {
|
2008-05-23 21:05:58 +00:00
|
|
|
BasicBlock::iterator InsertPt = UserBB->getFirstNonPHI();
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-12-05 23:58:20 +00:00
|
|
|
InsertedTrunc = new TruncInst(I, Src->getType(), "", InsertPt);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Replace a use of the {s|z}ext source with a use of the result.
|
|
|
|
TheUse = InsertedTrunc;
|
|
|
|
|
|
|
|
MadeChange = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
return MadeChange;
|
|
|
|
}
|
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
// In this pass we look for GEP and cast instructions that are used
|
|
|
|
// across basic blocks and rewrite them to improve basic-block-at-a-time
|
|
|
|
// selection.
|
|
|
|
bool CodeGenPrepare::OptimizeBlock(BasicBlock &BB) {
|
|
|
|
bool MadeChange = false;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2008-12-19 18:03:11 +00:00
|
|
|
// Split all critical edges where the dest block has a PHI.
|
2007-03-31 04:06:36 +00:00
|
|
|
TerminatorInst *BBTI = BB.getTerminator();
|
|
|
|
if (BBTI->getNumSuccessors() > 1) {
|
2008-12-19 18:03:11 +00:00
|
|
|
for (unsigned i = 0, e = BBTI->getNumSuccessors(); i != e; ++i) {
|
|
|
|
BasicBlock *SuccBB = BBTI->getSuccessor(i);
|
|
|
|
if (isa<PHINode>(SuccBB->begin()) && isCriticalEdge(BBTI, i, true))
|
|
|
|
SplitEdgeNicely(BBTI, i, BackEdges, this);
|
|
|
|
}
|
2007-03-31 04:06:36 +00:00
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
// Keep track of non-local addresses that have been sunk into this block.
|
|
|
|
// This allows us to avoid inserting duplicate code for blocks with multiple
|
|
|
|
// load/stores of the same address.
|
|
|
|
DenseMap<Value*, Value*> SunkAddrs;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
for (BasicBlock::iterator BBI = BB.begin(), E = BB.end(); BBI != E; ) {
|
|
|
|
Instruction *I = BBI++;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
if (CastInst *CI = dyn_cast<CastInst>(I)) {
|
2007-03-31 04:06:36 +00:00
|
|
|
// If the source of the cast is a constant, then this should have
|
|
|
|
// already been constant folded. The only reason NOT to constant fold
|
|
|
|
// it is if something (e.g. LSR) was careful to place the constant
|
|
|
|
// evaluation in a block other than then one that uses it (e.g. to hoist
|
|
|
|
// the address of globals out of a loop). If this is the case, we don't
|
|
|
|
// want to forward-subst the cast.
|
|
|
|
if (isa<Constant>(CI->getOperand(0)))
|
|
|
|
continue;
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-12-05 23:58:20 +00:00
|
|
|
bool Change = false;
|
|
|
|
if (TLI) {
|
|
|
|
Change = OptimizeNoopCopyExpression(CI, *TLI);
|
|
|
|
MadeChange |= Change;
|
|
|
|
}
|
|
|
|
|
2008-03-19 22:02:26 +00:00
|
|
|
if (!Change && (isa<ZExtInst>(I) || isa<SExtInst>(I)))
|
2007-12-05 23:58:20 +00:00
|
|
|
MadeChange |= OptimizeExtUses(I);
|
2007-06-12 16:50:17 +00:00
|
|
|
} else if (CmpInst *CI = dyn_cast<CmpInst>(I)) {
|
|
|
|
MadeChange |= OptimizeCmpExpression(CI);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
} else if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
|
|
|
|
if (TLI)
|
2008-11-25 07:09:13 +00:00
|
|
|
MadeChange |= OptimizeMemoryInst(I, I->getOperand(0), LI->getType(),
|
|
|
|
SunkAddrs);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
} else if (StoreInst *SI = dyn_cast<StoreInst>(I)) {
|
|
|
|
if (TLI)
|
2008-11-25 07:09:13 +00:00
|
|
|
MadeChange |= OptimizeMemoryInst(I, SI->getOperand(1),
|
|
|
|
SI->getOperand(0)->getType(),
|
|
|
|
SunkAddrs);
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
} else if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(I)) {
|
2007-04-14 00:17:39 +00:00
|
|
|
if (GEPI->hasAllZeroIndices()) {
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
/// The GEP operand must be a pointer, so must its result -> BitCast
|
2008-09-24 05:32:41 +00:00
|
|
|
Instruction *NC = new BitCastInst(GEPI->getOperand(0), GEPI->getType(),
|
Completely rewrite addressing-mode related sinking of code. In particular,
this fixes problems where codegenprepare would sink expressions into load/stores
that are not valid, and fixes cases where it would miss important valid ones.
This fixes several serious codesize and perf issues, particularly on targets
with complex addressing modes like arm and x86. For example, now we compile
CodeGen/X86/isel-sink.ll to:
_test:
movl 8(%esp), %eax
movl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx,%eax,4)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx,%eax,4), %eax
ret
instead of:
_test:
movl 8(%esp), %eax
leal (,%eax,4), %ecx
addl 4(%esp), %ecx
cmpl $1233, %eax
ja LBB1_2 #F
LBB1_1: #T
movl $4, (%ecx)
movl $141, %eax
ret
LBB1_2: #F
movl (%ecx), %eax
ret
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@35970 91177308-0d34-0410-b5e6-96231b3b80d8
2007-04-13 20:30:56 +00:00
|
|
|
GEPI->getName(), GEPI);
|
|
|
|
GEPI->replaceAllUsesWith(NC);
|
|
|
|
GEPI->eraseFromParent();
|
|
|
|
MadeChange = true;
|
|
|
|
BBI = NC;
|
|
|
|
}
|
|
|
|
} else if (CallInst *CI = dyn_cast<CallInst>(I)) {
|
|
|
|
// If we found an inline asm expession, and if the target knows how to
|
|
|
|
// lower it to normal LLVM code, do so now.
|
2009-07-20 17:52:52 +00:00
|
|
|
if (TLI && isa<InlineAsm>(CI->getCalledValue())) {
|
|
|
|
if (TLI->ExpandInlineAsm(CI)) {
|
|
|
|
BBI = BB.begin();
|
|
|
|
// Avoid processing instructions out of order, which could cause
|
|
|
|
// reuse before a value is defined.
|
|
|
|
SunkAddrs.clear();
|
|
|
|
} else
|
|
|
|
// Sink address computing for memory operands into the block.
|
|
|
|
MadeChange |= OptimizeInlineAsmInst(I, &(*CI), SunkAddrs);
|
|
|
|
}
|
2007-03-31 04:06:36 +00:00
|
|
|
}
|
|
|
|
}
|
2008-09-24 05:32:41 +00:00
|
|
|
|
2007-03-31 04:06:36 +00:00
|
|
|
return MadeChange;
|
|
|
|
}
|