summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorEvan Cheng <evan.cheng@apple.com>2006-05-03 20:32:03 +0000
committerEvan Cheng <evan.cheng@apple.com>2006-05-03 20:32:03 +0000
commit8b1cde2bbe3924f6fa85ced9074d216405fa5bb0 (patch)
tree78c8205ad7344acd01b58ea2b93db12a67122fe1
parente3a9c70ba0d798fec80b63ea1507fa546c41bfc3 (diff)
downloadbcm5719-llvm-8b1cde2bbe3924f6fa85ced9074d216405fa5bb0.tar.gz
bcm5719-llvm-8b1cde2bbe3924f6fa85ced9074d216405fa5bb0.zip
Use movsd to shuffle in the lowest two elements of a v4f32 / v4i32 vector when
movlps cannot be used (e.g. when load from m64 has multiple uses). llvm-svn: 28089
-rw-r--r--llvm/lib/Target/X86/X86InstrSSE.td8
-rw-r--r--llvm/test/Regression/CodeGen/X86/vec_shuffle-5.ll11
2 files changed, 19 insertions, 0 deletions
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index 1474881295e..6b799945b9d 100644
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -2481,6 +2481,14 @@ def : Pat<(v2i64 (vector_shuffle VR128:$src1, VR128:$src2,
MOVL_shuffle_mask)),
(v2i64 (MOVLPDrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
+// vector_shuffle v1, v2 <4, 5, 2, 3> using MOVLPDrr (movsd)
+def : Pat<(v4f32 (vector_shuffle VR128:$src1, VR128:$src2,
+ MOVLP_shuffle_mask)),
+ (v4f32 (MOVLPDrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
+def : Pat<(v4i32 (vector_shuffle VR128:$src1, VR128:$src2,
+ MOVLP_shuffle_mask)),
+ (v4i32 (MOVLPDrr VR128:$src1, VR128:$src2))>, Requires<[HasSSE2]>;
+
// Set lowest element and zero upper elements.
def : Pat<(bc_v2i64 (vector_shuffle immAllZerosV,
(v2f64 (scalar_to_vector (loadf64 addr:$src))),
diff --git a/llvm/test/Regression/CodeGen/X86/vec_shuffle-5.ll b/llvm/test/Regression/CodeGen/X86/vec_shuffle-5.ll
new file mode 100644
index 00000000000..ac9c667aa45
--- /dev/null
+++ b/llvm/test/Regression/CodeGen/X86/vec_shuffle-5.ll
@@ -0,0 +1,11 @@
+; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | grep movsd | wc -l | grep 1
+
+void %test() {
+ %tmp1 = load <4 x float>* null
+ %tmp2 = shufflevector <4 x float> %tmp1, <4 x float> < float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00 >, <4 x uint> < uint 0, uint 1, uint 6, uint 7 >
+ %tmp3 = shufflevector <4 x float> %tmp1, <4 x float> zeroinitializer, <4 x uint> < uint 2, uint 3, uint 6, uint 7 >
+ %tmp4 = add <4 x float> %tmp2, %tmp3
+ store <4 x float> %tmp4, <4 x float>* null
+ ret void
+}
+
OpenPOWER on IntegriCloud