Fix recent regression in gcc.dg/pr26719.c (6835035).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@70386 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Analysis/ScalarEvolution.cpp b/lib/Analysis/ScalarEvolution.cpp
index 7f6321792572c867644738e9e01c394544040e58..42d2fdd3629da5138975b770399443ce0fff21c2 100644 (file)
--- a/lib/Analysis/ScalarEvolution.cpp
+++ b/lib/Analysis/ScalarEvolution.cpp
@@ -812,11 +812,14 @@ SCEVHandle ScalarEvolution::getSignExtendExpr(const SCEVHandle &Op,
         SCEVHandle Step = AR->getStepRecurrence(*this);
 
         // Check whether the backedge-taken count can be losslessly casted to
-        // the addrec's type. The count is always unsigned.
+        // the addrec's type. The count needs to be the same whether sign
+        // extended or zero extended.
         SCEVHandle CastedBECount =
           getTruncateOrZeroExtend(BECount, Start->getType());
         if (BECount ==
-            getTruncateOrZeroExtend(CastedBECount, BECount->getType())) {
+            getTruncateOrZeroExtend(CastedBECount, BECount->getType()) &&
+            BECount ==
+            getTruncateOrSignExtend(CastedBECount, BECount->getType())) {
           const Type *WideTy =
             IntegerType::get(getTypeSizeInBits(Start->getType()) * 2);
           SCEVHandle SMul =

diff --git a/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll b/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll
new file mode 100644 (file)
index 0000000..153a181
--- /dev/null
+++ b/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll
@@ -0,0 +1,43 @@
+; RUN: llvm-as < %s | opt -loop-reduce | llvm-dis | grep mul | count 3
+; The multiply in bb2 must not be reduced to an add, as the sext causes the
+; %1 argument to become negative after a while.
+; ModuleID = '<stdin>'
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
+target triple = "i386-apple-darwin9.6"
+@table = common global [32 x [256 x i32]] zeroinitializer, align 32            ; <[32 x [256 x i32]]*> [#uses=2]
+
+define i32 @main() nounwind {
+bb4.thread:
+       br label %bb2
+
+bb2:           ; preds = %bb4, %bb2, %bb4.thread
+       %i.0.reg2mem.0.ph = phi i32 [ 0, %bb4.thread ], [ %i.0.reg2mem.0.ph, %bb2 ], [ %indvar.next9, %bb4 ]            ; <i32> [#uses=4]
+       %j.0.reg2mem.0 = phi i32 [ 0, %bb4.thread ], [ %indvar.next, %bb2 ], [ 0, %bb4 ]                ; <i32> [#uses=3]
+       %0 = trunc i32 %j.0.reg2mem.0 to i8             ; <i8> [#uses=1]
+       %1 = sext i8 %0 to i32          ; <i32> [#uses=1]
+       %2 = mul i32 %1, %i.0.reg2mem.0.ph              ; <i32> [#uses=1]
+       %3 = getelementptr [32 x [256 x i32]]* @table, i32 0, i32 %i.0.reg2mem.0.ph, i32 %j.0.reg2mem.0         ; <i32*> [#uses=1]
+       store i32 %2, i32* %3, align 4
+       %indvar.next = add i32 %j.0.reg2mem.0, 1                ; <i32> [#uses=2]
+       %exitcond = icmp eq i32 %indvar.next, 256               ; <i1> [#uses=1]
+       br i1 %exitcond, label %bb4, label %bb2
+
+bb4:           ; preds = %bb2
+       %indvar.next9 = add i32 %i.0.reg2mem.0.ph, 1            ; <i32> [#uses=2]
+       %exitcond10 = icmp eq i32 %indvar.next9, 32             ; <i1> [#uses=1]
+       br i1 %exitcond10, label %bb5, label %bb2
+
+bb5:           ; preds = %bb4
+       %4 = load i32* getelementptr ([32 x [256 x i32]]* @table, i32 0, i32 9, i32 132), align 16              ; <i32> [#uses=1]
+       %5 = icmp eq i32 %4, -1116              ; <i1> [#uses=1]
+       br i1 %5, label %bb7, label %bb6
+
+bb6:           ; preds = %bb5
+       tail call void @abort() noreturn nounwind
+       unreachable
+
+bb7:           ; preds = %bb5
+       ret i32 0
+}
+
+declare void @abort() noreturn nounwind