This patch implements quick look-up for block in loop by maintaining a hash set for blocks.
It improves the efficiency of loop analysis a lot, the biggest improvement could be 5-6%(458.sjeng).
Below are the compilation time for our benchmark in llc before & after the patch.
Benchmark llc - trunk llc - patched
401.bzip2 0.339081 100.00% 0.329657 102.86%
403.gcc 19.853966 100.00% 19.605466 101.27%
429.mcf 0.049823 100.00% 0.048451 102.83%
433.milc 0.514898 100.00% 0.510217 100.92%
444.namd 1.109328 100.00% 1.103481 100.53%
445.gobmk 4.988028 100.00% 4.929114 101.20%
456.hmmer 0.843871 100.00% 0.825865 102.18%
458.sjeng 0.754238 100.00% 0.714095 105.62%
464.h264ref 2.9668 100.00% 2.90612 102.09%
471.omnetpp 4.556533 100.00% 4.511886 100.99%
bitmnp01 0.038168 100.00% 0.0357 106.91%
idctrn01 0.037745 100.00% 0.037332 101.11%
libquake2 3.78689 100.00% 3.76209 100.66%
libquake_ 2.251525 100.00% 2.234104 100.78%
linpack 0.033159 100.00% 0.032788 101.13%
matrix01 0.045319 100.00% 0.043497 104.19%
nbench 0.333161 100.00% 0.329799 101.02%
tblook01 0.017863 100.00% 0.017666 101.12%
ttsprk01 0.054337 100.00% 0.053057 102.41%
Reviewer : Andrew Trick <atrick@apple.com>, Hal Finkel <hfinkel@anl.gov>
Approver : Andrew Trick <atrick@apple.com>
Test : Pass make check-all & llvm test-suite
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193460
91177308-0d34-0410-b5e6-
96231b3b80d8
// Blocks - The list of blocks in this loop. First entry is the header node.
std::vector<BlockT*> Blocks;
+ SmallPtrSet<const BlockT*, 8> DenseBlockSet;
+
LoopBase(const LoopBase<BlockT, LoopT> &) LLVM_DELETED_FUNCTION;
const LoopBase<BlockT, LoopT>&
operator=(const LoopBase<BlockT, LoopT> &) LLVM_DELETED_FUNCTION;
/// contains - Return true if the specified basic block is in this loop.
///
bool contains(const BlockT *BB) const {
- return std::find(block_begin(), block_end(), BB) != block_end();
+ return DenseBlockSet.count(BB);
}
/// contains - Return true if the specified instruction is in this loop.
/// getBlocks - Get a list of the basic blocks which make up this loop.
///
const std::vector<BlockT*> &getBlocks() const { return Blocks; }
- std::vector<BlockT*> &getBlocksVector() { return Blocks; }
typedef typename std::vector<BlockT*>::const_iterator block_iterator;
block_iterator block_begin() const { return Blocks.begin(); }
block_iterator block_end() const { return Blocks.end(); }
/// transformations should use addBasicBlockToLoop.
void addBlockEntry(BlockT *BB) {
Blocks.push_back(BB);
+ DenseBlockSet.insert(BB);
+ }
+
+ /// reverseBlocks - interface to reverse Blocks[from, end of loop] in this loop
+ void reverseBlock(unsigned from) {
+ std::reverse(Blocks.begin() + from, Blocks.end());
+ }
+
+ /// reserveBlocks- interface to do reserve() for Blocks
+ void reserveBlocks(unsigned size) {
+ Blocks.reserve(size);
}
/// moveToHeader - This method is used to move BB (which must be part of this
/// the mapping in the LoopInfo class.
void removeBlockFromLoop(BlockT *BB) {
RemoveFromVector(Blocks, BB);
+ DenseBlockSet.erase(BB);
}
/// verifyLoop - Verify loop structure
friend class LoopInfoBase<BlockT, LoopT>;
explicit LoopBase(BlockT *BB) : ParentLoop(0) {
Blocks.push_back(BB);
+ DenseBlockSet.insert(BB);
}
};
template<class BlockT, class LoopT>
void LoopBase<BlockT, LoopT>::
getExitingBlocks(SmallVectorImpl<BlockT *> &ExitingBlocks) const {
- // Sort the blocks vector so that we can use binary search to do quick
- // lookups.
- SmallVector<BlockT*, 128> LoopBBs(block_begin(), block_end());
- std::sort(LoopBBs.begin(), LoopBBs.end());
-
typedef GraphTraits<BlockT*> BlockTraits;
for (block_iterator BI = block_begin(), BE = block_end(); BI != BE; ++BI)
for (typename BlockTraits::ChildIteratorType I =
BlockTraits::child_begin(*BI), E = BlockTraits::child_end(*BI);
I != E; ++I)
- if (!std::binary_search(LoopBBs.begin(), LoopBBs.end(), *I)) {
+ if (!contains(*I)) {
// Not in current loop? It must be an exit block.
ExitingBlocks.push_back(*BI);
break;
template<class BlockT, class LoopT>
void LoopBase<BlockT, LoopT>::
getExitBlocks(SmallVectorImpl<BlockT*> &ExitBlocks) const {
- // Sort the blocks vector so that we can use binary search to do quick
- // lookups.
- SmallVector<BlockT*, 128> LoopBBs(block_begin(), block_end());
- std::sort(LoopBBs.begin(), LoopBBs.end());
-
typedef GraphTraits<BlockT*> BlockTraits;
for (block_iterator BI = block_begin(), BE = block_end(); BI != BE; ++BI)
for (typename BlockTraits::ChildIteratorType I =
BlockTraits::child_begin(*BI), E = BlockTraits::child_end(*BI);
I != E; ++I)
- if (!std::binary_search(LoopBBs.begin(), LoopBBs.end(), *I))
+ if (!contains(*I))
// Not in current loop? It must be an exit block.
ExitBlocks.push_back(*I);
}
template<class BlockT, class LoopT>
void LoopBase<BlockT, LoopT>::
getExitEdges(SmallVectorImpl<Edge> &ExitEdges) const {
- // Sort the blocks vector so that we can use binary search to do quick
- // lookups.
- SmallVector<BlockT*, 128> LoopBBs(block_begin(), block_end());
- array_pod_sort(LoopBBs.begin(), LoopBBs.end());
-
typedef GraphTraits<BlockT*> BlockTraits;
for (block_iterator BI = block_begin(), BE = block_end(); BI != BE; ++BI)
for (typename BlockTraits::ChildIteratorType I =
BlockTraits::child_begin(*BI), E = BlockTraits::child_end(*BI);
I != E; ++I)
- if (!std::binary_search(LoopBBs.begin(), LoopBBs.end(), *I))
+ if (!contains(*I))
// Not in current loop? It must be an exit block.
ExitEdges.push_back(Edge(*BI, *I));
}
// Add the basic block to this loop and all parent loops...
while (L) {
- L->Blocks.push_back(NewBB);
+ L->addBlockEntry(NewBB);
L = L->getParentLoop();
}
}
// Keep track of the number of BBs visited.
unsigned NumVisited = 0;
- // Sort the blocks vector so that we can use binary search to do quick
- // lookups.
- SmallVector<BlockT*, 128> LoopBBs(block_begin(), block_end());
- std::sort(LoopBBs.begin(), LoopBBs.end());
-
// Check the individual blocks.
for ( ; BI != BE; ++BI) {
BlockT *BB = *BI;
for (typename BlockTraits::ChildIteratorType SI =
BlockTraits::child_begin(BB), SE = BlockTraits::child_end(BB);
SI != SE; ++SI)
- if (std::binary_search(LoopBBs.begin(), LoopBBs.end(), *SI)) {
+ if (contains(*SI)) {
HasInsideLoopSuccs = true;
break;
}
InvBlockTraits::child_begin(BB), PE = InvBlockTraits::child_end(BB);
PI != PE; ++PI) {
BlockT *N = *PI;
- if (std::binary_search(LoopBBs.begin(), LoopBBs.end(), N))
+ if (contains(N))
HasInsideLoopPreds = true;
else
OutsideLoopPreds.push_back(N);
// Each block in each subloop should be contained within this loop.
for (block_iterator BI = (*I)->block_begin(), BE = (*I)->block_end();
BI != BE; ++BI) {
- assert(std::binary_search(LoopBBs.begin(), LoopBBs.end(), *BI) &&
+ assert(contains(*BI) &&
"Loop does not contain all the blocks of a subloop!");
}
}
}
L->getSubLoopsVector().reserve(NumSubloops);
- L->getBlocksVector().reserve(NumBlocks);
+ L->reserveBlocks(NumBlocks);
}
namespace {
// For convenience, Blocks and Subloops are inserted in postorder. Reverse
// the lists, except for the loop header, which is always at the beginning.
- std::reverse(Subloop->getBlocksVector().begin()+1,
- Subloop->getBlocksVector().end());
+ Subloop->reverseBlock(1);
std::reverse(Subloop->getSubLoopsVector().begin(),
Subloop->getSubLoopsVector().end());
Subloop = Subloop->getParentLoop();
}
for (; Subloop; Subloop = Subloop->getParentLoop())
- Subloop->getBlocksVector().push_back(Block);
+ Subloop->addBlockEntry(Block);
}
/// Analyze LoopInfo discovers loops during a postorder DominatorTree traversal
/// isLCSSAForm - Return true if the Loop is in LCSSA form
bool Loop::isLCSSAForm(DominatorTree &DT) const {
- // Sort the blocks vector so that we can use binary search to do quick
- // lookups.
- SmallPtrSet<BasicBlock*, 16> LoopBBs(block_begin(), block_end());
-
for (block_iterator BI = block_begin(), E = block_end(); BI != E; ++BI) {
BasicBlock *BB = *BI;
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E;++I)
// block they are defined in. Also, blocks not reachable from the
// entry are special; uses in them don't need to go through PHIs.
if (UserBB != BB &&
- !LoopBBs.count(UserBB) &&
+ !contains(UserBB) &&
DT.isReachableFromEntry(UserBB))
return false;
}
/// hasDedicatedExits - Return true if no exit block for the loop
/// has a predecessor that is outside the loop.
bool Loop::hasDedicatedExits() const {
- // Sort the blocks vector so that we can use binary search to do quick
- // lookups.
- SmallPtrSet<BasicBlock *, 16> LoopBBs(block_begin(), block_end());
// Each predecessor of each exit block of a normal loop is contained
// within the loop.
SmallVector<BasicBlock *, 4> ExitBlocks;
for (unsigned i = 0, e = ExitBlocks.size(); i != e; ++i)
for (pred_iterator PI = pred_begin(ExitBlocks[i]),
PE = pred_end(ExitBlocks[i]); PI != PE; ++PI)
- if (!LoopBBs.count(*PI))
+ if (!contains(*PI))
return false;
// All the requirements are met.
return true;
assert(hasDedicatedExits() &&
"getUniqueExitBlocks assumes the loop has canonical form exits!");
- // Sort the blocks vector so that we can use binary search to do quick
- // lookups.
- SmallVector<BasicBlock *, 128> LoopBBs(block_begin(), block_end());
- std::sort(LoopBBs.begin(), LoopBBs.end());
-
SmallVector<BasicBlock *, 32> switchExitBlocks;
for (block_iterator BI = block_begin(), BE = block_end(); BI != BE; ++BI) {
for (succ_iterator I = succ_begin(*BI), E = succ_end(*BI); I != E; ++I) {
// If block is inside the loop then it is not a exit block.
- if (std::binary_search(LoopBBs.begin(), LoopBBs.end(), *I))
+ if (contains(*I))
continue;
pred_iterator PI = pred_begin(*I);
DominatorTree *DT;
LoopInfo *LI;
ScalarEvolution *SE;
- std::vector<BasicBlock*> LoopBlocks;
PredIteratorCache PredCache;
Loop *L;
// Check the special guarantees that LCSSA makes.
assert(L->isLCSSAForm(*DT) && "LCSSA form not preserved!");
}
-
- /// inLoop - returns true if the given block is within the current loop
- bool inLoop(BasicBlock *B) const {
- return std::binary_search(LoopBlocks.begin(), LoopBlocks.end(), B);
- }
};
}
if (ExitBlocks.empty())
return false;
- // Speed up queries by creating a sorted vector of blocks.
- LoopBlocks.clear();
- LoopBlocks.insert(LoopBlocks.end(), L->block_begin(), L->block_end());
- array_pod_sort(LoopBlocks.begin(), LoopBlocks.end());
-
// Look at all the instructions in the loop, checking to see if they have uses
// outside the loop. If so, rewrite those uses.
bool MadeChange = false;
if (PHINode *PN = dyn_cast<PHINode>(U))
UserBB = PN->getIncomingBlock(UI);
- if (InstBB != UserBB && !inLoop(UserBB))
+ if (InstBB != UserBB && !L->contains(UserBB))
UsesToRewrite.push_back(&UI.getUse());
}
// If the exit block has a predecessor not within the loop, arrange for
// the incoming value use corresponding to that predecessor to be
// rewritten in terms of a different LCSSA PHI.
- if (!inLoop(*PI))
+ if (!L->contains(*PI))
UsesToRewrite.push_back(
&PN->getOperandUse(
PN->getOperandNumForIncomingValue(PN->getNumIncomingValues()-1)));
return false;
assert(TheLoop->getNumBlocks() > 1 && "Single block loops are vectorizable");
- std::vector<BasicBlock*> &LoopBlocks = TheLoop->getBlocksVector();
// A list of pointers that we can safely read and write to.
SmallPtrSet<Value *, 8> SafePointes;
// Collect safe addresses.
- for (unsigned i = 0, e = LoopBlocks.size(); i < e; ++i) {
- BasicBlock *BB = LoopBlocks[i];
+ for (Loop::block_iterator BI = TheLoop->block_begin(),
+ BE = TheLoop->block_end(); BI != BE; ++BI) {
+ BasicBlock *BB = *BI;
if (blockNeedsPredication(BB))
continue;
}
// Collect the blocks that need predication.
- for (unsigned i = 0, e = LoopBlocks.size(); i < e; ++i) {
- BasicBlock *BB = LoopBlocks[i];
+ for (Loop::block_iterator BI = TheLoop->block_begin(),
+ BE = TheLoop->block_end(); BI != BE; ++BI) {
+ BasicBlock *BB = *BI;
// We don't support switch statements inside loops.
if (!isa<BranchInst>(BB->getTerminator()))