[PATCH] D12289: [LV] When emitting an overflow check, test the trip count, not the backedge count.

[James Molloy via llvm-commits]

jmolloy created this revision.
jmolloy added reviewers: anemet, mzolotukhin.
jmolloy added a subscriber: llvm-commits.
jmolloy set the repository for this revision to rL LLVM.

We were testing the backedge-taken count with -1, when it's just as simple to test the trip count with zero. Given we need to calculate the trip count anyway, this removes code in the loop vectorizer.

Repository:
  rL LLVM

http://reviews.llvm.org/D12289

Files:
  lib/Transforms/Vectorize/LoopVectorize.cpp
  test/Transforms/LoopVectorize/induction.ll

Index: test/Transforms/LoopVectorize/induction.ll
===================================================================
--- test/Transforms/LoopVectorize/induction.ll
+++ test/Transforms/LoopVectorize/induction.ll
@@ -112,8 +112,8 @@
 ; condition and branch directly to the scalar loop.
 
 ; CHECK-LABEL: max_i32_backedgetaken
-; CHECK:  %backedge.overflow = icmp eq i32 -1, -1
-; CHECK:  br i1 %backedge.overflow, label %scalar.ph, label %overflow.checked
+; CHECK:  tc.overflow = icmp eq i32 0, 0
+; CHECK:  br i1 %tc.overflow, label %scalar.ph, label %overflow.checked
 
 ; CHECK: scalar.ph:
 ; CHECK:  %bc.resume.val = phi i32 [ %resume.val, %middle.block ], [ 0, %0 ]
Index: lib/Transforms/Vectorize/LoopVectorize.cpp
===================================================================
--- lib/Transforms/Vectorize/LoopVectorize.cpp
+++ lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -2650,29 +2650,22 @@
   // Notice that the pre-header does not change, only the loop body.
   SCEVExpander Exp(*SE, DL, "induction");
 
+  // Count holds the overall loop count (N).
+  Value *Count = Exp.expandCodeFor(ExitCount, ExitCount->getType(),
+                                   VectorPH->getTerminator());
+
   // We need to test whether the backedge-taken count is uint##_max. Adding one
   // to it will cause overflow and an incorrect loop trip count in the vector
   // body. In case of overflow we want to directly jump to the scalar remainder
   // loop.
-  Value *BackedgeCount =
-      Exp.expandCodeFor(BackedgeTakeCount, BackedgeTakeCount->getType(),
-                        VectorPH->getTerminator());
-  if (BackedgeCount->getType()->isPointerTy())
-    BackedgeCount = CastInst::CreatePointerCast(BackedgeCount, IdxTy,
-                                                "backedge.ptrcnt.to.int",
-                                                VectorPH->getTerminator());
   Instruction *CheckBCOverflow =
-      CmpInst::Create(Instruction::ICmp, CmpInst::ICMP_EQ, BackedgeCount,
-                      Constant::getAllOnesValue(BackedgeCount->getType()),
-                      "backedge.overflow", VectorPH->getTerminator());
+      CmpInst::Create(Instruction::ICmp, CmpInst::ICMP_EQ, Count,
+                      Constant::getNullValue(Count->getType()),
+                      "tc.overflow", VectorPH->getTerminator());
 
   Builder.SetInsertPoint(VectorPH->getTerminator());
   Value *StartIdx = ExtendedIdx = ConstantInt::get(IdxTy, 0);
 
-  // Count holds the overall loop count (N).
-  Value *Count = Exp.expandCodeFor(ExitCount, ExitCount->getType(),
-                                   VectorPH->getTerminator());
-
   LoopBypassBlocks.push_back(VectorPH);
 
   // Split the single block loop into the two loop structure described above.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D12289.32968.patch
Type: text/x-patch
Size: 2759 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150824/966a52ec/attachment.bin>