diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -5085,6 +5085,10 @@
   if (N1.isUndef())
     return N0;
 
+  // Fold (avg x, x) --> x
+  if (N0 == N1 && Level >= AfterLegalizeTypes)
+    return N0;
+
   // TODO If we use avg for scalars anywhere, we can add (avgfl x, 0) -> x >> 1
 
   return SDValue();
diff --git a/llvm/test/CodeGen/AArch64/arm64-vhadd.ll b/llvm/test/CodeGen/AArch64/arm64-vhadd.ll
--- a/llvm/test/CodeGen/AArch64/arm64-vhadd.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-vhadd.ll
@@ -1194,7 +1194,6 @@
 define <8 x i8> @shadd_v8i8(<8 x i8> %x) {
 ; CHECK-LABEL: shadd_v8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    shadd.8b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i8> @llvm.aarch64.neon.shadd.v8i8(<8 x i8> %x, <8 x i8> %x)
   ret <8 x i8> %r
@@ -1203,7 +1202,6 @@
 define <4 x i16> @shadd_v4i16(<4 x i16> %x) {
 ; CHECK-LABEL: shadd_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    shadd.4h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i16> @llvm.aarch64.neon.shadd.v4i16(<4 x i16> %x, <4 x i16> %x)
   ret <4 x i16> %r
@@ -1212,7 +1210,6 @@
 define <2 x i32> @shadd_v2i32(<2 x i32> %x) {
 ; CHECK-LABEL: shadd_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    shadd.2s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <2 x i32> @llvm.aarch64.neon.shadd.v2i32(<2 x i32> %x, <2 x i32> %x)
   ret <2 x i32> %r
@@ -1221,7 +1218,6 @@
 define <16 x i8> @shadd_v16i8(<16 x i8> %x) {
 ; CHECK-LABEL: shadd_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    shadd.16b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <16 x i8> @llvm.aarch64.neon.shadd.v16i8(<16 x i8> %x, <16 x i8> %x)
   ret <16 x i8> %r
@@ -1230,7 +1226,6 @@
 define <8 x i16> @shadd_v8i16(<8 x i16> %x) {
 ; CHECK-LABEL: shadd_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    shadd.8h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i16> @llvm.aarch64.neon.shadd.v8i16(<8 x i16> %x, <8 x i16> %x)
   ret <8 x i16> %r
@@ -1239,7 +1234,6 @@
 define <4 x i32> @shadd_v4i32(<4 x i32> %x) {
 ; CHECK-LABEL: shadd_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    shadd.4s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i32> @llvm.aarch64.neon.shadd.v4i32(<4 x i32> %x, <4 x i32> %x)
   ret <4 x i32> %r
@@ -1248,7 +1242,6 @@
 define <8 x i8> @uhadd_v8i8(<8 x i8> %x) {
 ; CHECK-LABEL: uhadd_v8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uhadd.8b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i8> @llvm.aarch64.neon.uhadd.v8i8(<8 x i8> %x, <8 x i8> %x)
   ret <8 x i8> %r
@@ -1257,7 +1250,6 @@
 define <4 x i16> @uhadd_v4i16(<4 x i16> %x) {
 ; CHECK-LABEL: uhadd_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uhadd.4h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i16> @llvm.aarch64.neon.uhadd.v4i16(<4 x i16> %x, <4 x i16> %x)
   ret <4 x i16> %r
@@ -1266,7 +1258,6 @@
 define <2 x i32> @uhadd_v2i32(<2 x i32> %x) {
 ; CHECK-LABEL: uhadd_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uhadd.2s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <2 x i32> @llvm.aarch64.neon.uhadd.v2i32(<2 x i32> %x, <2 x i32> %x)
   ret <2 x i32> %r
@@ -1275,7 +1266,6 @@
 define <16 x i8> @uhadd_v16i8(<16 x i8> %x) {
 ; CHECK-LABEL: uhadd_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uhadd.16b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <16 x i8> @llvm.aarch64.neon.uhadd.v16i8(<16 x i8> %x, <16 x i8> %x)
   ret <16 x i8> %r
@@ -1284,7 +1274,6 @@
 define <8 x i16> @uhadd_v8i16(<8 x i16> %x) {
 ; CHECK-LABEL: uhadd_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uhadd.8h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i16> @llvm.aarch64.neon.uhadd.v8i16(<8 x i16> %x, <8 x i16> %x)
   ret <8 x i16> %r
@@ -1293,7 +1282,6 @@
 define <4 x i32> @uhadd_v4i32(<4 x i32> %x) {
 ; CHECK-LABEL: uhadd_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    uhadd.4s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i32> @llvm.aarch64.neon.uhadd.v4i32(<4 x i32> %x, <4 x i32> %x)
   ret <4 x i32> %r
@@ -1301,7 +1289,6 @@
 define <8 x i8> @srhadd_v8i8(<8 x i8> %x) {
 ; CHECK-LABEL: srhadd_v8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    srhadd.8b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i8> @llvm.aarch64.neon.srhadd.v8i8(<8 x i8> %x, <8 x i8> %x)
   ret <8 x i8> %r
@@ -1310,7 +1297,6 @@
 define <4 x i16> @srhadd_v4i16(<4 x i16> %x) {
 ; CHECK-LABEL: srhadd_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    srhadd.4h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i16> @llvm.aarch64.neon.srhadd.v4i16(<4 x i16> %x, <4 x i16> %x)
   ret <4 x i16> %r
@@ -1319,7 +1305,6 @@
 define <2 x i32> @srhadd_v2i32(<2 x i32> %x) {
 ; CHECK-LABEL: srhadd_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    srhadd.2s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <2 x i32> @llvm.aarch64.neon.srhadd.v2i32(<2 x i32> %x, <2 x i32> %x)
   ret <2 x i32> %r
@@ -1328,7 +1313,6 @@
 define <16 x i8> @srhadd_v16i8(<16 x i8> %x) {
 ; CHECK-LABEL: srhadd_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    srhadd.16b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <16 x i8> @llvm.aarch64.neon.srhadd.v16i8(<16 x i8> %x, <16 x i8> %x)
   ret <16 x i8> %r
@@ -1337,7 +1321,6 @@
 define <8 x i16> @srhadd_v8i16(<8 x i16> %x) {
 ; CHECK-LABEL: srhadd_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    srhadd.8h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i16> @llvm.aarch64.neon.srhadd.v8i16(<8 x i16> %x, <8 x i16> %x)
   ret <8 x i16> %r
@@ -1346,7 +1329,6 @@
 define <4 x i32> @srhadd_v4i32(<4 x i32> %x) {
 ; CHECK-LABEL: srhadd_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    srhadd.4s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i32> @llvm.aarch64.neon.srhadd.v4i32(<4 x i32> %x, <4 x i32> %x)
   ret <4 x i32> %r
@@ -1355,7 +1337,6 @@
 define <8 x i8> @urhadd_v8i8(<8 x i8> %x) {
 ; CHECK-LABEL: urhadd_v8i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    urhadd.8b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i8> @llvm.aarch64.neon.urhadd.v8i8(<8 x i8> %x, <8 x i8> %x)
   ret <8 x i8> %r
@@ -1364,7 +1345,6 @@
 define <4 x i16> @urhadd_v4i16(<4 x i16> %x) {
 ; CHECK-LABEL: urhadd_v4i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    urhadd.4h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i16> @llvm.aarch64.neon.urhadd.v4i16(<4 x i16> %x, <4 x i16> %x)
   ret <4 x i16> %r
@@ -1373,7 +1353,6 @@
 define <2 x i32> @urhadd_v2i32(<2 x i32> %x) {
 ; CHECK-LABEL: urhadd_v2i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    urhadd.2s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <2 x i32> @llvm.aarch64.neon.urhadd.v2i32(<2 x i32> %x, <2 x i32> %x)
   ret <2 x i32> %r
@@ -1382,7 +1361,6 @@
 define <16 x i8> @urhadd_v16i8(<16 x i8> %x) {
 ; CHECK-LABEL: urhadd_v16i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    urhadd.16b v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <16 x i8> @llvm.aarch64.neon.urhadd.v16i8(<16 x i8> %x, <16 x i8> %x)
   ret <16 x i8> %r
@@ -1391,7 +1369,6 @@
 define <8 x i16> @urhadd_v8i16(<8 x i16> %x) {
 ; CHECK-LABEL: urhadd_v8i16:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    urhadd.8h v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <8 x i16> @llvm.aarch64.neon.urhadd.v8i16(<8 x i16> %x, <8 x i16> %x)
   ret <8 x i16> %r
@@ -1400,7 +1377,6 @@
 define <4 x i32> @urhadd_v4i32(<4 x i32> %x) {
 ; CHECK-LABEL: urhadd_v4i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    urhadd.4s v0, v0, v0
 ; CHECK-NEXT:    ret
   %r = tail call <4 x i32> @llvm.aarch64.neon.urhadd.v4i32(<4 x i32> %x, <4 x i32> %x)
   ret <4 x i32> %r
diff --git a/llvm/test/CodeGen/X86/avg.ll b/llvm/test/CodeGen/X86/avg.ll
--- a/llvm/test/CodeGen/X86/avg.ll
+++ b/llvm/test/CodeGen/X86/avg.ll
@@ -728,65 +728,40 @@
 define void @avg_v64i8_2(ptr %a, ptr %b) nounwind {
 ; SSE2-LABEL: avg_v64i8_2:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa (%rsi), %xmm0
-; SSE2-NEXT:    movdqa 16(%rsi), %xmm1
-; SSE2-NEXT:    movdqa 32(%rsi), %xmm2
-; SSE2-NEXT:    movdqa 48(%rsi), %xmm3
-; SSE2-NEXT:    pavgb %xmm0, %xmm0
-; SSE2-NEXT:    pavgb %xmm1, %xmm1
-; SSE2-NEXT:    pavgb %xmm2, %xmm2
-; SSE2-NEXT:    pavgb %xmm3, %xmm3
-; SSE2-NEXT:    movdqu %xmm3, (%rax)
-; SSE2-NEXT:    movdqu %xmm2, (%rax)
-; SSE2-NEXT:    movdqu %xmm1, (%rax)
-; SSE2-NEXT:    movdqu %xmm0, (%rax)
+; SSE2-NEXT:    movaps (%rsi), %xmm0
+; SSE2-NEXT:    movaps 16(%rsi), %xmm1
+; SSE2-NEXT:    movaps 32(%rsi), %xmm2
+; SSE2-NEXT:    movaps 48(%rsi), %xmm3
+; SSE2-NEXT:    movups %xmm3, (%rax)
+; SSE2-NEXT:    movups %xmm2, (%rax)
+; SSE2-NEXT:    movups %xmm1, (%rax)
+; SSE2-NEXT:    movups %xmm0, (%rax)
 ; SSE2-NEXT:    retq
 ;
 ; AVX1-LABEL: avg_v64i8_2:
 ; AVX1:       # %bb.0:
-; AVX1-NEXT:    vmovdqa (%rsi), %xmm0
-; AVX1-NEXT:    vmovdqa 16(%rsi), %xmm1
-; AVX1-NEXT:    vmovdqa 32(%rsi), %xmm2
-; AVX1-NEXT:    vmovdqa 48(%rsi), %xmm3
-; AVX1-NEXT:    vpavgb %xmm0, %xmm0, %xmm0
-; AVX1-NEXT:    vpavgb %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpavgb %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpavgb %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vmovdqu %xmm3, (%rax)
-; AVX1-NEXT:    vmovdqu %xmm2, (%rax)
-; AVX1-NEXT:    vmovdqu %xmm1, (%rax)
-; AVX1-NEXT:    vmovdqu %xmm0, (%rax)
+; AVX1-NEXT:    vmovaps (%rsi), %ymm0
+; AVX1-NEXT:    vmovaps 32(%rsi), %ymm1
+; AVX1-NEXT:    vmovups %ymm1, (%rax)
+; AVX1-NEXT:    vmovups %ymm0, (%rax)
+; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: avg_v64i8_2:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vmovdqa (%rsi), %ymm0
-; AVX2-NEXT:    vmovdqa 32(%rsi), %ymm1
-; AVX2-NEXT:    vpavgb %ymm0, %ymm0, %ymm0
-; AVX2-NEXT:    vpavgb %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vmovdqu %ymm1, (%rax)
-; AVX2-NEXT:    vmovdqu %ymm0, (%rax)
+; AVX2-NEXT:    vmovaps (%rsi), %ymm0
+; AVX2-NEXT:    vmovaps 32(%rsi), %ymm1
+; AVX2-NEXT:    vmovups %ymm1, (%rax)
+; AVX2-NEXT:    vmovups %ymm0, (%rax)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
-; AVX512F-LABEL: avg_v64i8_2:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vmovdqa (%rsi), %ymm0
-; AVX512F-NEXT:    vmovdqa 32(%rsi), %ymm1
-; AVX512F-NEXT:    vpavgb %ymm0, %ymm0, %ymm0
-; AVX512F-NEXT:    vpavgb %ymm1, %ymm1, %ymm1
-; AVX512F-NEXT:    vmovdqu %ymm1, (%rax)
-; AVX512F-NEXT:    vmovdqu %ymm0, (%rax)
-; AVX512F-NEXT:    vzeroupper
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: avg_v64i8_2:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vmovdqa64 (%rsi), %zmm0
-; AVX512BW-NEXT:    vpavgb %zmm0, %zmm0, %zmm0
-; AVX512BW-NEXT:    vmovdqu64 %zmm0, (%rax)
-; AVX512BW-NEXT:    vzeroupper
-; AVX512BW-NEXT:    retq
+; AVX512-LABEL: avg_v64i8_2:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovaps (%rsi), %zmm0
+; AVX512-NEXT:    vmovups %zmm0, (%rax)
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
   %1 = load <64 x i8>, ptr %a
   %2 = load <64 x i8>, ptr %b
   %3 = zext <64 x i8> %1 to <64 x i32>
diff --git a/llvm/test/CodeGen/X86/combine-pavg.ll b/llvm/test/CodeGen/X86/combine-pavg.ll
--- a/llvm/test/CodeGen/X86/combine-pavg.ll
+++ b/llvm/test/CodeGen/X86/combine-pavg.ll
@@ -6,16 +6,13 @@
 declare <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8>, <16 x i8>) nounwind readnone
 declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>) nounwind readnone
 
-; TODO: AVG(X,X) -> X
 define <16 x i8> @combine_pavgb_self(<16 x i8> %a0) {
 ; SSE-LABEL: combine_pavgb_self:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    pavgb %xmm0, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: combine_pavgb_self:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    vpavgb %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %1 = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %a0, <16 x i8> %a0)
   ret <16 x i8> %1