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/vec_mul.ll b/test/CodeGen/PowerPC/vec_mul.ll
new file mode 100644
index 0000000..eea1def
--- /dev/null
+++ b/test/CodeGen/PowerPC/vec_mul.ll
@@ -0,0 +1,24 @@
+; RUN: llvm-upgrade < %s | llvm-as | llc -march=ppc32 -mcpu=g5 | not grep mullw
+; RUN: llvm-upgrade < %s | llvm-as | llc -march=ppc32 -mcpu=g5 | grep vmsumuhm
+
+<4 x int> %test_v4i32(<4 x int>* %X, <4 x int>* %Y) {
+ %tmp = load <4 x int>* %X
+ %tmp2 = load <4 x int>* %Y
+ %tmp3 = mul <4 x int> %tmp, %tmp2
+ ret <4 x int> %tmp3
+}
+
+<8 x short> %test_v8i16(<8 x short>* %X, <8 x short>* %Y) {
+ %tmp = load <8 x short>* %X
+ %tmp2 = load <8 x short>* %Y
+ %tmp3 = mul <8 x short> %tmp, %tmp2
+ ret <8 x short> %tmp3
+}
+
+<16 x sbyte> %test_v16i8(<16 x sbyte>* %X, <16 x sbyte>* %Y) {
+ %tmp = load <16 x sbyte>* %X
+ %tmp2 = load <16 x sbyte>* %Y
+ %tmp3 = mul <16 x sbyte> %tmp, %tmp2
+ ret <16 x sbyte> %tmp3
+}
+