diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td
index 4b283a2988a..3874c1968b5 100644
--- a/lib/Target/X86/X86InstrSSE.td
+++ b/lib/Target/X86/X86InstrSSE.td
@@ -3939,6 +3939,14 @@ def MOVNTI_64mr : RI<0xC3, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src),
                   PS, Requires<[HasSSE2]>;
 } // SchedRW = [WriteStore]
 
+let Predicates = [HasAVX, NoVLX] in {
+  def : Pat<(alignednontemporalstore (v4i32 VR128:$src), addr:$dst),
+            (VMOVNTPSmr addr:$dst, VR128:$src)>;
+}
+
+def : Pat<(alignednontemporalstore (v4i32 VR128:$src), addr:$dst),
+          (MOVNTPSmr addr:$dst, VR128:$src)>;
+
 } // AddedComplexity
 
 //===----------------------------------------------------------------------===//
diff --git a/test/CodeGen/X86/nontemporal-2.ll b/test/CodeGen/X86/nontemporal-2.ll
new file mode 100644
index 00000000000..9d0cb9a5edf
--- /dev/null
+++ b/test/CodeGen/X86/nontemporal-2.ll
@@ -0,0 +1,31 @@
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=corei7 | FileCheck %s -check-prefix=CHECK -check-prefix=SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=corei7-avx | FileCheck %s -check-prefix=CHECK -check-prefix=AVX
+
+
+; Make sure that we generate non-temporal stores for the test cases below.
+
+define void @test1(<4 x float>* %dst) {
+; CHECK-LABEL: test1:
+; SSE: movntps
+; AVX: vmovntps
+  store <4 x float> zeroinitializer, <4 x float>* %dst, align 16, !nontemporal !1
+  ret void
+}
+
+define void @test2(<4 x i32>* %dst) {
+; CHECK-LABEL: test2:
+; SSE: movntps
+; AVX: vmovntps
+  store <4 x i32> zeroinitializer, <4 x i32>* %dst, align 16, !nontemporal !1
+  ret void
+}
+
+define void @test3(<2 x double>* %dst) {
+; CHECK-LABEL: test3:
+; SSE: movntps
+; AVX: vmovntps
+  store <2 x double> zeroinitializer, <2 x double>* %dst, align 16, !nontemporal !1
+  ret void
+}
+
+!1 = metadata !{i32 1}