It's not necessary to do rounding for alloca operations when the requested

alignment is equal to the stack alignment.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@40004 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 29 insertions, 0 deletions

diff --git a/test/Transforms/LoopStrengthReduce/phi_node_update_multiple_preds.ll b/test/Transforms/LoopStrengthReduce/phi_node_update_multiple_preds.ll
new file mode 100644
index 0000000..cd0afe1
--- /dev/null
+++ b/test/Transforms/LoopStrengthReduce/phi_node_update_multiple_preds.ll
@@ -0,0 +1,29 @@
+; RUN: llvm-upgrade < %s | llvm-as | opt -loop-reduce -disable-output
+; LSR should not crash on this.
+
+fastcc void %loadloop() {
+entry:
+	switch sbyte 0, label %shortcirc_next [
+		 sbyte 32, label %loopexit.2
+		 sbyte 59, label %loopexit.2
+	]
+
+shortcirc_next:		; preds = %no_exit.2, %entry
+	%indvar37 = phi uint [ 0, %entry ], [ %indvar.next38, %no_exit.2 ]		; <uint> [#uses=3]
+	%wp.2.4 = getelementptr sbyte* null, uint %indvar37		; <sbyte*> [#uses=1]
+	br bool false, label %loopexit.2, label %no_exit.2
+
+no_exit.2:		; preds = %shortcirc_next
+	%wp.2.4.rec = cast uint %indvar37 to int		; <int> [#uses=1]
+	%inc.1.rec = add int %wp.2.4.rec, 1		; <int> [#uses=1]
+	%inc.1 = getelementptr sbyte* null, int %inc.1.rec		; <sbyte*> [#uses=2]
+	%indvar.next38 = add uint %indvar37, 1		; <uint> [#uses=1]
+	switch sbyte 0, label %shortcirc_next [
+		 sbyte 32, label %loopexit.2
+		 sbyte 59, label %loopexit.2
+	]
+
+loopexit.2:		; preds = %no_exit.2, %no_exit.2, %shortcirc_next, %entry, %entry
+	%wp.2.7 = phi sbyte* [ null, %entry ], [ null, %entry ], [ %wp.2.4, %shortcirc_next ], [ %inc.1, %no_exit.2 ], [ %inc.1, %no_exit.2 ]		; <sbyte*> [#uses=0]
+	ret void
+}