Use PMULDQ for v2i64 multiplies when SSE4.1 is available. And add

load-folding table entries for PMULDQ and PMULLD.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51489 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 32 insertions, 0 deletions

diff --git a/test/CodeGen/X86/pmul.ll b/test/CodeGen/X86/pmul.ll
new file mode 100644
index 0000000..3af47f6
--- /dev/null
+++ b/test/CodeGen/X86/pmul.ll
@@ -0,0 +1,32 @@
+; RUN: llvm-as < %s | llc -march=x86 -mattr=sse41 > %t
+; RUN: grep pmul %t | count 6
+; RUN: grep mov %t | count 8
+
+define <4 x i32> @a(<4 x i32> %i) nounwind  {
+        %A = mul <4 x i32> %i, < i32 117, i32 117, i32 117, i32 117 >
+        ret <4 x i32> %A
+}
+define <2 x i64> @b(<2 x i64> %i) nounwind  {
+        %A = mul <2 x i64> %i, < i64 117, i64 117 >
+        ret <2 x i64> %A
+}
+define <4 x i32> @c(<4 x i32> %i, <4 x i32> %j) nounwind  {
+        %A = mul <4 x i32> %i, %j
+        ret <4 x i32> %A
+}
+define <2 x i64> @d(<2 x i64> %i, <2 x i64> %j) nounwind  {
+        %A = mul <2 x i64> %i, %j
+        ret <2 x i64> %A
+}
+; Use a call to force spills.
+declare void @foo()
+define <4 x i32> @e(<4 x i32> %i, <4 x i32> %j) nounwind  {
+        call void @foo()
+        %A = mul <4 x i32> %i, %j
+        ret <4 x i32> %A
+}
+define <2 x i64> @f(<2 x i64> %i, <2 x i64> %j) nounwind  {
+        call void @foo()
+        %A = mul <2 x i64> %i, %j
+        ret <2 x i64> %A
+}