diff --git a/llvm/test/CodeGen/X86/haddsub-shuf.ll b/llvm/test/CodeGen/X86/haddsub-shuf.ll
index 429175a10818b3..5f15d07d0bab39 100644
--- a/llvm/test/CodeGen/X86/haddsub-shuf.ll
+++ b/llvm/test/CodeGen/X86/haddsub-shuf.ll
@@ -1,6 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3           | FileCheck %s --check-prefixes=SSSE3,SSSE3_SLOW
-; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3,fast-hops | FileCheck %s --check-prefixes=SSSE3,SSSE3_FAST
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse3           | FileCheck %s --check-prefixes=SSE,SSE_SLOW,SSE3,SSE3_SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse3,fast-hops | FileCheck %s --check-prefixes=SSE,SSE_FAST,SSE3,SSE3_FAST
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3           | FileCheck %s --check-prefixes=SSE,SSE_SLOW,SSSE3,SSSE3_SLOW
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3,fast-hops | FileCheck %s --check-prefixes=SSE,SSE_FAST,SSSE3,SSSE3_FAST
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx             | FileCheck %s --check-prefixes=AVX,AVX1,AVX1_SLOW
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx,fast-hops   | FileCheck %s --check-prefixes=AVX,AVX1,AVX1_FAST
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2            | FileCheck %s --check-prefixes=AVX,AVX2,AVX2_SLOW
@@ -10,10 +12,10 @@
 ; PR34111 - https://bugs.llvm.org/show_bug.cgi?id=34111
 
 define <4 x float> @hadd_v4f32(<4 x float> %a) {
-; SSSE3-LABEL: hadd_v4f32:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm0, %xmm0
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hadd_v4f32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm0, %xmm0
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: hadd_v4f32:
 ; AVX:       # %bb.0:
@@ -27,21 +29,21 @@ define <4 x float> @hadd_v4f32(<4 x float> %a) {
 }
 
 define <8 x float> @hadd_v8f32a(<8 x float> %a) {
-; SSSE3_SLOW-LABEL: hadd_v8f32a:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movaps %xmm0, %xmm2
-; SSSE3_SLOW-NEXT:    haddps %xmm1, %xmm2
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
-; SSSE3_SLOW-NEXT:    movaps %xmm2, %xmm1
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hadd_v8f32a:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    movaps %xmm0, %xmm2
-; SSSE3_FAST-NEXT:    haddps %xmm1, %xmm2
-; SSSE3_FAST-NEXT:    haddps %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    movaps %xmm2, %xmm1
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hadd_v8f32a:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movaps %xmm0, %xmm2
+; SSE_SLOW-NEXT:    haddps %xmm1, %xmm2
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
+; SSE_SLOW-NEXT:    movaps %xmm2, %xmm1
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hadd_v8f32a:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    movaps %xmm0, %xmm2
+; SSE_FAST-NEXT:    haddps %xmm1, %xmm2
+; SSE_FAST-NEXT:    haddps %xmm0, %xmm0
+; SSE_FAST-NEXT:    movaps %xmm2, %xmm1
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX1_SLOW-LABEL: hadd_v8f32a:
 ; AVX1_SLOW:       # %bb.0:
@@ -71,11 +73,11 @@ define <8 x float> @hadd_v8f32a(<8 x float> %a) {
 }
 
 define <8 x float> @hadd_v8f32b(<8 x float> %a) {
-; SSSE3-LABEL: hadd_v8f32b:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm0, %xmm0
-; SSSE3-NEXT:    haddps %xmm1, %xmm1
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hadd_v8f32b:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm0, %xmm0
+; SSE-NEXT:    haddps %xmm1, %xmm1
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: hadd_v8f32b:
 ; AVX:       # %bb.0:
@@ -89,10 +91,10 @@ define <8 x float> @hadd_v8f32b(<8 x float> %a) {
 }
 
 define <4 x float> @hsub_v4f32(<4 x float> %a) {
-; SSSE3-LABEL: hsub_v4f32:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    hsubps %xmm0, %xmm0
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hsub_v4f32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    hsubps %xmm0, %xmm0
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: hsub_v4f32:
 ; AVX:       # %bb.0:
@@ -106,21 +108,21 @@ define <4 x float> @hsub_v4f32(<4 x float> %a) {
 }
 
 define <8 x float> @hsub_v8f32a(<8 x float> %a) {
-; SSSE3_SLOW-LABEL: hsub_v8f32a:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movaps %xmm0, %xmm2
-; SSSE3_SLOW-NEXT:    hsubps %xmm1, %xmm2
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
-; SSSE3_SLOW-NEXT:    movaps %xmm2, %xmm1
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hsub_v8f32a:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    movaps %xmm0, %xmm2
-; SSSE3_FAST-NEXT:    hsubps %xmm1, %xmm2
-; SSSE3_FAST-NEXT:    hsubps %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    movaps %xmm2, %xmm1
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hsub_v8f32a:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movaps %xmm0, %xmm2
+; SSE_SLOW-NEXT:    hsubps %xmm1, %xmm2
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
+; SSE_SLOW-NEXT:    movaps %xmm2, %xmm1
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hsub_v8f32a:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    movaps %xmm0, %xmm2
+; SSE_FAST-NEXT:    hsubps %xmm1, %xmm2
+; SSE_FAST-NEXT:    hsubps %xmm0, %xmm0
+; SSE_FAST-NEXT:    movaps %xmm2, %xmm1
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX1_SLOW-LABEL: hsub_v8f32a:
 ; AVX1_SLOW:       # %bb.0:
@@ -150,11 +152,11 @@ define <8 x float> @hsub_v8f32a(<8 x float> %a) {
 }
 
 define <8 x float> @hsub_v8f32b(<8 x float> %a) {
-; SSSE3-LABEL: hsub_v8f32b:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    hsubps %xmm0, %xmm0
-; SSSE3-NEXT:    hsubps %xmm1, %xmm1
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hsub_v8f32b:
+; SSE:       # %bb.0:
+; SSE-NEXT:    hsubps %xmm0, %xmm0
+; SSE-NEXT:    hsubps %xmm1, %xmm1
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: hsub_v8f32b:
 ; AVX:       # %bb.0:
@@ -168,18 +170,18 @@ define <8 x float> @hsub_v8f32b(<8 x float> %a) {
 }
 
 define <2 x double> @hadd_v2f64(<2 x double> %a) {
-; SSSE3_SLOW-LABEL: hadd_v2f64:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSSE3_SLOW-NEXT:    addsd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm1[0,0]
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hadd_v2f64:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    haddpd %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hadd_v2f64:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE_SLOW-NEXT:    addsd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm1[0,0]
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hadd_v2f64:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    haddpd %xmm0, %xmm0
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX1_SLOW-LABEL: hadd_v2f64:
 ; AVX1_SLOW:       # %bb.0:
@@ -212,18 +214,18 @@ define <2 x double> @hadd_v2f64(<2 x double> %a) {
 }
 
 define <2 x double> @hadd_v2f64_scalar_splat(<2 x double> %a) {
-; SSSE3_SLOW-LABEL: hadd_v2f64_scalar_splat:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSSE3_SLOW-NEXT:    addsd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm1[0,0]
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hadd_v2f64_scalar_splat:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    haddpd %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hadd_v2f64_scalar_splat:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE_SLOW-NEXT:    addsd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm1[0,0]
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hadd_v2f64_scalar_splat:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    haddpd %xmm0, %xmm0
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX1_SLOW-LABEL: hadd_v2f64_scalar_splat:
 ; AVX1_SLOW:       # %bb.0:
@@ -257,23 +259,23 @@ define <2 x double> @hadd_v2f64_scalar_splat(<2 x double> %a) {
 }
 
 define <4 x double> @hadd_v4f64_scalar_splat(<4 x double> %a) {
-; SSSE3_SLOW-LABEL: hadd_v4f64_scalar_splat:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm2
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
-; SSSE3_SLOW-NEXT:    addsd %xmm0, %xmm2
-; SSSE3_SLOW-NEXT:    movapd %xmm1, %xmm3
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
-; SSSE3_SLOW-NEXT:    addsd %xmm1, %xmm3
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm1 = xmm3[0,0]
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hadd_v4f64_scalar_splat:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    haddpd %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    haddpd %xmm1, %xmm1
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hadd_v4f64_scalar_splat:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm2
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE_SLOW-NEXT:    addsd %xmm0, %xmm2
+; SSE_SLOW-NEXT:    movapd %xmm1, %xmm3
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1]
+; SSE_SLOW-NEXT:    addsd %xmm1, %xmm3
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm1 = xmm3[0,0]
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hadd_v4f64_scalar_splat:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    haddpd %xmm0, %xmm0
+; SSE_FAST-NEXT:    haddpd %xmm1, %xmm1
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX-LABEL: hadd_v4f64_scalar_splat:
 ; AVX:       # %bb.0:
@@ -292,20 +294,20 @@ define <4 x double> @hadd_v4f64_scalar_splat(<4 x double> %a) {
 }
 
 define <4 x double> @hadd_v4f64_scalar_broadcast(<4 x double> %a) {
-; SSSE3_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSSE3_SLOW-NEXT:    addsd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm1[0,0]
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hadd_v4f64_scalar_broadcast:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    haddpd %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    movapd %xmm0, %xmm1
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE_SLOW-NEXT:    addsd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm1[0,0]
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hadd_v4f64_scalar_broadcast:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    haddpd %xmm0, %xmm0
+; SSE_FAST-NEXT:    movapd %xmm0, %xmm1
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX1_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
 ; AVX1_SLOW:       # %bb.0:
@@ -346,23 +348,23 @@ define <4 x double> @hadd_v4f64_scalar_broadcast(<4 x double> %a) {
 }
 
 define <4 x double> @hadd_v4f64(<4 x double> %a) {
-; SSSE3_SLOW-LABEL: hadd_v4f64:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm2
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
-; SSSE3_SLOW-NEXT:    addsd %xmm0, %xmm2
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
-; SSSE3_SLOW-NEXT:    movapd %xmm1, %xmm2
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm1[1]
-; SSSE3_SLOW-NEXT:    addsd %xmm1, %xmm2
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm1 = xmm2[0,0]
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hadd_v4f64:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    haddpd %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    haddpd %xmm1, %xmm1
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hadd_v4f64:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm2
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE_SLOW-NEXT:    addsd %xmm0, %xmm2
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm2[0,0]
+; SSE_SLOW-NEXT:    movapd %xmm1, %xmm2
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm1[1]
+; SSE_SLOW-NEXT:    addsd %xmm1, %xmm2
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm1 = xmm2[0,0]
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hadd_v4f64:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    haddpd %xmm0, %xmm0
+; SSE_FAST-NEXT:    haddpd %xmm1, %xmm1
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX-LABEL: hadd_v4f64:
 ; AVX:       # %bb.0:
@@ -376,18 +378,18 @@ define <4 x double> @hadd_v4f64(<4 x double> %a) {
 }
 
 define <2 x double> @hsub_v2f64(<2 x double> %a) {
-; SSSE3_SLOW-LABEL: hsub_v2f64:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm1
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSSE3_SLOW-NEXT:    subsd %xmm1, %xmm0
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm0[0,0]
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hsub_v2f64:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    hsubpd %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hsub_v2f64:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm1
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE_SLOW-NEXT:    subsd %xmm1, %xmm0
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm0[0,0]
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hsub_v2f64:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    hsubpd %xmm0, %xmm0
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX1_SLOW-LABEL: hsub_v2f64:
 ; AVX1_SLOW:       # %bb.0:
@@ -420,23 +422,23 @@ define <2 x double> @hsub_v2f64(<2 x double> %a) {
 }
 
 define <4 x double> @hsub_v4f64(<4 x double> %a) {
-; SSSE3_SLOW-LABEL: hsub_v4f64:
-; SSSE3_SLOW:       # %bb.0:
-; SSSE3_SLOW-NEXT:    movapd %xmm0, %xmm2
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
-; SSSE3_SLOW-NEXT:    subsd %xmm2, %xmm0
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm0[0,0]
-; SSSE3_SLOW-NEXT:    movapd %xmm1, %xmm2
-; SSSE3_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm1[1]
-; SSSE3_SLOW-NEXT:    subsd %xmm2, %xmm1
-; SSSE3_SLOW-NEXT:    movddup {{.*#+}} xmm1 = xmm1[0,0]
-; SSSE3_SLOW-NEXT:    retq
-;
-; SSSE3_FAST-LABEL: hsub_v4f64:
-; SSSE3_FAST:       # %bb.0:
-; SSSE3_FAST-NEXT:    hsubpd %xmm0, %xmm0
-; SSSE3_FAST-NEXT:    hsubpd %xmm1, %xmm1
-; SSSE3_FAST-NEXT:    retq
+; SSE_SLOW-LABEL: hsub_v4f64:
+; SSE_SLOW:       # %bb.0:
+; SSE_SLOW-NEXT:    movapd %xmm0, %xmm2
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE_SLOW-NEXT:    subsd %xmm2, %xmm0
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm0 = xmm0[0,0]
+; SSE_SLOW-NEXT:    movapd %xmm1, %xmm2
+; SSE_SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm1[1]
+; SSE_SLOW-NEXT:    subsd %xmm2, %xmm1
+; SSE_SLOW-NEXT:    movddup {{.*#+}} xmm1 = xmm1[0,0]
+; SSE_SLOW-NEXT:    retq
+;
+; SSE_FAST-LABEL: hsub_v4f64:
+; SSE_FAST:       # %bb.0:
+; SSE_FAST-NEXT:    hsubpd %xmm0, %xmm0
+; SSE_FAST-NEXT:    hsubpd %xmm1, %xmm1
+; SSE_FAST-NEXT:    retq
 ;
 ; AVX-LABEL: hsub_v4f64:
 ; AVX:       # %bb.0:
@@ -450,6 +452,13 @@ define <4 x double> @hsub_v4f64(<4 x double> %a) {
 }
 
 define <4 x i32> @hadd_v4i32(<4 x i32> %a) {
+; SSE3-LABEL: hadd_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2]
+; SSE3-NEXT:    paddd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hadd_v4i32:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddd %xmm0, %xmm0
@@ -467,6 +476,16 @@ define <4 x i32> @hadd_v4i32(<4 x i32> %a) {
 }
 
 define <8 x i32> @hadd_v8i32a(<8 x i32> %a) {
+; SSE3-LABEL: hadd_v8i32a:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movaps %xmm0, %xmm2
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,3],xmm1[1,3]
+; SSE3-NEXT:    paddd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,0,1]
+; SSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3_SLOW-LABEL: hadd_v8i32a:
 ; SSSE3_SLOW:       # %bb.0:
 ; SSSE3_SLOW-NEXT:    movdqa %xmm0, %xmm2
@@ -513,6 +532,16 @@ define <8 x i32> @hadd_v8i32a(<8 x i32> %a) {
 }
 
 define <8 x i32> @hadd_v8i32b(<8 x i32> %a) {
+; SSE3-LABEL: hadd_v8i32b:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,3,1,3]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,0,2]
+; SSE3-NEXT:    paddd %xmm2, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,3,1,3]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,0,2]
+; SSE3-NEXT:    paddd %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hadd_v8i32b:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddd %xmm0, %xmm0
@@ -539,6 +568,13 @@ define <8 x i32> @hadd_v8i32b(<8 x i32> %a) {
 }
 
 define <4 x i32> @hsub_v4i32(<4 x i32> %a) {
+; SSE3-LABEL: hsub_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,3,1,3]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,0,3]
+; SSE3-NEXT:    psubd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hsub_v4i32:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phsubd %xmm0, %xmm0
@@ -556,6 +592,16 @@ define <4 x i32> @hsub_v4i32(<4 x i32> %a) {
 }
 
 define <8 x i32> @hsub_v8i32a(<8 x i32> %a) {
+; SSE3-LABEL: hsub_v8i32a:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movaps %xmm0, %xmm2
+; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; SSE3-NEXT:    psubd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,0,1]
+; SSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3_SLOW-LABEL: hsub_v8i32a:
 ; SSSE3_SLOW:       # %bb.0:
 ; SSSE3_SLOW-NEXT:    movdqa %xmm0, %xmm2
@@ -602,6 +648,16 @@ define <8 x i32> @hsub_v8i32a(<8 x i32> %a) {
 }
 
 define <8 x i32> @hsub_v8i32b(<8 x i32> %a) {
+; SSE3-LABEL: hsub_v8i32b:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,3,1,3]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,0,2]
+; SSE3-NEXT:    psubd %xmm2, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,3,1,3]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,0,2]
+; SSE3-NEXT:    psubd %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hsub_v8i32b:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phsubd %xmm0, %xmm0
@@ -628,6 +684,18 @@ define <8 x i32> @hsub_v8i32b(<8 x i32> %a) {
 }
 
 define <8 x i16> @hadd_v8i16(<8 x i16> %a) {
+; SSE3-LABEL: hadd_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; SSE3-NEXT:    paddw %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hadd_v8i16:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddw %xmm0, %xmm0
@@ -645,6 +713,28 @@ define <8 x i16> @hadd_v8i16(<8 x i16> %a) {
 }
 
 define <16 x i16> @hadd_v16i16a(<16 x i16> %a) {
+; SSE3-LABEL: hadd_v16i16a:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE3-NEXT:    paddw %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,1,0,1]
+; SSE3-NEXT:    retq
+;
 ; SSSE3_SLOW-LABEL: hadd_v16i16a:
 ; SSSE3_SLOW:       # %bb.0:
 ; SSSE3_SLOW-NEXT:    movdqa %xmm0, %xmm2
@@ -691,6 +781,32 @@ define <16 x i16> @hadd_v16i16a(<16 x i16> %a) {
 }
 
 define <16 x i16> @hadd_v16i16b(<16 x i16> %a) {
+; SSE3-LABEL: hadd_v16i16b:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[3,1,1,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,5,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,7,5,4]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,4]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,2,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,4,5]
+; SSE3-NEXT:    paddw %xmm2, %xmm0
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[3,1,1,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,5,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,7,5,4]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,0,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,4]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,2,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,4,5]
+; SSE3-NEXT:    paddw %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hadd_v16i16b:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddw %xmm0, %xmm0
@@ -717,6 +833,14 @@ define <16 x i16> @hadd_v16i16b(<16 x i16> %a) {
 }
 
 define <8 x i16> @hsub_v8i16(<8 x i16> %a) {
+; SSE3-LABEL: hsub_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[1,1,3,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,6,6]
+; SSE3-NEXT:    psubw %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hsub_v8i16:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phsubw %xmm0, %xmm0
@@ -734,6 +858,29 @@ define <8 x i16> @hsub_v8i16(<8 x i16> %a) {
 }
 
 define <16 x i16> @hsub_v16i16a(<16 x i16> %a) {
+; SSE3-LABEL: hsub_v16i16a:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE3-NEXT:    psubw %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,0,1]
+; SSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3_SLOW-LABEL: hsub_v16i16a:
 ; SSSE3_SLOW:       # %bb.0:
 ; SSSE3_SLOW-NEXT:    movdqa %xmm0, %xmm2
@@ -780,6 +927,32 @@ define <16 x i16> @hsub_v16i16a(<16 x i16> %a) {
 }
 
 define <16 x i16> @hsub_v16i16b(<16 x i16> %a) {
+; SSE3-LABEL: hsub_v16i16b:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[3,1,1,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,5,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,7,5,4]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,0,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,4]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,2,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,4,5]
+; SSE3-NEXT:    psubw %xmm2, %xmm0
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[3,1,1,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,5,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,7,5,4]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,0,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,4]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,2,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,4,5]
+; SSE3-NEXT:    psubw %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hsub_v16i16b:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phsubw %xmm0, %xmm0
@@ -806,11 +979,11 @@ define <16 x i16> @hsub_v16i16b(<16 x i16> %a) {
 }
 
 define <4 x float> @broadcast_haddps_v4f32(<4 x float> %a0) {
-; SSSE3-LABEL: broadcast_haddps_v4f32:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm0, %xmm0
-; SSSE3-NEXT:    movsldup {{.*#+}} xmm0 = xmm0[0,0,2,2]
-; SSSE3-NEXT:    retq
+; SSE-LABEL: broadcast_haddps_v4f32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm0, %xmm0
+; SSE-NEXT:    movsldup {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: broadcast_haddps_v4f32:
 ; AVX1:       # %bb.0:
@@ -831,10 +1004,10 @@ define <4 x float> @broadcast_haddps_v4f32(<4 x float> %a0) {
 declare <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float>, <4 x float>)
 
 define <4 x float> @PR34724_1(<4 x float> %a, <4 x float> %b) {
-; SSSE3-LABEL: PR34724_1:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm1, %xmm0
-; SSSE3-NEXT:    retq
+; SSE-LABEL: PR34724_1:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm1, %xmm0
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: PR34724_1:
 ; AVX:       # %bb.0:
@@ -851,10 +1024,10 @@ define <4 x float> @PR34724_1(<4 x float> %a, <4 x float> %b) {
 }
 
 define <4 x float> @PR34724_2(<4 x float> %a, <4 x float> %b) {
-; SSSE3-LABEL: PR34724_2:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm1, %xmm0
-; SSSE3-NEXT:    retq
+; SSE-LABEL: PR34724_2:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm1, %xmm0
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: PR34724_2:
 ; AVX:       # %bb.0:
@@ -876,11 +1049,11 @@ define <4 x float> @PR34724_2(<4 x float> %a, <4 x float> %b) {
 ;
 
 define <4 x float> @hadd_4f32_v8f32_shuffle(<8 x float> %a0) {
-; SSSE3-LABEL: hadd_4f32_v8f32_shuffle:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm1, %xmm0
-; SSSE3-NEXT:    movshdup {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hadd_4f32_v8f32_shuffle:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm1, %xmm0
+; SSE-NEXT:    movshdup {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: hadd_4f32_v8f32_shuffle:
 ; AVX:       # %bb.0:
@@ -899,11 +1072,11 @@ define <4 x float> @hadd_4f32_v8f32_shuffle(<8 x float> %a0) {
 }
 
 define <4 x float> @hsub_4f32_v8f32_shuffle(<8 x float> %a0) {
-; SSSE3-LABEL: hsub_4f32_v8f32_shuffle:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm1, %xmm0
-; SSSE3-NEXT:    movshdup {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hsub_4f32_v8f32_shuffle:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm1, %xmm0
+; SSE-NEXT:    movshdup {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: hsub_4f32_v8f32_shuffle:
 ; AVX:       # %bb.0:
@@ -922,6 +1095,14 @@ define <4 x float> @hsub_4f32_v8f32_shuffle(<8 x float> %a0) {
 }
 
 define <4 x i32> @hadd_4i32_v8i32_shuffle(<8 x i32> %a0) {
+; SSE3-LABEL: hadd_4i32_v8i32_shuffle:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movaps %xmm0, %xmm2
+; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,2],xmm1[2,2]
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE3-NEXT:    paddd %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hadd_4i32_v8i32_shuffle:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddd %xmm1, %xmm0
@@ -953,6 +1134,14 @@ define <4 x i32> @hadd_4i32_v8i32_shuffle(<8 x i32> %a0) {
 }
 
 define <4 x i32> @hsub_4i32_v8i32_shuffle(<8 x i32> %a0) {
+; SSE3-LABEL: hsub_4i32_v8i32_shuffle:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movaps %xmm0, %xmm2
+; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,2],xmm1[2,2]
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE3-NEXT:    paddd %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hsub_4i32_v8i32_shuffle:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddd %xmm1, %xmm0
@@ -988,12 +1177,12 @@ define <4 x i32> @hsub_4i32_v8i32_shuffle(<8 x i32> %a0) {
 ;
 
 define <4 x double> @hadd_4f64_v4f64_shuffle(<4 x double> %a0, <4 x double> %a1) {
-; SSSE3-LABEL: hadd_4f64_v4f64_shuffle:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddpd %xmm1, %xmm0
-; SSSE3-NEXT:    haddpd %xmm3, %xmm2
-; SSSE3-NEXT:    movapd %xmm2, %xmm1
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hadd_4f64_v4f64_shuffle:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddpd %xmm1, %xmm0
+; SSE-NEXT:    haddpd %xmm3, %xmm2
+; SSE-NEXT:    movapd %xmm2, %xmm1
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: hadd_4f64_v4f64_shuffle:
 ; AVX1:       # %bb.0:
@@ -1016,12 +1205,12 @@ define <4 x double> @hadd_4f64_v4f64_shuffle(<4 x double> %a0, <4 x double> %a1)
 }
 
 define <4 x double> @hsub_4f64_v4f64_shuffle(<4 x double> %a0, <4 x double> %a1) {
-; SSSE3-LABEL: hsub_4f64_v4f64_shuffle:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    hsubpd %xmm1, %xmm0
-; SSSE3-NEXT:    hsubpd %xmm3, %xmm2
-; SSSE3-NEXT:    movapd %xmm2, %xmm1
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hsub_4f64_v4f64_shuffle:
+; SSE:       # %bb.0:
+; SSE-NEXT:    hsubpd %xmm1, %xmm0
+; SSE-NEXT:    hsubpd %xmm3, %xmm2
+; SSE-NEXT:    movapd %xmm2, %xmm1
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: hsub_4f64_v4f64_shuffle:
 ; AVX1:       # %bb.0:
@@ -1044,12 +1233,12 @@ define <4 x double> @hsub_4f64_v4f64_shuffle(<4 x double> %a0, <4 x double> %a1)
 }
 
 define <8 x float> @hadd_8f32_v8f32_shuffle(<8 x float> %a0, <8 x float> %a1) {
-; SSSE3-LABEL: hadd_8f32_v8f32_shuffle:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm1, %xmm0
-; SSSE3-NEXT:    haddps %xmm3, %xmm2
-; SSSE3-NEXT:    movaps %xmm2, %xmm1
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hadd_8f32_v8f32_shuffle:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm1, %xmm0
+; SSE-NEXT:    haddps %xmm3, %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: hadd_8f32_v8f32_shuffle:
 ; AVX1:       # %bb.0:
@@ -1072,12 +1261,12 @@ define <8 x float> @hadd_8f32_v8f32_shuffle(<8 x float> %a0, <8 x float> %a1) {
 }
 
 define <8 x float> @hsub_8f32_v8f32_shuffle(<8 x float> %a0, <8 x float> %a1) {
-; SSSE3-LABEL: hsub_8f32_v8f32_shuffle:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    haddps %xmm1, %xmm0
-; SSSE3-NEXT:    haddps %xmm3, %xmm2
-; SSSE3-NEXT:    movaps %xmm2, %xmm1
-; SSSE3-NEXT:    retq
+; SSE-LABEL: hsub_8f32_v8f32_shuffle:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm1, %xmm0
+; SSE-NEXT:    haddps %xmm3, %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: hsub_8f32_v8f32_shuffle:
 ; AVX1:       # %bb.0:
@@ -1100,6 +1289,19 @@ define <8 x float> @hsub_8f32_v8f32_shuffle(<8 x float> %a0, <8 x float> %a1) {
 }
 
 define <8 x i32> @hadd_8i32_v8i32_shuffle(<8 x i32> %a0, <8 x i32> %a1) {
+; SSE3-LABEL: hadd_8i32_v8i32_shuffle:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movaps %xmm2, %xmm4
+; SSE3-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm3[0,2]
+; SSE3-NEXT:    movaps %xmm0, %xmm5
+; SSE3-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm1[0,2]
+; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,3],xmm3[1,3]
+; SSE3-NEXT:    paddd %xmm4, %xmm2
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; SSE3-NEXT:    paddd %xmm5, %xmm0
+; SSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hadd_8i32_v8i32_shuffle:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddd %xmm1, %xmm0
@@ -1130,6 +1332,20 @@ define <8 x i32> @hadd_8i32_v8i32_shuffle(<8 x i32> %a0, <8 x i32> %a1) {
 }
 
 define <8 x i32> @hsub_8i32_v8i32_shuffle(<8 x i32> %a0, <8 x i32> %a1) {
+; SSE3-LABEL: hsub_8i32_v8i32_shuffle:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movaps %xmm2, %xmm4
+; SSE3-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm3[0,2]
+; SSE3-NEXT:    movaps %xmm0, %xmm5
+; SSE3-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm1[0,2]
+; SSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,3],xmm3[1,3]
+; SSE3-NEXT:    psubd %xmm2, %xmm4
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; SSE3-NEXT:    psubd %xmm0, %xmm5
+; SSE3-NEXT:    movdqa %xmm5, %xmm0
+; SSE3-NEXT:    movdqa %xmm4, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hsub_8i32_v8i32_shuffle:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phsubd %xmm1, %xmm0
@@ -1160,6 +1376,45 @@ define <8 x i32> @hsub_8i32_v8i32_shuffle(<8 x i32> %a0, <8 x i32> %a1) {
 }
 
 define <16 x i16> @hadd_16i16_16i16_shuffle(<16 x i16> %a0, <16 x i16> %a1) {
+; SSE3-LABEL: hadd_16i16_16i16_shuffle:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm5 = xmm2[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm4 = xmm1[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[0,2,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm4[0]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE3-NEXT:    paddw %xmm5, %xmm2
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE3-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE3-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
+; SSE3-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE3-NEXT:    paddw %xmm6, %xmm0
+; SSE3-NEXT:    movdqa %xmm2, %xmm1
+; SSE3-NEXT:    retq
+;
 ; SSSE3-LABEL: hadd_16i16_16i16_shuffle:
 ; SSSE3:       # %bb.0:
 ; SSSE3-NEXT:    phaddw %xmm1, %xmm0