diff --git a/llvm/include/llvm/Target/TargetSelectionDAG.td b/llvm/include/llvm/Target/TargetSelectionDAG.td
index d94bdc67bf02..18ff04c079a4 100644
--- a/llvm/include/llvm/Target/TargetSelectionDAG.td
+++ b/llvm/include/llvm/Target/TargetSelectionDAG.td
@@ -492,6 +492,12 @@ def intrinsic_wo_chain : SDNode<"ISD::INTRINSIC_WO_CHAIN",
 // Do not use cvt directly. Use cvt forms below
 def cvt : SDNode<"ISD::CONVERT_RNDSAT", SDTConvertOp>;
 
+def SDT_assertext : SDTypeProfile<1, 1,
+  [SDTCisInt<0>, SDTCisInt<1>, SDTCisSameAs<1, 0>]>;
+def assertsext : SDNode<"ISD::AssertSext", SDT_assertext>;
+def assertzext : SDNode<"ISD::AssertZext", SDT_assertext>;
+
+
 //===----------------------------------------------------------------------===//
 // Selection DAG Condition Codes
 
diff --git a/llvm/lib/Target/AArch64/AArch64InstrNEON.td b/llvm/lib/Target/AArch64/AArch64InstrNEON.td
index 1b7e0f93b954..68a499b75336 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrNEON.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrNEON.td
@@ -64,11 +64,6 @@ def Neon_vextract : SDNode<"AArch64ISD::NEON_VEXTRACT", SDTypeProfile<1, 3,
                            [SDTCisVec<0>,  SDTCisSameAs<0, 1>,
                            SDTCisSameAs<0, 2>, SDTCisVT<3, i64>]>>;
 
-def SDT_assertext : SDTypeProfile<1, 1,
-  [SDTCisInt<0>, SDTCisInt<1>, SDTCisSameAs<1, 0>]>;
-def assertsext : SDNode<"ISD::AssertSext", SDT_assertext>;
-def assertzext : SDNode<"ISD::AssertZext", SDT_assertext>;
-
 //===----------------------------------------------------------------------===//
 // Addressing-mode instantiations
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/X86/README-SSE.txt b/llvm/lib/Target/X86/README-SSE.txt
index adfa7fa1232c..71329b066923 100644
--- a/llvm/lib/Target/X86/README-SSE.txt
+++ b/llvm/lib/Target/X86/README-SSE.txt
@@ -494,11 +494,6 @@ is memory.
 
 //===---------------------------------------------------------------------===//
 
-SSE4 extract-to-mem ops aren't being pattern matched because of the AssertZext
-sitting between the truncate and the extract.
-
-//===---------------------------------------------------------------------===//
-
 INSERTPS can match any insert (extract, imm1), imm2 for 4 x float, and insert
 any number of 0.0 simultaneously.  Currently we only use it for simple
 insertions.
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index 34b3b855bf1d..ea91b5b10423 100644
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -6210,10 +6210,8 @@ multiclass SS41I_extract8<bits<8> opc, string OpcodeStr> {
                  (ins i8mem:$dst, VR128:$src1, i32i8imm:$src2),
                  !strconcat(OpcodeStr,
                             "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                 []>;
-// FIXME:
-// There's an AssertZext in the way of writing the store pattern
-// (store (i8 (trunc (X86pextrb (v16i8 VR128:$src1), imm:$src2))), addr:$dst)
+                 [(store (i8 (trunc (assertzext (X86pextrb (v16i8 VR128:$src1),
+						 imm:$src2)))), addr:$dst)]>;
 }
 
 let Predicates = [HasAVX] in
@@ -6236,10 +6234,8 @@ multiclass SS41I_extract16<bits<8> opc, string OpcodeStr> {
                  (ins i16mem:$dst, VR128:$src1, i32i8imm:$src2),
                  !strconcat(OpcodeStr,
                   "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                 []>;
-// FIXME:
-// There's an AssertZext in the way of writing the store pattern
-// (store (i16 (trunc (X86pextrw (v16i8 VR128:$src1), imm:$src2))), addr:$dst)
+                 [(store (i16 (trunc (assertzext (X86pextrw (v8i16 VR128:$src1),
+						  imm:$src2)))), addr:$dst)]>;
 }
 
 let Predicates = [HasAVX] in
diff --git a/llvm/test/CodeGen/X86/extract-store.ll b/llvm/test/CodeGen/X86/extract-store.ll
new file mode 100644
index 000000000000..27d93804ba60
--- /dev/null
+++ b/llvm/test/CodeGen/X86/extract-store.ll
@@ -0,0 +1,22 @@
+; RUN: llc < %s -o - -mcpu=generic -march=x86-64 -mattr=+sse4.1 | FileCheck %s -check-prefix=SSE41
+; RUN: llc < %s -o - -mcpu=generic -march=x86-64 -mattr=+avx | FileCheck %s -check-prefix=AVX
+
+define void @pextrb(i8* nocapture %dst, <16 x i8> %foo) {
+; AVX: vpextrb
+; SSE41: pextrb
+; AVX-NOT: movb
+; SSE41-NOT: movb
+  %vecext = extractelement <16 x i8> %foo, i32 15
+  store i8 %vecext, i8* %dst, align 1
+  ret void
+}
+
+define void @pextrw(i16* nocapture %dst, <8 x i16> %foo) {
+; AVX: vpextrw
+; SSE41: pextrw
+; AVX-NOT: movw
+; SSE41-NOT: movw
+  %vecext = extractelement <8 x i16> %foo, i32 15
+  store i16 %vecext, i16* %dst, align 1
+  ret void
+}
diff --git a/llvm/test/CodeGen/X86/widen_load-2.ll b/llvm/test/CodeGen/X86/widen_load-2.ll
index 26815a422ec8..41bea859f474 100644
--- a/llvm/test/CodeGen/X86/widen_load-2.ll
+++ b/llvm/test/CodeGen/X86/widen_load-2.ll
@@ -149,9 +149,9 @@ define void @add31i8(%i8vec31* nocapture sret %ret, %i8vec31* %ap, %i8vec31* %bp
 ; CHECK: movdqa
 ; CHECK: paddb
 ; CHECK: paddb
-; CHECK: movq
 ; CHECK: pextrb
 ; CHECK: pextrw
+; CHECK: movq
 ; CHECK: ret
 	%a = load %i8vec31* %ap, align 16
 	%b = load %i8vec31* %bp, align 16