[llvm] r217102 - IndVarSimplify: Don't let LFTR compare against a poison value
David Majnemer
david.majnemer at gmail.com
Wed Sep 3 16:03:18 PDT 2014
Author: majnemer
Date: Wed Sep 3 18:03:18 2014
New Revision: 217102
URL: http://llvm.org/viewvc/llvm-project?rev=217102&view=rev
Log:
IndVarSimplify: Don't let LFTR compare against a poison value
LinearFunctionTestReplace tries to use the *next* indvar to compare
against when possible. However, it may be the case that the calculation
for the next indvar has NUW/NSW flags and that it may only be safely
used inside the loop. Using it in a comparison to calculate the exit
condition could result in observing poison.
This fixes PR20680.
Differential Revision: http://reviews.llvm.org/D5174
Added:
llvm/trunk/test/Transforms/IndVarSimplify/pr20680.ll
Modified:
llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
llvm/trunk/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll
llvm/trunk/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll
llvm/trunk/test/Transforms/IndVarSimplify/lftr-extend-const.ll
llvm/trunk/test/Transforms/IndVarSimplify/lftr-reuse.ll
Modified: llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp?rev=217102&r1=217101&r2=217102&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp Wed Sep 3 18:03:18 2014
@@ -1641,15 +1641,27 @@ 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.
- CmpIndVar = IndVar->getIncomingValueForBlock(L->getExitingBlock());
+ llvm::Value *IncrementedIndvar = IndVar->getIncomingValueForBlock(L->getExitingBlock());
+ const auto *IncrementedIndvarSCEV =
+ cast<SCEVAddRecExpr>(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.
+ if (ScalarEvolution::maskFlags(IncrementedIndvarSCEV->getNoWrapFlags(),
+ SCEV::FlagNUW | SCEV::FlagNSW) ==
+ SCEV::FlagAnyWrap) {
+ // 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;
+ }
}
Value *ExitCnt = genLoopLimit(IndVar, IVCount, L, Rewriter, SE);
Modified: llvm/trunk/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll?rev=217102&r1=217101&r2=217102&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll (original)
+++ llvm/trunk/test/Transforms/IndVarSimplify/2011-10-27-lftrnull.ll Wed Sep 3 18:03:18 2014
@@ -6,7 +6,7 @@ target triple = "thumbv7-apple-darwin"
; CHECK-LABEL: @test(
; CHECK: if.end.i126:
-; CHECK: %exitcond = icmp ne i8* %incdec.ptr.i, getelementptr (i8* null, i32 undef)
+; CHECK: %exitcond = icmp ne i8* %destYPixelPtr.010.i, getelementptr (i8* null, i32 undef)
define void @test() nounwind {
entry:
br label %while.cond
Modified: llvm/trunk/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll?rev=217102&r1=217101&r2=217102&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll (original)
+++ llvm/trunk/test/Transforms/IndVarSimplify/lftr-address-space-pointers.ll Wed Sep 3 18:03:18 2014
@@ -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 addrspace(2)* %base, i8 %0
+; CHECK: %lftr.limit = getelementptr i8 addrspace(2)* %base, i8
; 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 addrspace(3)* %base, i16 %0
+; CHECK: %lftr.limit = getelementptr i8 addrspace(3)* %base, i16
; CHECK: for.body:
; CHECK: phi i8 addrspace(3)*
Modified: llvm/trunk/test/Transforms/IndVarSimplify/lftr-extend-const.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/lftr-extend-const.ll?rev=217102&r1=217101&r2=217102&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/lftr-extend-const.ll (original)
+++ llvm/trunk/test/Transforms/IndVarSimplify/lftr-extend-const.ll Wed Sep 3 18:03:18 2014
@@ -2,7 +2,7 @@
; CHECK-LABEL: @foo(
; CHECK-NOT: %lftr.wideiv = trunc i32 %indvars.iv.next to i16
-; CHECK: %exitcond = icmp ne i32 %indvars.iv.next, 512
+; CHECK: %exitcond = icmp ne i32 %indvars.iv, 511
define void @foo() #0 {
entry:
br label %for.body
@@ -21,7 +21,7 @@ for.end:
; Check that post-incrementing the backedge taken count does not overflow.
; CHECK-LABEL: @postinc(
-; CHECK: icmp eq i32 %indvars.iv.next, 256
+; CHECK: icmp eq i32 %indvars.iv, 255
define i32 @postinc() #0 {
entry:
br label %do.body
Modified: llvm/trunk/test/Transforms/IndVarSimplify/lftr-reuse.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/lftr-reuse.ll?rev=217102&r1=217101&r2=217102&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/lftr-reuse.ll (original)
+++ llvm/trunk/test/Transforms/IndVarSimplify/lftr-reuse.ll Wed Sep 3 18:03:18 2014
@@ -82,15 +82,23 @@ 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: entry:
-; CHECK-NOT: zext
-; CHECK-NOT: add
-; CHECK: loop:
-; CHECK: phi i64
-; CHECK: phi i64
+; 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-NOT: phi
-; CHECK: icmp ne
-; CHECK: br i1
+; 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
entry:
%cmp = icmp slt i32 1, %irow
br i1 %cmp, label %loop, label %return
Added: llvm/trunk/test/Transforms/IndVarSimplify/pr20680.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/pr20680.ll?rev=217102&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/pr20680.ll (added)
+++ llvm/trunk/test/Transforms/IndVarSimplify/pr20680.ll Wed Sep 3 18:03:18 2014
@@ -0,0 +1,219 @@
+; RUN: opt < %s -indvars -S | FileCheck %s
+
+ at a = common global i32 0, align 4
+ at c = common global i32 0, align 4
+ at b = common global i32 0, align 4
+
+define void @f() {
+; CHECK-LABEL: @f(
+; CHECK-LABEL: entry:
+; CHECK: br label %[[for_cond2_preheader:.*]]
+
+; CHECK: [[for_cond2_preheader]]:
+; CHECK-NEXT: %[[indvars_iv:.*]] = phi i32 [ %[[indvars_iv_next:.*]], %[[for_inc13:.*]] ], [ -14, %entry ]
+; br i1 {{.*}}, label %[[for_inc13]], label %
+entry:
+ %0 = load i32* @a, align 4
+ %tobool2 = icmp eq i32 %0, 0
+ %1 = load i32* @a, align 4
+ %tobool = icmp eq i32 %1, 0
+ br label %for.cond2.preheader
+
+for.cond2.preheader: ; preds = %for.inc13, %entry
+ %storemerge15 = phi i8 [ -14, %entry ], [ %inc14, %for.inc13 ]
+ br i1 %tobool2, label %for.inc13, label %for.body3.lr.ph
+
+for.body3.lr.ph: ; preds = %for.cond2.preheader
+ %tobool5 = icmp eq i8 %storemerge15, 0
+ %conv7 = sext i8 %storemerge15 to i32
+ %2 = add nsw i32 %conv7, 1
+ %3 = icmp ult i32 %2, 3
+ %div = select i1 %3, i32 %conv7, i32 0
+ br i1 %tobool5, label %for.body3.lr.ph.split.us, label %for.body3.lr.ph.for.body3.lr.ph.split_crit_edge
+
+for.body3.lr.ph.for.body3.lr.ph.split_crit_edge: ; preds = %for.body3.lr.ph
+ br label %for.body3.lr.ph.split
+
+for.body3.lr.ph.split.us: ; preds = %for.body3.lr.ph
+ br i1 %tobool, label %for.body3.lr.ph.split.us.split.us, label %for.body3.lr.ph.split.us.for.body3.lr.ph.split.us.split_crit_edge
+
+for.body3.lr.ph.split.us.for.body3.lr.ph.split.us.split_crit_edge: ; preds = %for.body3.lr.ph.split.us
+ br label %for.body3.lr.ph.split.us.split
+
+for.body3.lr.ph.split.us.split.us: ; preds = %for.body3.lr.ph.split.us
+ br label %for.body3.us.us
+
+for.body3.us.us: ; preds = %for.cond2.loopexit.us.us, %for.body3.lr.ph.split.us.split.us
+ br i1 true, label %cond.false.us.us, label %cond.end.us.us
+
+cond.false.us.us: ; preds = %for.body3.us.us
+ br label %cond.end.us.us
+
+cond.end.us.us: ; preds = %cond.false.us.us, %for.body3.us.us
+ %cond.us.us = phi i32 [ %div, %cond.false.us.us ], [ %conv7, %for.body3.us.us ]
+ %4 = load i32* @b, align 4
+ %cmp91.us.us = icmp slt i32 %4, 1
+ br i1 %cmp91.us.us, label %for.inc.lr.ph.us.us, label %for.cond2.loopexit.us.us
+
+for.cond2.loopexit.us.us: ; preds = %for.cond8.for.cond2.loopexit_crit_edge.us.us, %cond.end.us.us
+ br i1 true, label %for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa.us, label %for.body3.us.us
+
+for.inc.lr.ph.us.us: ; preds = %cond.end.us.us
+ br label %for.inc.us.us
+
+for.cond8.for.cond2.loopexit_crit_edge.us.us: ; preds = %for.inc.us.us
+ %inc.lcssa.us.us = phi i32 [ %inc.us.us, %for.inc.us.us ]
+ store i32 %inc.lcssa.us.us, i32* @b, align 4
+ br label %for.cond2.loopexit.us.us
+
+for.inc.us.us: ; preds = %for.inc.us.us, %for.inc.lr.ph.us.us
+ %5 = phi i32 [ %4, %for.inc.lr.ph.us.us ], [ %inc.us.us, %for.inc.us.us ]
+ %inc.us.us = add nsw i32 %5, 1
+ %cmp9.us.us = icmp slt i32 %inc.us.us, 1
+ br i1 %cmp9.us.us, label %for.inc.us.us, label %for.cond8.for.cond2.loopexit_crit_edge.us.us
+
+for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa.us: ; preds = %for.cond2.loopexit.us.us
+ %cond.lcssa.ph.us.ph.us = phi i32 [ %cond.us.us, %for.cond2.loopexit.us.us ]
+ br label %for.cond2.for.inc13_crit_edge.us-lcssa.us
+
+for.body3.lr.ph.split.us.split: ; preds = %for.body3.lr.ph.split.us.for.body3.lr.ph.split.us.split_crit_edge
+ br label %for.body3.us
+
+for.body3.us: ; preds = %for.cond2.loopexit.us, %for.body3.lr.ph.split.us.split
+ br i1 true, label %cond.false.us, label %cond.end.us
+
+cond.false.us: ; preds = %for.body3.us
+ br label %cond.end.us
+
+cond.end.us: ; preds = %cond.false.us, %for.body3.us
+ %cond.us = phi i32 [ %div, %cond.false.us ], [ %conv7, %for.body3.us ]
+ %6 = load i32* @b, align 4
+ %cmp91.us = icmp slt i32 %6, 1
+ br i1 %cmp91.us, label %for.inc.lr.ph.us, label %for.cond2.loopexit.us
+
+for.inc.us: ; preds = %for.inc.lr.ph.us, %for.inc.us
+ %7 = phi i32 [ %6, %for.inc.lr.ph.us ], [ %inc.us, %for.inc.us ]
+ %inc.us = add nsw i32 %7, 1
+ %cmp9.us = icmp slt i32 %inc.us, 1
+ br i1 %cmp9.us, label %for.inc.us, label %for.cond8.for.cond2.loopexit_crit_edge.us
+
+for.cond2.loopexit.us: ; preds = %for.cond8.for.cond2.loopexit_crit_edge.us, %cond.end.us
+ br i1 false, label %for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa, label %for.body3.us
+
+for.inc.lr.ph.us: ; preds = %cond.end.us
+ br label %for.inc.us
+
+for.cond8.for.cond2.loopexit_crit_edge.us: ; preds = %for.inc.us
+ %inc.lcssa.us = phi i32 [ %inc.us, %for.inc.us ]
+ store i32 %inc.lcssa.us, i32* @b, align 4
+ br label %for.cond2.loopexit.us
+
+for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa: ; preds = %for.cond2.loopexit.us
+ %cond.lcssa.ph.us.ph = phi i32 [ %cond.us, %for.cond2.loopexit.us ]
+ br label %for.cond2.for.inc13_crit_edge.us-lcssa.us
+
+for.cond2.for.inc13_crit_edge.us-lcssa.us: ; preds = %for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa, %for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa.us
+ %cond.lcssa.ph.us = phi i32 [ %cond.lcssa.ph.us.ph, %for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa ], [ %cond.lcssa.ph.us.ph.us, %for.cond2.for.inc13_crit_edge.us-lcssa.us.us-lcssa.us ]
+ br label %for.cond2.for.inc13_crit_edge
+
+for.body3.lr.ph.split: ; preds = %for.body3.lr.ph.for.body3.lr.ph.split_crit_edge
+ br i1 %tobool, label %for.body3.lr.ph.split.split.us, label %for.body3.lr.ph.split.for.body3.lr.ph.split.split_crit_edge
+
+for.body3.lr.ph.split.for.body3.lr.ph.split.split_crit_edge: ; preds = %for.body3.lr.ph.split
+ br label %for.body3.lr.ph.split.split
+
+for.body3.lr.ph.split.split.us: ; preds = %for.body3.lr.ph.split
+ br label %for.body3.us3
+
+for.body3.us3: ; preds = %for.cond2.loopexit.us11, %for.body3.lr.ph.split.split.us
+ br i1 false, label %cond.false.us4, label %cond.end.us5
+
+cond.false.us4: ; preds = %for.body3.us3
+ br label %cond.end.us5
+
+cond.end.us5: ; preds = %cond.false.us4, %for.body3.us3
+ %cond.us6 = phi i32 [ %div, %cond.false.us4 ], [ %conv7, %for.body3.us3 ]
+ %8 = load i32* @b, align 4
+ %cmp91.us7 = icmp slt i32 %8, 1
+ br i1 %cmp91.us7, label %for.inc.lr.ph.us12, label %for.cond2.loopexit.us11
+
+for.inc.us8: ; preds = %for.inc.lr.ph.us12, %for.inc.us8
+ %9 = phi i32 [ %8, %for.inc.lr.ph.us12 ], [ %inc.us9, %for.inc.us8 ]
+ %inc.us9 = add nsw i32 %9, 1
+ %cmp9.us10 = icmp slt i32 %inc.us9, 1
+ br i1 %cmp9.us10, label %for.inc.us8, label %for.cond8.for.cond2.loopexit_crit_edge.us13
+
+for.cond2.loopexit.us11: ; preds = %for.cond8.for.cond2.loopexit_crit_edge.us13, %cond.end.us5
+ br i1 true, label %for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa.us, label %for.body3.us3
+
+for.inc.lr.ph.us12: ; preds = %cond.end.us5
+ br label %for.inc.us8
+
+for.cond8.for.cond2.loopexit_crit_edge.us13: ; preds = %for.inc.us8
+ %inc.lcssa.us14 = phi i32 [ %inc.us9, %for.inc.us8 ]
+ store i32 %inc.lcssa.us14, i32* @b, align 4
+ br label %for.cond2.loopexit.us11
+
+for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa.us: ; preds = %for.cond2.loopexit.us11
+ %cond.lcssa.ph.ph.us = phi i32 [ %cond.us6, %for.cond2.loopexit.us11 ]
+ br label %for.cond2.for.inc13_crit_edge.us-lcssa
+
+for.body3.lr.ph.split.split: ; preds = %for.body3.lr.ph.split.for.body3.lr.ph.split.split_crit_edge
+ br label %for.body3
+
+for.cond8.for.cond2.loopexit_crit_edge: ; preds = %for.inc
+ %inc.lcssa = phi i32 [ %inc, %for.inc ]
+ store i32 %inc.lcssa, i32* @b, align 4
+ br label %for.cond2.loopexit
+
+for.cond2.loopexit: ; preds = %cond.end, %for.cond8.for.cond2.loopexit_crit_edge
+ br i1 false, label %for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa, label %for.body3
+
+for.body3: ; preds = %for.cond2.loopexit, %for.body3.lr.ph.split.split
+ br i1 false, label %cond.false, label %cond.end
+
+cond.false: ; preds = %for.body3
+ br label %cond.end
+
+cond.end: ; preds = %cond.false, %for.body3
+ %cond = phi i32 [ %div, %cond.false ], [ %conv7, %for.body3 ]
+ %10 = load i32* @b, align 4
+ %cmp91 = icmp slt i32 %10, 1
+ br i1 %cmp91, label %for.inc.lr.ph, label %for.cond2.loopexit
+
+for.inc.lr.ph: ; preds = %cond.end
+ br label %for.inc
+
+for.inc: ; preds = %for.inc, %for.inc.lr.ph
+ %11 = phi i32 [ %10, %for.inc.lr.ph ], [ %inc, %for.inc ]
+ %inc = add nsw i32 %11, 1
+ %cmp9 = icmp slt i32 %inc, 1
+ br i1 %cmp9, label %for.inc, label %for.cond8.for.cond2.loopexit_crit_edge
+
+for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa: ; preds = %for.cond2.loopexit
+ %cond.lcssa.ph.ph = phi i32 [ %cond, %for.cond2.loopexit ]
+ br label %for.cond2.for.inc13_crit_edge.us-lcssa
+
+for.cond2.for.inc13_crit_edge.us-lcssa: ; preds = %for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa, %for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa.us
+ %cond.lcssa.ph = phi i32 [ %cond.lcssa.ph.ph, %for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa ], [ %cond.lcssa.ph.ph.us, %for.cond2.for.inc13_crit_edge.us-lcssa.us-lcssa.us ]
+ br label %for.cond2.for.inc13_crit_edge
+
+for.cond2.for.inc13_crit_edge: ; preds = %for.cond2.for.inc13_crit_edge.us-lcssa, %for.cond2.for.inc13_crit_edge.us-lcssa.us
+ %cond.lcssa = phi i32 [ %cond.lcssa.ph, %for.cond2.for.inc13_crit_edge.us-lcssa ], [ %cond.lcssa.ph.us, %for.cond2.for.inc13_crit_edge.us-lcssa.us ]
+ store i32 %cond.lcssa, i32* @c, align 4
+ br label %for.inc13
+
+; CHECK: [[for_inc13]]:
+; CHECK-NEXT: %[[indvars_iv_next]] = add nuw nsw i32 %[[indvars_iv]], 1
+; CHECK-NEXT: %[[exitcond4:.*]] = icmp ne i32 %[[indvars_iv]], -1
+; CHECK-NEXT: br i1 %[[exitcond4]], label %[[for_cond2_preheader]], label %[[for_end15:.*]]
+for.inc13: ; preds = %for.cond2.for.inc13_crit_edge, %for.cond2.preheader
+ %inc14 = add i8 %storemerge15, 1
+ %cmp = icmp ugt i8 %inc14, 50
+ br i1 %cmp, label %for.cond2.preheader, label %for.end15
+
+; CHECK: [[for_end15]]:
+; CHECK-NEXT: ret void
+for.end15: ; preds = %for.inc13
+ ret void
+}
More information about the llvm-commits
mailing list