Index: lib/CodeGen/SelectionDAG/DAGCombiner.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -8779,6 +8779,22 @@
     return DAG.getNode(N0.getOpcode(), SL, VTs, X, Y, N0.getOperand(2));
   }
 
+  // fold (truncate (extract_subvector(ext x))) ->
+  //      (extract_subvector x)
+  // TODO: This can be generalized to cover cases where the truncate and extract
+  // do not fully cancel each other out.
+  if (!LegalTypes && N0.getOpcode() == ISD::EXTRACT_SUBVECTOR) {
+    SDValue N00 = N0.getOperand(0);
+    if (N00.getOpcode() == ISD::SIGN_EXTEND ||
+        N00.getOpcode() == ISD::ZERO_EXTEND ||
+        N00.getOpcode() == ISD::ANY_EXTEND) {
+      if (N00.getOperand(0)->getValueType(0).getVectorElementType() ==
+          VT.getVectorElementType())
+        return DAG.getNode(ISD::EXTRACT_SUBVECTOR, SDLoc(N0->getOperand(0)), VT,
+                           N00.getOperand(0), N0.getOperand(1));
+    }
+  }
+
   if (SDValue NewVSel = matchVSelectOpSizesWithSetCC(N))
     return NewVSel;
 
Index: test/CodeGen/X86/madd.ll
===================================================================
--- test/CodeGen/X86/madd.ll
+++ test/CodeGen/X86/madd.ll
@@ -412,28 +412,8 @@
 define <8 x i32> @pmaddwd_16(<16 x i16> %A, <16 x i16> %B) {
 ; SSE2-LABEL: pmaddwd_16:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
-; SSE2-NEXT:    psrad $16, %xmm4
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
-; SSE2-NEXT:    psrad $16, %xmm5
-; SSE2-NEXT:    packssdw %xmm4, %xmm5
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE2-NEXT:    psrad $16, %xmm4
-; SSE2-NEXT:    packssdw %xmm1, %xmm4
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
-; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    packssdw %xmm0, %xmm1
-; SSE2-NEXT:    pmaddwd %xmm5, %xmm1
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE2-NEXT:    psrad $16, %xmm3
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    packssdw %xmm3, %xmm0
-; SSE2-NEXT:    pmaddwd %xmm4, %xmm0
+; SSE2-NEXT:    pmaddwd %xmm2, %xmm0
+; SSE2-NEXT:    pmaddwd %xmm3, %xmm1
 ; SSE2-NEXT:    retq
 ;
 ; AVX-LABEL: pmaddwd_16:
@@ -452,76 +432,16 @@
 define <16 x i32> @pmaddwd_32(<32 x i16> %A, <32 x i16> %B) {
 ; SSE2-LABEL: pmaddwd_32:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; SSE2-NEXT:    psrad $16, %xmm8
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm3[0],xmm11[1],xmm3[1],xmm11[2],xmm3[2],xmm11[3],xmm3[3]
-; SSE2-NEXT:    psrad $16, %xmm11
-; SSE2-NEXT:    packssdw %xmm8, %xmm11
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE2-NEXT:    psrad $16, %xmm3
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm2[0],xmm10[1],xmm2[1],xmm10[2],xmm2[2],xmm10[3],xmm2[3]
-; SSE2-NEXT:    psrad $16, %xmm10
-; SSE2-NEXT:    packssdw %xmm3, %xmm10
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE2-NEXT:    psrad $16, %xmm2
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
-; SSE2-NEXT:    psrad $16, %xmm9
-; SSE2-NEXT:    packssdw %xmm2, %xmm9
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; SSE2-NEXT:    psrad $16, %xmm8
-; SSE2-NEXT:    packssdw %xmm1, %xmm8
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3]
-; SSE2-NEXT:    psrad $16, %xmm3
-; SSE2-NEXT:    packssdw %xmm0, %xmm3
-; SSE2-NEXT:    pmaddwd %xmm11, %xmm3
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
-; SSE2-NEXT:    psrad $16, %xmm2
-; SSE2-NEXT:    packssdw %xmm0, %xmm2
-; SSE2-NEXT:    pmaddwd %xmm10, %xmm2
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
-; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    packssdw %xmm0, %xmm1
-; SSE2-NEXT:    pmaddwd %xmm9, %xmm1
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
-; SSE2-NEXT:    psrad $16, %xmm5
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    packssdw %xmm5, %xmm0
-; SSE2-NEXT:    pmaddwd %xmm8, %xmm0
+; SSE2-NEXT:    pmaddwd %xmm4, %xmm0
+; SSE2-NEXT:    pmaddwd %xmm5, %xmm1
+; SSE2-NEXT:    pmaddwd %xmm6, %xmm2
+; SSE2-NEXT:    pmaddwd %xmm7, %xmm3
 ; SSE2-NEXT:    retq
 ;
 ; AVX2-LABEL: pmaddwd_32:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm4
-; AVX2-NEXT:    vpmovsxwd %xmm4, %ymm4
-; AVX2-NEXT:    vpmovsxwd %xmm1, %ymm1
-; AVX2-NEXT:    vpackssdw %ymm4, %ymm1, %ymm1
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm4
-; AVX2-NEXT:    vpmovsxwd %xmm4, %ymm4
-; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm0
-; AVX2-NEXT:    vpackssdw %ymm4, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm3, %xmm4
-; AVX2-NEXT:    vpmovsxwd %xmm4, %ymm4
-; AVX2-NEXT:    vpmovsxwd %xmm3, %ymm3
-; AVX2-NEXT:    vpackssdw %ymm4, %ymm3, %ymm3
-; AVX2-NEXT:    vextracti128 $1, %ymm2, %xmm4
-; AVX2-NEXT:    vpmovsxwd %xmm4, %ymm4
-; AVX2-NEXT:    vpmovsxwd %xmm2, %ymm2
-; AVX2-NEXT:    vpackssdw %ymm4, %ymm2, %ymm2
-; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm2[0,2,1,3]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
 ; AVX2-NEXT:    vpmaddwd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm2 = ymm3[0,2,1,3]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,1,3]
-; AVX2-NEXT:    vpmaddwd %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vpmaddwd %ymm3, %ymm1, %ymm1
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: pmaddwd_32:
Index: test/CodeGen/X86/trunc-subvector.ll
===================================================================
--- test/CodeGen/X86/trunc-subvector.ll
+++ test/CodeGen/X86/trunc-subvector.ll
@@ -7,26 +7,13 @@
 define <4 x i32> @test1(<8 x i32> %v) {
 ; SSE2-LABEL: test1:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE2-NEXT:    retq
 ;
-; AVX2-LABEL: test1:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpmovsxdq %xmm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    # kill: def %xmm0 killed %xmm0 killed %ymm0
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
-;
-; AVX512-LABEL: test1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    vpmovsxdq %ymm0, %zmm0
-; AVX512-NEXT:    vpmovqd %zmm0, %ymm0
-; AVX512-NEXT:    # kill: def %xmm0 killed %xmm0 killed %ymm0
-; AVX512-NEXT:    vzeroupper
-; AVX512-NEXT:    retq
+; AVX-LABEL: test1:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def %xmm0 killed %xmm0 killed %ymm0
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
   %x = sext <8 x i32> %v to <8 x i64>
   %s = shufflevector <8 x i64> %x, <8 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %t = trunc <4 x i64> %s to <4 x i32>
@@ -36,29 +23,14 @@
 define <4 x i32> @test2(<8 x i32> %v) {
 ; SSE2-LABEL: test2:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    movaps %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; AVX2-LABEL: test2:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT:    vpmovsxdq %xmm0, %ymm0
-; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[0,2,2,3,4,6,6,7]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    # kill: def %xmm0 killed %xmm0 killed %ymm0
-; AVX2-NEXT:    vzeroupper
-; AVX2-NEXT:    retq
-;
-; AVX512-LABEL: test2:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    vpmovsxdq %ymm0, %zmm0
-; AVX512-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
-; AVX512-NEXT:    vpmovqd %zmm0, %ymm0
-; AVX512-NEXT:    # kill: def %xmm0 killed %xmm0 killed %ymm0
-; AVX512-NEXT:    vzeroupper
-; AVX512-NEXT:    retq
+; AVX-LABEL: test2:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX-NEXT:    vzeroupper
+; AVX-NEXT:    retq
   %x = sext <8 x i32> %v to <8 x i64>
   %s = shufflevector <8 x i64> %x, <8 x i64> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
   %t = trunc <4 x i64> %s to <4 x i32>