Index: lib/Target/X86/X86InstrSSE.td
===================================================================
--- lib/Target/X86/X86InstrSSE.td
+++ lib/Target/X86/X86InstrSSE.td
@@ -8517,6 +8517,14 @@
             (VCVTPH2PSrm addr:$src)>;
   def : Pat<(int_x86_vcvtph2ps_128 (vzload_v2i64 addr:$src)),
             (VCVTPH2PSrm addr:$src)>;
+
+  def : Pat<(store (f64 (vector_extract (v2f64 (bitconvert (v8i16
+                  (int_x86_vcvtps2ph_128 VR128:$src1, i32:$src2)))), (iPTR 0))),
+                   addr:$dst),
+                   (VCVTPS2PHmr addr:$dst, VR128:$src1, imm:$src2)>;
+  def : Pat<(store (v8i16 (int_x86_vcvtps2ph_256 VR256:$src1, i32:$src2)),
+                   addr:$dst),
+                   (VCVTPS2PHYmr addr:$dst, VR256:$src1, imm:$src2)>;
 }
 
 // Patterns for  matching conversions from float to half-float and vice versa.
Index: test/CodeGen/X86/f16c-intrinsics.ll
===================================================================
--- test/CodeGen/X86/f16c-intrinsics.ll
+++ test/CodeGen/X86/f16c-intrinsics.ll
@@ -39,6 +39,16 @@
 }
 declare <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float>, i32) nounwind readonly
 
+define void @test_x86_vcvtps2ph_128_m(<4 x i16>* nocapture %d, <4 x float> %a) nounwind {
+entry:
+  ; CHECK-LABEL: test_x86_vcvtps2ph_128_m:
+  ; CHECK-NOT: vmov
+  ; CHECK: vcvtps2ph  $3, %xmm0, (%
+  %0 = tail call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %a, i32 3)
+  %1 = shufflevector <8 x i16> %0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  store <4 x i16> %1, <4 x i16>* %d, align 8
+  ret void
+}
 
 define <8 x i16> @test_x86_vcvtps2ph_256(<8 x float> %a0) {
   ; CHECK-LABEL: test_x86_vcvtps2ph_256
@@ -61,3 +71,13 @@
   %res = tail call <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16> %bc) #2
   ret <4 x float> %res
 }
+
+define void @test_x86_vcvtps2ph_256_m(<8 x i16>* nocapture %d, <8 x float> %a) nounwind {
+entry:
+  ; CHECK-LABEL: test_x86_vcvtps2ph_256_m:
+  ; CHECK-NOT: vmov
+  ; CHECK: vcvtps2ph  $3, %ymm0, (%
+  %0 = tail call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %a, i32 3)
+  store <8 x i16> %0, <8 x i16>* %d, align 16
+  ret void
+}