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/buildvec_canonicalize.ll b/test/CodeGen/PowerPC/buildvec_canonicalize.ll
new file mode 100644
index 0000000..54cbdae
--- /dev/null
+++ b/test/CodeGen/PowerPC/buildvec_canonicalize.ll
@@ -0,0 +1,27 @@
+; There should be exactly one vxor here.
+; RUN: llvm-upgrade < %s | llvm-as | \
+; RUN:   llc -march=ppc32 -mcpu=g5 --enable-unsafe-fp-math | \
+; RUN:   grep vxor | wc -l | grep 1
+
+; There should be exactly one vsplti here.
+; RUN: llvm-upgrade < %s | llvm-as | \
+; RUN:   llc -march=ppc32 -mcpu=g5 --enable-unsafe-fp-math | \
+; RUN:   grep vsplti | wc -l | grep 1
+
+
+void %VXOR(<4 x float>* %P1, <4 x int>* %P2, <4 x float>* %P3) {
+        %tmp = load <4 x float>* %P3
+        %tmp3 = load <4 x float>* %P1
+        %tmp4 = mul <4 x float> %tmp, %tmp3
+        store <4 x float> %tmp4, <4 x float>* %P3
+        store <4 x float> zeroinitializer, <4 x float>* %P1
+        store <4 x int> zeroinitializer, <4 x int>* %P2
+        ret void
+}
+
+void %VSPLTI(<4 x int>* %P2, <8 x short>* %P3) {
+        store <4 x int> cast (<16 x sbyte> < sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1, sbyte -1 > to <4 x int>), <4 x int>* %P2
+        store <8 x short> < short -1, short -1, short -1, short -1, short -1, short -1, short -1, short -1 >, <8 x short>* %P3
+        ret void
+}
+