[X86] Add some optsize patterns for 256-bit X86vzmovl.

These patterns use VMOVSS/SD. Without optsize we use BLENDI instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@337119 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 5c8b612f49be59d19221288b6e44448cba77fb07..eb6a33234911d7ba7717b2ce335df098139125e4 100644 (file)
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -304,6 +304,25 @@ let Predicates = [UseAVX, OptForSize] in {
             (VMOVSSrr (v4f32 (V_SET0)), VR128:$src)>;
   def : Pat<(v4i32 (X86vzmovl (v4i32 VR128:$src))),
             (VMOVSSrr (v4i32 (V_SET0)), VR128:$src)>;
+
+  // Move low f32 and clear high bits.
+  def : Pat<(v8f32 (X86vzmovl (v8f32 VR256:$src))),
+            (SUBREG_TO_REG (i32 0),
+             (VMOVSSrr (v4f32 (V_SET0)),
+              (EXTRACT_SUBREG (v8f32 VR256:$src), sub_xmm)), sub_xmm)>;
+  def : Pat<(v8i32 (X86vzmovl (v8i32 VR256:$src))),
+            (SUBREG_TO_REG (i32 0),
+             (VMOVSSrr (v4i32 (V_SET0)),
+              (EXTRACT_SUBREG (v8i32 VR256:$src), sub_xmm)), sub_xmm)>;
+
+  def : Pat<(v4f64 (X86vzmovl (v4f64 VR256:$src))),
+            (SUBREG_TO_REG (i32 0),
+             (VMOVSDrr (v2f64 (V_SET0)),
+                       (EXTRACT_SUBREG (v4f64 VR256:$src), sub_xmm)), sub_xmm)>;
+  def : Pat<(v4i64 (X86vzmovl (v4i64 VR256:$src))),
+            (SUBREG_TO_REG (i32 0),
+             (VMOVSDrr (v2i64 (V_SET0)),
+                       (EXTRACT_SUBREG (v4i64 VR256:$src), sub_xmm)), sub_xmm)>;
 }
 
 let Predicates = [UseSSE1] in {

diff --git a/test/CodeGen/X86/vector-shuffle-256-v4.ll b/test/CodeGen/X86/vector-shuffle-256-v4.ll
index 03e159f3d4444334d991929c757b7cf7968e9259..e125601a5889f323b8134b4928c11a941a6a49bb 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v4.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v4.ll
@@ -1882,3 +1882,67 @@ entry:
   %add = add <4 x i64> %shuffle, %shuffle1
   ret <4 x i64> %add
 }
+
+define <4 x double> @shuffle_v4f64_0zzz_optsize(<4 x double> %a) optsize {
+; AVX1OR2-LABEL: shuffle_v4f64_0zzz_optsize:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
+; AVX1OR2-NEXT:    vmovsd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v4f64_0zzz_optsize:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-NEXT:    vmovsd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; AVX512VL-NEXT:    retq
+  %b = shufflevector <4 x double> %a, <4 x double> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
+  ret <4 x double> %b
+}
+
+define <4 x i64> @shuffle_v4i64_0zzz_optsize(<4 x i64> %a) optsize {
+; AVX1OR2-LABEL: shuffle_v4i64_0zzz_optsize:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vxorpd %xmm1, %xmm1, %xmm1
+; AVX1OR2-NEXT:    vmovsd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v4i64_0zzz_optsize:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-NEXT:    vmovsd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; AVX512VL-NEXT:    retq
+  %b = shufflevector <4 x i64> %a, <4 x i64> zeroinitializer, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
+  ret <4 x i64> %b
+}
+
+define <8 x float> @shuffle_v8f32_0zzzzzzz_optsize(<8 x float> %a) optsize {
+; AVX1OR2-LABEL: shuffle_v8f32_0zzzzzzz_optsize:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; AVX1OR2-NEXT:    vmovss {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8f32_0zzzzzzz_optsize:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-NEXT:    vmovss {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
+; AVX512VL-NEXT:    retq
+  %b = shufflevector <8 x float> %a, <8 x float> zeroinitializer, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  ret <8 x float> %b
+}
+
+define <8 x i32> @shuffle_v8i32_0zzzzzzz_optsize(<8 x i32> %a) optsize {
+; AVX1OR2-LABEL: shuffle_v8i32_0zzzzzzz_optsize:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; AVX1OR2-NEXT:    vmovss {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
+; AVX1OR2-NEXT:    retq
+;
+; AVX512VL-LABEL: shuffle_v8i32_0zzzzzzz_optsize:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-NEXT:    vmovss {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
+; AVX512VL-NEXT:    retq
+  %b = shufflevector <8 x i32> %a, <8 x i32> zeroinitializer, <8 x i32> <i32 0, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  ret <8 x i32> %b
+}