6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-06-17 20:23:59 +00:00
Code Issues Projects Releases Wiki Activity

Fix a trip-count overflow issue in LoopUnroll.

Browse Source 
Currently LoopUnroll generates a prologue loop before the main loop
body to execute first N%UnrollFactor iterations. Also, this loop is
used if trip-count can overflow - it's determined by a runtime check.

However, we've been mistakenly optimizing this loop to a linear code for
UnrollFactor = 2, not taking into account that it also serves as a safe
version of the loop if its trip-count overflows.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@222451 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Michael Zolotukhin
2014-11-20 20:19:55 +00:00
parent d23f04a165
commit 4e7b10b07f
3 changed files with 42 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
lib/Transforms/Utils/LoopUnrollRuntime.cpp
View File

@ -295,6 +295,10 @@ bool llvm::UnrollRuntimeLoopProlog(Loop *L, unsigned Count, LoopInfo *LI,
if (isa<SCEVCouldNotCompute>(BECount) || !BECount->getType()->isIntegerTy())
return false;
// If BECount is INT_MAX, we can't compute trip-count without overflow.
if (BECount->isAllOnesValue())
return false;
// Add 1 since the backedge count doesn't include the first loop iteration
const SCEV *TripCountSC =
SE->getAddExpr(BECount, SE->getConstant(BECount->getType(), 1));
@ -357,11 +361,16 @@ bool llvm::UnrollRuntimeLoopProlog(Loop *L, unsigned Count, LoopInfo *LI,
std::vector<BasicBlock *> NewBlocks;
ValueToValueMapTy VMap;
// If unroll count is 2 and we can't overflow in tripcount computation (which
// is BECount + 1), then we don't need a loop for prologue, and we can unroll
// it. We can be sure that we don't overflow only if tripcount is a constant.
bool UnrollPrologue = (Count == 2 && isa<ConstantInt>(TripCount));
// Clone all the basic blocks in the loop. If Count is 2, we don't clone
// the loop, otherwise we create a cloned loop to execute the extra
// iterations. This function adds the appropriate CFG connections.
CloneLoopBlocks(L, ModVal, Count == 2, PH, PEnd, NewBlocks, LoopBlocks, VMap,
LI);
CloneLoopBlocks(L, ModVal, UnrollPrologue, PH, PEnd, NewBlocks, LoopBlocks,
VMap, LI);
// Insert the cloned blocks into function just before the original loop
F->getBasicBlockList().splice(PEnd, F->getBasicBlockList(), NewBlocks[0],