[llvm] r299526 - [LAA] Correctly return a half-open range in expandBounds

[James Molloy via llvm-commits]

Author: jamesm
Date: Wed Apr  5 04:24:26 2017
New Revision: 299526

URL: http://llvm.org/viewvc/llvm-project?rev=299526&view=rev
Log:
[LAA] Correctly return a half-open range in expandBounds

This is a latent bug that's been hanging around for a while. For a loop-invariant
pointer, expandBounds would return the range {Ptr, Ptr}, but this was interpreted
as a half-open range, not a closed range. So we ended up planting incorrect
bounds checks. Even worse, they were tautological, so we ended up incorrectly
executing the optimized loop.

Modified:
    llvm/trunk/lib/Analysis/LoopAccessAnalysis.cpp
    llvm/trunk/test/Transforms/LoopVersioning/loop-invariant-bound.ll
    llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM1.ll
    llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM2.ll

Modified: llvm/trunk/lib/Analysis/LoopAccessAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LoopAccessAnalysis.cpp?rev=299526&r1=299525&r2=299526&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LoopAccessAnalysis.cpp (original)
+++ llvm/trunk/lib/Analysis/LoopAccessAnalysis.cpp Wed Apr  5 04:24:26 2017
@@ -1939,7 +1939,10 @@ expandBounds(const RuntimePointerCheckin
     Value *NewPtr = (Inst && TheLoop->contains(Inst))
                         ? Exp.expandCodeFor(Sc, PtrArithTy, Loc)
                         : Ptr;
-    return {NewPtr, NewPtr};
+    // We must return a half-open range, which means incrementing Sc.
+    const SCEV *ScPlusOne = SE->getAddExpr(Sc, SE->getOne(PtrArithTy));
+    Value *NewPtrPlusOne = Exp.expandCodeFor(ScPlusOne, PtrArithTy, Loc);
+    return {NewPtr, NewPtrPlusOne};
   } else {
     Value *Start = nullptr, *End = nullptr;
     DEBUG(dbgs() << "LAA: Adding RT check for range:\n");

Modified: llvm/trunk/test/Transforms/LoopVersioning/loop-invariant-bound.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopVersioning/loop-invariant-bound.ll?rev=299526&r1=299525&r2=299526&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/LoopVersioning/loop-invariant-bound.ll (original)
+++ llvm/trunk/test/Transforms/LoopVersioning/loop-invariant-bound.ll Wed Apr  5 04:24:26 2017
@@ -8,12 +8,13 @@ target datalayout = "e-m:e-i64:64-f80:12
 %Partials.215 = type { [2 x %Dual.213] }
 
 ; Function Attrs: sspreq
-define void @"julia_axpy!_65480"(%Dual.212*) {
+define void @"julia_axpy!_65480"(%Dual.212*, %Dual.212* %other) {
 top:
   br label %if24
 
 ; CHECK-NOT: %bc = bitcast i64* %v2.sroa.0.0..sroa_cast
-; CHECK: %bound0
+; CHECK: %bound0 = icmp ult i8* %[[x:[a-z0-9]+]], %[[y:[a-z0-9]+]]
+; CHECK-NOT: %bound1 = icmp ult i8* %[[y]], %[[x]]
 
 if24:                                             ; preds = %if24, %top
   %"#temp#1.sroa.3.02" = phi i64 [ undef, %top ], [ %2, %if24 ]
@@ -24,7 +25,7 @@ if24:
   %v2.sroa.0.0..sroa_cast = bitcast %Dual.212* %0 to i64*
   %v2.sroa.0.0.copyload = load i64, i64* %v2.sroa.0.0..sroa_cast, align 1
   %3 = add i64 %"#temp#1.sroa.0.01", -1
-  %4 = getelementptr inbounds %Dual.212, %Dual.212* undef, i64 %3, i32 1, i32 0, i64 0, i32 1, i32 0, i64 0
+  %4 = getelementptr inbounds %Dual.212, %Dual.212* %other, i64 0, i32 1, i32 0, i64 0, i32 1, i32 0, i64 0
   %5 = bitcast double* %4 to i64*
   store i64 undef, i64* %5, align 8
   %notlhs27 = icmp eq i64 %2, undef

Modified: llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM1.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM1.ll?rev=299526&r1=299525&r2=299526&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM1.ll (original)
+++ llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM1.ll Wed Apr  5 04:24:26 2017
@@ -16,7 +16,7 @@
 ; CHECK-NEXT: %add8 = add nsw i32 %[[induction]], %add
 ; CHECK-NEXT: %inc = add nuw i32 %j.113, 1
 ; CHECK-NEXT: %cmp2 = icmp ult i32 %inc, %itr
-; CHECK-NEXT: br i1 %cmp2, label %for.body3, label %for.inc11.loopexit.loopexit6, !llvm.loop !5
+; CHECK-NEXT: br i1 %cmp2, label %for.body3, label %for.inc11.loopexit.loopexit7, !llvm.loop !5
 define i32 @foo(i32* nocapture %var1, i32* nocapture readnone %var2, i32* nocapture %var3, i32 %itr) #0 {
 entry:
   %cmp14 = icmp eq i32 %itr, 0

Modified: llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM2.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM2.ll?rev=299526&r1=299525&r2=299526&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM2.ll (original)
+++ llvm/trunk/test/Transforms/LoopVersioningLICM/loopversioningLICM2.ll Wed Apr  5 04:24:26 2017
@@ -7,7 +7,7 @@
 ; CHECK: Loop: Loop at depth 2 containing: %for.body3.us<header><latch><exiting>
 ; CHECK-NEXT:     Loop Versioning found to be beneficial
 ;
-; CHECK: for.cond1.for.inc17_crit_edge.us.loopexit5:       ; preds = %for.body3.us
+; CHECK: for.cond1.for.inc17_crit_edge.us.loopexit6:       ; preds = %for.body3.us
 ; CHECK-NEXT: %add14.us.lcssa = phi float [ %add14.us, %for.body3.us ]
 ; CHECK-NEXT: store float %add14.us.lcssa, float* %arrayidx.us, align 4, !alias.scope !0, !noalias !0
 ; CHECK-NEXT: br label %for.cond1.for.inc17_crit_edge.us