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
diff --git a/test/CodeGen/PowerPC/2006-12-07-LargeAlloca.ll b/test/CodeGen/PowerPC/2006-12-07-LargeAlloca.ll
new file mode 100644
index 0000000..f6103e5
--- /dev/null
+++ b/test/CodeGen/PowerPC/2006-12-07-LargeAlloca.ll
@@ -0,0 +1,26 @@
+; RUN: llvm-upgrade < %s | llvm-as | llc -march=ppc64
+; RUN: llvm-upgrade < %s | llvm-as | llc -march=ppc32
+; RUN: llvm-upgrade < %s | llvm-as | llc
+
+void %bitap() {
+entry:
+ %RMask.i = alloca [256 x uint], align 16 ; <[256 x uint]*> [#uses=1]
+ %buffer = alloca [147456 x sbyte], align 16 ; <[147456 x sbyte]*> [#uses=0]
+ br bool false, label %bb19, label %bb.preheader
+
+bb.preheader: ; preds = %entry
+ ret void
+
+bb19: ; preds = %entry
+ br bool false, label %bb12.i, label %cond_next39
+
+bb12.i: ; preds = %bb12.i, %bb19
+ %i.0.i = phi uint [ %tmp11.i, %bb12.i ], [ 0, %bb19 ] ; <uint> [#uses=2]
+ %tmp9.i = getelementptr [256 x uint]* %RMask.i, int 0, uint %i.0.i ; <uint*> [#uses=1]
+ store uint 0, uint* %tmp9.i
+ %tmp11.i = add uint %i.0.i, 1 ; <uint> [#uses=1]
+ br label %bb12.i
+
+cond_next39: ; preds = %bb19
+ ret void
+}