Index: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
@@ -13935,7 +13935,33 @@
     if (SDValue Res = LowerEXTRACT_VECTOR_ELT_SSE4(Op, DAG))
       return Res;
 
-  // TODO: handle v16i8.
+  // TODO: We only extract a single element from v16i8, we can probably afford
+  // to be more aggressive here before using the default approach of spilling to
+  // stack.
+  if (VT.getSizeInBits() == 8 && Op->isOnlyUserOf(Vec.getNode())) {
+    // Extract either the lowest i32 or any i16, and extract the sub-byte.
+    int DWordIdx = IdxVal / 4;
+    if (DWordIdx == 0) {
+      SDValue Res = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i32,
+                                DAG.getBitcast(MVT::v4i32, Vec),
+                                DAG.getIntPtrConstant(DWordIdx, dl));
+      int ShiftVal = (IdxVal % 4) * 8;
+      if (ShiftVal != 0)
+        Res = DAG.getNode(ISD::SRL, dl, MVT::i32, Res,
+                          DAG.getConstant(ShiftVal, dl, MVT::i32));
+      return DAG.getNode(ISD::TRUNCATE, dl, VT, Res);
+    }
+
+    int WordIdx = IdxVal / 2;
+    SDValue Res = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i16,
+                              DAG.getBitcast(MVT::v8i16, Vec),
+                              DAG.getIntPtrConstant(WordIdx, dl));
+    int ShiftVal = (IdxVal % 2) * 8;
+    if (ShiftVal != 0)
+      Res = DAG.getNode(ISD::SRL, dl, MVT::i16, Res,
+                        DAG.getConstant(ShiftVal, dl, MVT::i16));
+    return DAG.getNode(ISD::TRUNCATE, dl, VT, Res);
+  }
 
   if (VT.getSizeInBits() == 32) {
     if (IdxVal == 0)
Index: llvm/trunk/test/CodeGen/X86/extract-store.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/extract-store.ll
+++ llvm/trunk/test/CodeGen/X86/extract-store.ll
@@ -9,22 +9,14 @@
 define void @extract_i8_0(i8* nocapture %dst, <16 x i8> %foo) nounwind {
 ; SSE2-X32-LABEL: extract_i8_0:
 ; SSE2-X32:       # BB#0:
-; SSE2-X32-NEXT:    pushl %ebp
-; SSE2-X32-NEXT:    movl %esp, %ebp
-; SSE2-X32-NEXT:    andl $-16, %esp
-; SSE2-X32-NEXT:    subl $32, %esp
-; SSE2-X32-NEXT:    movl 8(%ebp), %eax
-; SSE2-X32-NEXT:    movaps %xmm0, (%esp)
-; SSE2-X32-NEXT:    movb (%esp), %cl
+; SSE2-X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; SSE2-X32-NEXT:    movd %xmm0, %ecx
 ; SSE2-X32-NEXT:    movb %cl, (%eax)
-; SSE2-X32-NEXT:    movl %ebp, %esp
-; SSE2-X32-NEXT:    popl %ebp
 ; SSE2-X32-NEXT:    retl
 ;
 ; SSE2-X64-LABEL: extract_i8_0:
 ; SSE2-X64:       # BB#0:
-; SSE2-X64-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-X64-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-X64-NEXT:    movd %xmm0, %eax
 ; SSE2-X64-NEXT:    movb %al, (%rdi)
 ; SSE2-X64-NEXT:    retq
 ;
@@ -57,22 +49,16 @@
 define void @extract_i8_3(i8* nocapture %dst, <16 x i8> %foo) nounwind {
 ; SSE2-X32-LABEL: extract_i8_3:
 ; SSE2-X32:       # BB#0:
-; SSE2-X32-NEXT:    pushl %ebp
-; SSE2-X32-NEXT:    movl %esp, %ebp
-; SSE2-X32-NEXT:    andl $-16, %esp
-; SSE2-X32-NEXT:    subl $32, %esp
-; SSE2-X32-NEXT:    movl 8(%ebp), %eax
-; SSE2-X32-NEXT:    movaps %xmm0, (%esp)
-; SSE2-X32-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; SSE2-X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; SSE2-X32-NEXT:    movd %xmm0, %ecx
+; SSE2-X32-NEXT:    shrl $24, %ecx
 ; SSE2-X32-NEXT:    movb %cl, (%eax)
-; SSE2-X32-NEXT:    movl %ebp, %esp
-; SSE2-X32-NEXT:    popl %ebp
 ; SSE2-X32-NEXT:    retl
 ;
 ; SSE2-X64-LABEL: extract_i8_3:
 ; SSE2-X64:       # BB#0:
-; SSE2-X64-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-X64-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-X64-NEXT:    movd %xmm0, %eax
+; SSE2-X64-NEXT:    shrl $24, %eax
 ; SSE2-X64-NEXT:    movb %al, (%rdi)
 ; SSE2-X64-NEXT:    retq
 ;
@@ -105,23 +91,15 @@
 define void @extract_i8_15(i8* nocapture %dst, <16 x i8> %foo) nounwind {
 ; SSE2-X32-LABEL: extract_i8_15:
 ; SSE2-X32:       # BB#0:
-; SSE2-X32-NEXT:    pushl %ebp
-; SSE2-X32-NEXT:    movl %esp, %ebp
-; SSE2-X32-NEXT:    andl $-16, %esp
-; SSE2-X32-NEXT:    subl $32, %esp
-; SSE2-X32-NEXT:    movl 8(%ebp), %eax
-; SSE2-X32-NEXT:    movaps %xmm0, (%esp)
-; SSE2-X32-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; SSE2-X32-NEXT:    movb %cl, (%eax)
-; SSE2-X32-NEXT:    movl %ebp, %esp
-; SSE2-X32-NEXT:    popl %ebp
+; SSE2-X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; SSE2-X32-NEXT:    pextrw $7, %xmm0, %ecx
+; SSE2-X32-NEXT:    movb %ch, (%eax)
 ; SSE2-X32-NEXT:    retl
 ;
 ; SSE2-X64-LABEL: extract_i8_15:
 ; SSE2-X64:       # BB#0:
-; SSE2-X64-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-X64-NEXT:    movb -{{[0-9]+}}(%rsp), %al
-; SSE2-X64-NEXT:    movb %al, (%rdi)
+; SSE2-X64-NEXT:    pextrw $7, %xmm0, %eax
+; SSE2-X64-NEXT:    movb %ah, (%rdi) # NOREX
 ; SSE2-X64-NEXT:    retq
 ;
 ; SSE41-X32-LABEL: extract_i8_15:
Index: llvm/trunk/test/CodeGen/X86/extractelement-index.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/extractelement-index.ll
+++ llvm/trunk/test/CodeGen/X86/extractelement-index.ll
@@ -11,8 +11,9 @@
 define i8 @extractelement_v16i8_1(<16 x i8> %a) nounwind {
 ; SSE2-LABEL: extractelement_v16i8_1:
 ; SSE2:       # BB#0:
-; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    shrl $8, %eax
+; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: extractelement_v16i8_1:
@@ -33,8 +34,9 @@
 define i8 @extractelement_v16i8_11(<16 x i8> %a) nounwind {
 ; SSE2-LABEL: extractelement_v16i8_11:
 ; SSE2:       # BB#0:
-; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-NEXT:    pextrw $5, %xmm0, %eax
+; SSE2-NEXT:    shrl $8, %eax
+; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: extractelement_v16i8_11:
@@ -55,8 +57,8 @@
 define i8 @extractelement_v16i8_14(<16 x i8> %a) nounwind {
 ; SSE2-LABEL: extractelement_v16i8_14:
 ; SSE2:       # BB#0:
-; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-NEXT:    pextrw $7, %xmm0, %eax
+; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: extractelement_v16i8_14:
@@ -77,8 +79,9 @@
 define i8 @extractelement_v32i8_1(<32 x i8> %a) nounwind {
 ; SSE2-LABEL: extractelement_v32i8_1:
 ; SSE2:       # BB#0:
-; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    shrl $8, %eax
+; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: extractelement_v32i8_1:
@@ -100,8 +103,9 @@
 define i8 @extractelement_v32i8_17(<32 x i8> %a) nounwind {
 ; SSE2-LABEL: extractelement_v32i8_17:
 ; SSE2:       # BB#0:
-; SSE2-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-NEXT:    movd %xmm1, %eax
+; SSE2-NEXT:    shrl $8, %eax
+; SSE2-NEXT:    # kill: %AL<def> %AL<kill> %EAX<kill>
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: extractelement_v32i8_17: