Index: llvm/lib/Transforms/Scalar/IndVarSimplify.cpp =================================================================== --- llvm/lib/Transforms/Scalar/IndVarSimplify.cpp +++ llvm/lib/Transforms/Scalar/IndVarSimplify.cpp @@ -2309,7 +2309,8 @@ // Returns true if the condition of \p BI being checked is invariant and can be // proved to be trivially true during at least first \p MaxIter iterations. static bool isTrivialCond(const Loop *L, BranchInst *BI, ScalarEvolution *SE, - bool ProvingLoopExit, const SCEV *MaxIter) { + bool ProvingLoopExit, const SCEV *MaxIter, + bool SkipLastIter) { ICmpInst::Predicate Pred; Value *LHS, *RHS; using namespace PatternMatch; @@ -2341,6 +2342,11 @@ ICmpInst::Predicate InvariantPred; const SCEV *InvariantLHS, *InvariantRHS; + if (SkipLastIter) { + const SCEV *One = SE->getOne(MaxIter->getType()); + MaxIter = SE->getMinusSCEV(MaxIter, One); + } + // Check if there is a loop-invariant predicate equivalent to our check. if (!SE->isLoopInvariantExitCondDuringFirstIterations( Pred, LHSS, RHSS, L, BI, MaxIter, InvariantPred, InvariantLHS, @@ -2422,6 +2428,7 @@ }; bool Changed = false; + bool SkipLastIter = false; SmallSet DominatingExitCounts; for (BasicBlock *ExitingBB : ExitingBlocks) { const SCEV *ExitCount = SE->getExitCount(L, ExitingBB); @@ -2429,18 +2436,42 @@ // Okay, we do not know the exit count here. Can we at least prove that it // will remain the same within iteration space? auto *BI = cast(ExitingBB->getTerminator()); - auto OptimizeCond = [&](bool Inverted) { - if (isTrivialCond(L, BI, SE, Inverted, MaxExitCount)) { + auto OptimizeCond = [&](bool Inverted, bool SkipLastIter) { + if (isTrivialCond(L, BI, SE, Inverted, MaxExitCount, SkipLastIter)) { FoldExit(ExitingBB, Inverted); return true; } return false; }; - if (OptimizeCond(false) || OptimizeCond(true)) + if (OptimizeCond(false, false) || OptimizeCond(true, false)) Changed = true; + else if (SkipLastIter) + // If we cannot prove it on the last loop iteration, try to benefit + // from the fact that this check will not be executed on the last + // iteration. + // TODO: What we really want is to query with (MaxExitCount - 1), + // and the check should only check the condition on pre-last iteration + // if the backedge was taken at least once. Consider loop: + // + // for (i = len; i != 0; i--) { ... check (i ult X) ... } + // + // If len = 0 then MaxExitCount = 0, and the pre-last iteration will + // not be executed at all. But SCEV does not always recognize this and + // may start thinking that i on the iteration (MaxExitCount - 1) is a + // huge value. So we just query this separately. + // + // The checks against MaxExitCount above can be removed once we find a + // good way to resolve this problem. + if (OptimizeCond(false, true) || OptimizeCond(true, true)) + Changed = true; continue; } + if (MaxExitCount == ExitCount) + // If the loop has more than 1 iteration, all further checks will be + // executed 1 iteration less. + SkipLastIter = true; + // If we know we'd exit on the first iteration, rewrite the exit to // reflect this. This does not imply the loop must exit through this // exit; there may be an earlier one taken on the first iteration. Index: llvm/test/Transforms/IndVarSimplify/predicated_ranges.ll =================================================================== --- llvm/test/Transforms/IndVarSimplify/predicated_ranges.ll +++ llvm/test/Transforms/IndVarSimplify/predicated_ranges.ll @@ -2,7 +2,7 @@ ; RUN: opt -indvars -S < %s | FileCheck %s ; RUN: opt -passes=indvars -S < %s | FileCheck %s -; TODO: should be able to remove the range check basing on the following facts: +; Check that we are able to remove the range check basing on the following facts: ; 0 <= len <= MAX_INT [1]; ; iv starts from len and goes down stopping at zero and [1], therefore ; 0 <= iv <= len [2]; @@ -21,8 +21,7 @@ ; CHECK-NEXT: br i1 [[ZERO_COND]], label [[EXIT:%.*]], label [[RANGE_CHECK_BLOCK:%.*]] ; CHECK: range_check_block: ; CHECK-NEXT: [[IV_NEXT]] = sub i32 [[IV]], 1 -; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[IV_NEXT]], [[LEN]] -; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[FAIL:%.*]] +; CHECK-NEXT: br i1 true, label [[BACKEDGE]], label [[FAIL:%.*]] ; CHECK: backedge: ; CHECK-NEXT: [[EL_PTR:%.*]] = getelementptr i32, i32* [[P]], i32 [[IV]] ; CHECK-NEXT: [[EL:%.*]] = load i32, i32* [[EL_PTR]], align 4