From caee94bbb4fb44971f594fe09fd61692dc4aa719 Mon Sep 17 00:00:00 2001 From: Sanjoy Das Date: Mon, 2 Mar 2015 21:41:07 +0000 Subject: [PATCH] Revert some changes that were made to fix PR20680. This re-lands change r230921. r230921 was reverted because it broke a clang test; a checkin fixing the clang test will be commited shortly. Summary: As far as I can tell, the real bug causing the issue was fixed in r230533. SCEVExpander should mark an increment operation as nuw or nsw only if it can *prove* that the operation does not overflow. There shouldn't be any situation where we have to do something different because of no-wrap flags generated by SCEVExpander. Revert "IndVarSimplify: Allow LFTR to fire more often" This reverts commit 1ade0f0faa98877b688e0b9da58e876052c1e04e (SVN: 222213). Revert "IndVarSimplify: Don't let LFTR compare against a poison value" This reverts commit c0f2b8b528d8a37b0a1522aae90af649d6357eb5 (SVN: 217102). Reviewers: majnemer, atrick, spatel Differential Revision: http://reviews.llvm.org/D7979 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@231018 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Transforms/Scalar/IndVarSimplify.cpp | 48 +++---------------- .../IndVarSimplify/2011-10-27-lftrnull.ll | 2 +- .../lftr-address-space-pointers.ll | 4 +- test/Transforms/IndVarSimplify/lftr-reuse.ll | 24 ++++------ 4 files changed, 17 insertions(+), 61 deletions(-) diff --git a/lib/Transforms/Scalar/IndVarSimplify.cpp b/lib/Transforms/Scalar/IndVarSimplify.cpp index f99ebbc453f..462c2b6f90d 100644 --- a/lib/Transforms/Scalar/IndVarSimplify.cpp +++ b/lib/Transforms/Scalar/IndVarSimplify.cpp @@ -1705,51 +1705,15 @@ LinearFunctionTestReplace(Loop *L, // compare against the post-incremented value, otherwise we must compare // against the preincremented value. if (L->getExitingBlock() == L->getLoopLatch()) { + // Add one to the "backedge-taken" count to get the trip count. + // This addition may overflow, which is valid as long as the comparison is + // truncated to BackedgeTakenCount->getType(). + IVCount = SE->getAddExpr(BackedgeTakenCount, + SE->getConstant(BackedgeTakenCount->getType(), 1)); // The BackedgeTaken expression contains the number of times that the // backedge branches to the loop header. This is one less than the // number of times the loop executes, so use the incremented indvar. - llvm::Value *IncrementedIndvar = - IndVar->getIncomingValueForBlock(L->getExitingBlock()); - const auto *IncrementedIndvarSCEV = - cast(SE->getSCEV(IncrementedIndvar)); - // It is unsafe to use the incremented indvar if it has a wrapping flag, we - // don't want to compare against a poison value. Check the SCEV that - // corresponds to the incremented indvar, the SCEVExpander will only insert - // flags in the IR if the SCEV originally had wrapping flags. - // FIXME: In theory, SCEV could drop flags even though they exist in IR. - // A more robust solution would involve getting a new expression for - // CmpIndVar by applying non-NSW/NUW AddExprs. - auto WrappingFlags = - ScalarEvolution::setFlags(SCEV::FlagNUW, SCEV::FlagNSW); - const SCEV *IVInit = IncrementedIndvarSCEV->getStart(); - if (SE->getTypeSizeInBits(IVInit->getType()) > - SE->getTypeSizeInBits(IVCount->getType())) - IVInit = SE->getTruncateExpr(IVInit, IVCount->getType()); - unsigned BitWidth = SE->getTypeSizeInBits(IVCount->getType()); - Type *WideTy = IntegerType::get(SE->getContext(), BitWidth + 1); - // Check if InitIV + BECount+1 requires sign/zero extension. - // If not, clear the corresponding flag from WrappingFlags because it is not - // necessary for those flags in the IncrementedIndvarSCEV expression. - if (SE->getSignExtendExpr(SE->getAddExpr(IVInit, BackedgeTakenCount), - WideTy) == - SE->getAddExpr(SE->getSignExtendExpr(IVInit, WideTy), - SE->getSignExtendExpr(BackedgeTakenCount, WideTy))) - WrappingFlags = ScalarEvolution::clearFlags(WrappingFlags, SCEV::FlagNSW); - if (SE->getZeroExtendExpr(SE->getAddExpr(IVInit, BackedgeTakenCount), - WideTy) == - SE->getAddExpr(SE->getZeroExtendExpr(IVInit, WideTy), - SE->getZeroExtendExpr(BackedgeTakenCount, WideTy))) - WrappingFlags = ScalarEvolution::clearFlags(WrappingFlags, SCEV::FlagNUW); - if (!ScalarEvolution::maskFlags(IncrementedIndvarSCEV->getNoWrapFlags(), - WrappingFlags)) { - // Add one to the "backedge-taken" count to get the trip count. - // This addition may overflow, which is valid as long as the comparison is - // truncated to BackedgeTakenCount->getType(). - IVCount = - SE->getAddExpr(BackedgeTakenCount, - SE->getConstant(BackedgeTakenCount->getType(), 1)); - CmpIndVar = IncrementedIndvar; - } + CmpIndVar = IndVar->getIncomingValueForBlock(L->getExitingBlock()); } Value *ExitCnt = genLoopLimit(IndVar, IVCount, L, Rewriter, SE); diff --git a/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll b/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll index 75eaf918bcf..f1d1139aef8 100644 --- a/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll +++ b/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll @@ -6,7 +6,7 @@ target triple = "thumbv7-apple-darwin" ; CHECK-LABEL: @test( ; CHECK: if.end.i126: -; CHECK: %exitcond = icmp ne i8* %destYPixelPtr.010.i, getelementptr (i8* null, i32 undef) +; CHECK: %exitcond = icmp ne i8* %incdec.ptr.i, getelementptr (i8* null, i32 undef) define void @test() nounwind { entry: br label %while.cond diff --git a/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll b/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll index b23e08dfad7..feb4b35a20a 100644 --- a/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll +++ b/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll @@ -11,7 +11,7 @@ entry: br i1 %cmp1, label %for.body, label %for.end ; Make sure the added GEP has the right index type -; CHECK: %lftr.limit = getelementptr i8, i8 addrspace(2)* %base, i8 +; CHECK: %lftr.limit = getelementptr i8, i8 addrspace(2)* %base, i8 %0 ; CHECK: for.body: ; CHECK: phi i8 addrspace(2)* @@ -43,7 +43,7 @@ entry: br i1 %cmp1, label %for.body, label %for.end ; Make sure the added GEP has the right index type -; CHECK: %lftr.limit = getelementptr i8, i8 addrspace(3)* %base, i16 +; CHECK: %lftr.limit = getelementptr i8, i8 addrspace(3)* %base, i16 %0 ; CHECK: for.body: ; CHECK: phi i8 addrspace(3)* diff --git a/test/Transforms/IndVarSimplify/lftr-reuse.ll b/test/Transforms/IndVarSimplify/lftr-reuse.ll index befbb9e3f15..b4e1fdef488 100644 --- a/test/Transforms/IndVarSimplify/lftr-reuse.ll +++ b/test/Transforms/IndVarSimplify/lftr-reuse.ll @@ -82,23 +82,15 @@ exit: ; Perform LFTR without generating extra preheader code. define void @guardedloop([0 x double]* %matrix, [0 x double]* %vector, i32 %irow, i32 %ilead) nounwind { -; CHECK-LABEL: @guardedloop( -; CHECK-LABEL: entry: -; CHECK-NEXT: %[[cmp:.*]] = icmp slt i32 1, %irow -; CHECK-NEXT: br i1 %[[cmp]], label %[[loop_preheader:.*]], label %[[return:.*]] - -; CHECK: [[loop_preheader]]: -; CHECK-NEXT: %[[sext:.*]] = sext i32 %ilead to i64 -; CHECK-NEXT: %[[add:.*]] = add i32 %irow, -1 -; CHECK-NEXT: br label %[[loop:.*]] - -; CHECK: [[loop]]: -; CHECK-NEXT: %[[indvars_iv2:.*]] = phi i64 -; CHECK-NEXT: phi i64 +; CHECK: entry: +; CHECK-NOT: zext +; CHECK-NOT: add +; CHECK: loop: +; CHECK: phi i64 +; CHECK: phi i64 ; CHECK-NOT: phi -; CHECK: %[[lftr_wideiv:.*]] = trunc i64 %[[indvars_iv2]] to i32 -; CHECK-NEXT: %[[exitcond:.*]] = icmp ne i32 %[[lftr_wideiv]], %[[add]] -; CHECK-NEXT: br i1 %[[exitcond]], label %[[loop]], label +; CHECK: icmp ne +; CHECK: br i1 entry: %cmp = icmp slt i32 1, %irow br i1 %cmp, label %loop, label %return