aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--lib/Transforms/Scalar/IndVarSimplify.cpp71
-rw-r--r--test/Transforms/IndVarSimplify/eliminate-rem.ll121
2 files changed, 192 insertions, 0 deletions
diff --git a/lib/Transforms/Scalar/IndVarSimplify.cpp b/lib/Transforms/Scalar/IndVarSimplify.cpp
index 2693f3a9a9..21cb73cbf2 100644
--- a/lib/Transforms/Scalar/IndVarSimplify.cpp
+++ b/lib/Transforms/Scalar/IndVarSimplify.cpp
@@ -98,6 +98,7 @@ namespace {
private:
void EliminateIVComparisons();
+ void EliminateIVRemainders();
void RewriteNonIntegerIVs(Loop *L);
ICmpInst *LinearFunctionTestReplace(Loop *L, const SCEV *BackedgeTakenCount,
@@ -398,6 +399,73 @@ void IndVarSimplify::EliminateIVComparisons() {
RecursivelyDeleteTriviallyDeadInstructions(Inst);
}
+void IndVarSimplify::EliminateIVRemainders() {
+ SmallVector<WeakVH, 16> DeadInsts;
+
+ // Look for SRem and URem users.
+ for (IVUsers::iterator I = IU->begin(), E = IU->end(); I != E; ++I) {
+ IVStrideUse &UI = *I;
+ BinaryOperator *Rem = dyn_cast<BinaryOperator>(UI.getUser());
+ if (!Rem) continue;
+
+ bool isSigned = Rem->getOpcode() == Instruction::SRem;
+ if (!isSigned && Rem->getOpcode() != Instruction::URem)
+ continue;
+
+ // We're only interested in the case where we know something about
+ // the numerator.
+ if (UI.getOperandValToReplace() != Rem->getOperand(0))
+ continue;
+
+ // Get the SCEVs for the ICmp operands.
+ const SCEV *S = SE->getSCEV(Rem->getOperand(0));
+ const SCEV *X = SE->getSCEV(Rem->getOperand(1));
+
+ // Simplify unnecessary loops away.
+ const Loop *ICmpLoop = LI->getLoopFor(Rem->getParent());
+ S = SE->getSCEVAtScope(S, ICmpLoop);
+ X = SE->getSCEVAtScope(X, ICmpLoop);
+
+ // i % n --> i if i is in [0,n).
+ if ((!isSigned || SE->isKnownNonNegative(S)) &&
+ SE->isKnownPredicate(isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT,
+ S, X))
+ Rem->replaceAllUsesWith(Rem->getOperand(0));
+ else {
+ // (i+1) % n --> (i+1)==n?0:(i+1) if i is in [0,n).
+ const SCEV *LessOne =
+ SE->getMinusSCEV(S, SE->getIntegerSCEV(1, S->getType()));
+ if ((!isSigned || SE->isKnownNonNegative(LessOne)) &&
+ SE->isKnownPredicate(isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT,
+ LessOne, X)) {
+ ICmpInst *ICmp = new ICmpInst(Rem, ICmpInst::ICMP_EQ,
+ Rem->getOperand(0), Rem->getOperand(1),
+ "tmp");
+ SelectInst *Sel =
+ SelectInst::Create(ICmp,
+ ConstantInt::get(Rem->getType(), 0),
+ Rem->getOperand(0), "tmp", Rem);
+ Rem->replaceAllUsesWith(Sel);
+ } else
+ continue;
+ }
+
+ // Inform IVUsers about the new users.
+ if (Instruction *I = dyn_cast<Instruction>(Rem->getOperand(0)))
+ IU->AddUsersIfInteresting(I);
+
+ DEBUG(dbgs() << "INDVARS: Simplified rem: " << *Rem << '\n');
+ DeadInsts.push_back(Rem);
+ }
+
+ // Now that we're done iterating through lists, clean up any instructions
+ // which are now dead.
+ while (!DeadInsts.empty())
+ if (Instruction *Inst =
+ dyn_cast_or_null<Instruction>(DeadInsts.pop_back_val()))
+ RecursivelyDeleteTriviallyDeadInstructions(Inst);
+}
+
bool IndVarSimplify::runOnLoop(Loop *L, LPPassManager &LPM) {
IU = &getAnalysis<IVUsers>();
LI = &getAnalysis<LoopInfo>();
@@ -427,6 +495,9 @@ bool IndVarSimplify::runOnLoop(Loop *L, LPPassManager &LPM) {
// Simplify ICmp IV users.
EliminateIVComparisons();
+ // Simplify SRem and URem IV users.
+ EliminateIVRemainders();
+
// Compute the type of the largest recurrence expression, and decide whether
// a canonical induction variable should be inserted.
const Type *LargestType = 0;
diff --git a/test/Transforms/IndVarSimplify/eliminate-rem.ll b/test/Transforms/IndVarSimplify/eliminate-rem.ll
new file mode 100644
index 0000000000..f756389398
--- /dev/null
+++ b/test/Transforms/IndVarSimplify/eliminate-rem.ll
@@ -0,0 +1,121 @@
+; RUN: opt -indvars -S < %s | FileCheck %s
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+
+; Indvars should be able to eliminate this srem.
+; CHECK: @simple
+; CHECK-NOT: rem
+; CHECK: ret
+
+define void @simple(i64 %arg, double* %arg3) nounwind {
+bb:
+ %t = icmp slt i64 0, %arg ; <i1> [#uses=1]
+ br i1 %t, label %bb4, label %bb12
+
+bb4: ; preds = %bb
+ br label %bb5
+
+bb5: ; preds = %bb4, %bb5
+ %t6 = phi i64 [ %t9, %bb5 ], [ 0, %bb4 ] ; <i64> [#uses=2]
+ %t7 = srem i64 %t6, %arg ; <i64> [#uses=1]
+ %t8 = getelementptr inbounds double* %arg3, i64 %t7 ; <double*> [#uses=1]
+ store double 0.000000e+00, double* %t8
+ %t9 = add nsw i64 %t6, 1 ; <i64> [#uses=2]
+ %t10 = icmp slt i64 %t9, %arg ; <i1> [#uses=1]
+ br i1 %t10, label %bb5, label %bb11
+
+bb11: ; preds = %bb5
+ br label %bb12
+
+bb12: ; preds = %bb11, %bb
+ ret void
+}
+
+; Indvars should be able to eliminate the (i+1)%n.
+; CHECK: @f
+; CHECK-NOT: rem
+; CHECK: rem
+; CHECK-NOT: rem
+; CHECK: ret
+
+define i32 @f(i64* %arg, i64 %arg1, i64 %arg2, i64 %arg3) nounwind {
+bb:
+ %t = icmp sgt i64 %arg1, 0 ; <i1> [#uses=1]
+ br i1 %t, label %bb4, label %bb54
+
+bb4: ; preds = %bb
+ br label %bb5
+
+bb5: ; preds = %bb49, %bb4
+ %t6 = phi i64 [ %t51, %bb49 ], [ 0, %bb4 ] ; <i64> [#uses=4]
+ %t7 = phi i32 [ %t50, %bb49 ], [ 0, %bb4 ] ; <i32> [#uses=2]
+ %t8 = add nsw i64 %t6, %arg1 ; <i64> [#uses=1]
+ %t9 = add nsw i64 %t8, -2 ; <i64> [#uses=1]
+ %t10 = srem i64 %t9, %arg1 ; <i64> [#uses=1]
+ %t11 = add nsw i64 %t10, 1 ; <i64> [#uses=1]
+ %t12 = add nsw i64 %t6, 1 ; <i64> [#uses=1]
+ %t13 = srem i64 %t12, %arg1 ; <i64> [#uses=1]
+ %t14 = icmp sgt i64 %arg1, 0 ; <i1> [#uses=1]
+ br i1 %t14, label %bb15, label %bb49
+
+bb15: ; preds = %bb5
+ br label %bb16
+
+bb16: ; preds = %bb44, %bb15
+ %t17 = phi i64 [ %t46, %bb44 ], [ 0, %bb15 ] ; <i64> [#uses=1]
+ %t18 = phi i32 [ %t45, %bb44 ], [ %t7, %bb15 ] ; <i32> [#uses=2]
+ %t19 = icmp sgt i64 %arg1, 0 ; <i1> [#uses=1]
+ br i1 %t19, label %bb20, label %bb44
+
+bb20: ; preds = %bb16
+ br label %bb21
+
+bb21: ; preds = %bb21, %bb20
+ %t22 = phi i64 [ %t41, %bb21 ], [ 0, %bb20 ] ; <i64> [#uses=4]
+ %t23 = phi i32 [ %t40, %bb21 ], [ %t18, %bb20 ] ; <i32> [#uses=1]
+ %t24 = mul i64 %t6, %arg1 ; <i64> [#uses=1]
+ %t25 = mul i64 %t13, %arg1 ; <i64> [#uses=1]
+ %t26 = add nsw i64 %t24, %t22 ; <i64> [#uses=1]
+ %t27 = mul i64 %t11, %arg1 ; <i64> [#uses=1]
+ %t28 = add nsw i64 %t25, %t22 ; <i64> [#uses=1]
+ %t29 = getelementptr inbounds i64* %arg, i64 %t26 ; <i64*> [#uses=1]
+ %t30 = add nsw i64 %t27, %t22 ; <i64> [#uses=1]
+ %t31 = getelementptr inbounds i64* %arg, i64 %t28 ; <i64*> [#uses=1]
+ %t32 = zext i32 %t23 to i64 ; <i64> [#uses=1]
+ %t33 = load i64* %t29 ; <i64> [#uses=1]
+ %t34 = getelementptr inbounds i64* %arg, i64 %t30 ; <i64*> [#uses=1]
+ %t35 = load i64* %t31 ; <i64> [#uses=1]
+ %t36 = add nsw i64 %t32, %t33 ; <i64> [#uses=1]
+ %t37 = add nsw i64 %t36, %t35 ; <i64> [#uses=1]
+ %t38 = load i64* %t34 ; <i64> [#uses=1]
+ %t39 = add nsw i64 %t37, %t38 ; <i64> [#uses=1]
+ %t40 = trunc i64 %t39 to i32 ; <i32> [#uses=2]
+ %t41 = add nsw i64 %t22, 1 ; <i64> [#uses=2]
+ %t42 = icmp slt i64 %t41, %arg1 ; <i1> [#uses=1]
+ br i1 %t42, label %bb21, label %bb43
+
+bb43: ; preds = %bb21
+ br label %bb44
+
+bb44: ; preds = %bb43, %bb16
+ %t45 = phi i32 [ %t18, %bb16 ], [ %t40, %bb43 ] ; <i32> [#uses=2]
+ %t46 = add nsw i64 %t17, 1 ; <i64> [#uses=2]
+ %t47 = icmp slt i64 %t46, %arg1 ; <i1> [#uses=1]
+ br i1 %t47, label %bb16, label %bb48
+
+bb48: ; preds = %bb44
+ br label %bb49
+
+bb49: ; preds = %bb48, %bb5
+ %t50 = phi i32 [ %t7, %bb5 ], [ %t45, %bb48 ] ; <i32> [#uses=2]
+ %t51 = add nsw i64 %t6, 1 ; <i64> [#uses=2]
+ %t52 = icmp slt i64 %t51, %arg1 ; <i1> [#uses=1]
+ br i1 %t52, label %bb5, label %bb53
+
+bb53: ; preds = %bb49
+ br label %bb54
+
+bb54: ; preds = %bb53, %bb
+ %t55 = phi i32 [ 0, %bb ], [ %t50, %bb53 ] ; <i32> [#uses=1]
+ ret i32 %t55
+}
generated by cgit v1.2.3-18-g5258 (git 2.32.0) at 2025-05-09 19:04:06 +0000