From 362488a72474f16c1af68ad328395ee6e196c489 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Sat, 9 Jan 2016 19:59:27 +0000
Subject: [PATCH] [X86][AVX] Add support for i64 broadcast loads on 32-bit
 targets

Added 32-bit AVX1/AVX2 broadcast tests.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@257264 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86ISelLowering.cpp  |  11 +-
 test/CodeGen/X86/avx-vbroadcast.ll  | 443 +++++++++++-----
 test/CodeGen/X86/avx2-vbroadcast.ll | 763 +++++++++++++++++++---------
 3 files changed, 862 insertions(+), 355 deletions(-)

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index d32146b32cf..7b97e79de5d 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -8173,6 +8173,8 @@ static SDValue lowerVectorShuffleAsBroadcast(SDLoc DL, MVT VT, SDValue V,
             DL, VT, V.getOperand(0), BroadcastIdx, Subtarget, DAG))
       return TruncBroadcast;
 
+  MVT BroadcastVT = VT;
+
   // Also check the simpler case, where we can directly reuse the scalar.
   if (V.getOpcode() == ISD::BUILD_VECTOR ||
       (V.getOpcode() == ISD::SCALAR_TO_VECTOR && BroadcastIdx == 0)) {
@@ -8183,12 +8185,16 @@ static SDValue lowerVectorShuffleAsBroadcast(SDLoc DL, MVT VT, SDValue V,
     if (!Subtarget->hasAVX2() && !isShuffleFoldableLoad(V))
       return SDValue();
   } else if (MayFoldLoad(V) && !cast<LoadSDNode>(V)->isVolatile()) {
+    // 32-bit targets need to load i64 as a f64 and then bitcast the result.
+    if (!Subtarget->is64Bit() && VT.getScalarType() == MVT::i64)
+      BroadcastVT = MVT::getVectorVT(MVT::f64, VT.getVectorNumElements());
+
     // If we are broadcasting a load that is only used by the shuffle
     // then we can reduce the vector load to the broadcasted scalar load.
     LoadSDNode *Ld = cast<LoadSDNode>(V);
     SDValue BaseAddr = Ld->getOperand(1);
     EVT AddrVT = BaseAddr.getValueType();
-    EVT SVT = VT.getScalarType();
+    EVT SVT = BroadcastVT.getScalarType();
     unsigned Offset = BroadcastIdx * SVT.getStoreSize();
     SDValue NewAddr = DAG.getNode(
         ISD::ADD, DL, AddrVT, BaseAddr,
@@ -8202,7 +8208,8 @@ static SDValue lowerVectorShuffleAsBroadcast(SDLoc DL, MVT VT, SDValue V,
     return SDValue();
   }
 
-  return DAG.getNode(X86ISD::VBROADCAST, DL, VT, V);
+  V = DAG.getNode(X86ISD::VBROADCAST, DL, BroadcastVT, V);
+  return DAG.getBitcast(VT, V);
 }
 
 // Check for whether we can use INSERTPS to perform the shuffle. We only use
diff --git a/test/CodeGen/X86/avx-vbroadcast.ll b/test/CodeGen/X86/avx-vbroadcast.ll
index 86b0628aa0b..a98a86b733e 100644
--- a/test/CodeGen/X86/avx-vbroadcast.ll
+++ b/test/CodeGen/X86/avx-vbroadcast.ll
@@ -1,11 +1,24 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx | FileCheck %s
+; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+avx | FileCheck %s --check-prefix=X32
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx | FileCheck %s --check-prefix=X64
 
 define <4 x i64> @A(i64* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: A:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: A:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl (%eax), %ecx
+; X32-NEXT:    movl 4(%eax), %eax
+; X32-NEXT:    vmovd %ecx, %xmm0
+; X32-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
+; X32-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: A:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load i64, i64* %ptr, align 8
   %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
@@ -16,10 +29,16 @@ entry:
 }
 
 define <8 x i32> @B(i32* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: B:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: B:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: B:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load i32, i32* %ptr, align 4
   %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
@@ -30,10 +49,16 @@ entry:
 }
 
 define <4 x double> @C(double* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: C:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: C:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: C:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load double, double* %ptr, align 8
   %vecinit.i = insertelement <4 x double> undef, double %q, i32 0
@@ -44,10 +69,16 @@ entry:
 }
 
 define <8 x float> @D(float* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: D:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: D:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: D:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load float, float* %ptr, align 4
   %vecinit.i = insertelement <8 x float> undef, float %q, i32 0
@@ -60,10 +91,16 @@ entry:
 ;;;; 128-bit versions
 
 define <4 x float> @e(float* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: e:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: e:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: e:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %q = load float, float* %ptr, align 4
   %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
@@ -75,10 +112,15 @@ entry:
 
 ; Don't broadcast constants on pre-AVX2 hardware.
 define <4 x float> @_e2(float* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _e2:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovaps {{.*#+}} xmm0 = [-7.812500e-03,-7.812500e-03,-7.812500e-03,-7.812500e-03]
-; CHECK-NEXT:    retq
+; X32-LABEL: _e2:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    vmovaps {{.*#+}} xmm0 = [-7.812500e-03,-7.812500e-03,-7.812500e-03,-7.812500e-03]
+; X32-NEXT:    retl
+;
+; X64-LABEL: _e2:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovaps {{.*#+}} xmm0 = [-7.812500e-03,-7.812500e-03,-7.812500e-03,-7.812500e-03]
+; X64-NEXT:    retq
 entry:
    %vecinit.i = insertelement <4 x float> undef, float       0xbf80000000000000, i32 0
   %vecinit2.i = insertelement <4 x float> %vecinit.i, float  0xbf80000000000000, i32 1
@@ -89,10 +131,16 @@ entry:
 
 
 define <4 x i32> @F(i32* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: F:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: F:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: F:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %q = load i32, i32* %ptr, align 4
   %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
@@ -105,10 +153,16 @@ entry:
 ; FIXME: Pointer adjusted broadcasts
 
 define <4 x i32> @load_splat_4i32_4i32_1111(<4 x i32>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4i32_4i32_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpshufd {{.*#+}} xmm0 = mem[1,1,1,1]
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4i32_4i32_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpshufd {{.*#+}} xmm0 = mem[1,1,1,1]
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4i32_4i32_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpshufd {{.*#+}} xmm0 = mem[1,1,1,1]
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x i32>, <4 x i32>* %ptr
   %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -116,11 +170,18 @@ entry:
 }
 
 define <8 x i32> @load_splat_8i32_4i32_33333333(<4 x i32>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8i32_4i32_33333333:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
-; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8i32_4i32_33333333:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
+; X32-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8i32_4i32_33333333:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
+; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x i32>, <4 x i32>* %ptr
   %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
@@ -128,13 +189,22 @@ entry:
 }
 
 define <8 x i32> @load_splat_8i32_8i32_55555555(<8 x i32>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8i32_8i32_55555555:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovaps (%rdi), %ymm0
-; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; CHECK-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8i32_8i32_55555555:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovaps (%eax), %ymm0
+; X32-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; X32-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X32-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8i32_8i32_55555555:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovaps (%rdi), %ymm0
+; X64-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; X64-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <8 x i32>, <8 x i32>* %ptr
   %ret = shufflevector <8 x i32> %ld, <8 x i32> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
@@ -142,10 +212,16 @@ entry:
 }
 
 define <4 x float> @load_splat_4f32_4f32_1111(<4 x float>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4f32_4f32_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 4(%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4f32_4f32_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 4(%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4f32_4f32_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 4(%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x float>, <4 x float>* %ptr
   %ret = shufflevector <4 x float> %ld, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -153,10 +229,16 @@ entry:
 }
 
 define <8 x float> @load_splat_8f32_4f32_33333333(<4 x float>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8f32_4f32_33333333:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 12(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8f32_4f32_33333333:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 12(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8f32_4f32_33333333:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 12(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x float>, <4 x float>* %ptr
   %ret = shufflevector <4 x float> %ld, <4 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
@@ -164,10 +246,16 @@ entry:
 }
 
 define <8 x float> @load_splat_8f32_8f32_55555555(<8 x float>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8f32_8f32_55555555:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 20(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8f32_8f32_55555555:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 20(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8f32_8f32_55555555:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 20(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <8 x float>, <8 x float>* %ptr
   %ret = shufflevector <8 x float> %ld, <8 x float> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
@@ -175,10 +263,16 @@ entry:
 }
 
 define <2 x i64> @load_splat_2i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_2i64_2i64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,3,2,3]
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_2i64_2i64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,3,2,3]
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_2i64_2i64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpshufd {{.*#+}} xmm0 = mem[2,3,2,3]
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x i64>, <2 x i64>* %ptr
   %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
@@ -186,12 +280,20 @@ entry:
 }
 
 define <4 x i64> @load_splat_4i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4i64_2i64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovaps (%rdi), %xmm0
-; CHECK-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
-; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4i64_2i64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovaps (%eax), %xmm0
+; X32-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
+; X32-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4i64_2i64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovaps (%rdi), %xmm0
+; X64-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
+; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x i64>, <2 x i64>* %ptr
   %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -199,13 +301,22 @@ entry:
 }
 
 define <4 x i64> @load_splat_4i64_4i64_2222(<4 x i64>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4i64_4i64_2222:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovapd (%rdi), %ymm0
-; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; CHECK-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
-; CHECK-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4i64_4i64_2222:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovapd (%eax), %ymm0
+; X32-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; X32-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X32-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4i64_4i64_2222:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovapd (%rdi), %ymm0
+; X64-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; X64-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x i64>, <4 x i64>* %ptr
   %ret = shufflevector <4 x i64> %ld, <4 x i64> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
@@ -213,11 +324,18 @@ entry:
 }
 
 define <2 x double> @load_splat_2f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_2f64_2f64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovaps (%rdi), %xmm0
-; CHECK-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_2f64_2f64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovaps (%eax), %xmm0
+; X32-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_2f64_2f64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovaps (%rdi), %xmm0
+; X64-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x double>, <2 x double>* %ptr
   %ret = shufflevector <2 x double> %ld, <2 x double> undef, <2 x i32> <i32 1, i32 1>
@@ -225,10 +343,16 @@ entry:
 }
 
 define <4 x double> @load_splat_4f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4f64_2f64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd 8(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4f64_2f64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd 8(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4f64_2f64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd 8(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x double>, <2 x double>* %ptr
   %ret = shufflevector <2 x double> %ld, <2 x double> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -236,10 +360,16 @@ entry:
 }
 
 define <4 x double> @load_splat_4f64_4f64_2222(<4 x double>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4f64_4f64_2222:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd 16(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4f64_4f64_2222:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd 16(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4f64_4f64_2222:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd 16(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x double>, <4 x double>* %ptr
   %ret = shufflevector <4 x double> %ld, <4 x double> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
@@ -249,11 +379,22 @@ entry:
 ; Unsupported vbroadcasts
 
 define <2 x i64> @G(i64* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: G:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
-; CHECK-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; CHECK-NEXT:    retq
+; X32-LABEL: G:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl (%eax), %ecx
+; X32-NEXT:    movl 4(%eax), %eax
+; X32-NEXT:    vmovd %ecx, %xmm0
+; X32-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: G:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; X64-NEXT:    retq
 entry:
   %q = load i64, i64* %ptr, align 8
   %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
@@ -262,20 +403,31 @@ entry:
 }
 
 define <4 x i32> @H(<4 x i32> %a) {
-; CHECK-LABEL: H:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
-; CHECK-NEXT:    retq
+; X32-LABEL: H:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; X32-NEXT:    retl
+;
+; X64-LABEL: H:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
+; X64-NEXT:    retq
 entry:
   %x = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
   ret <4 x i32> %x
 }
 
 define <2 x double> @I(double* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: I:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; CHECK-NEXT:    retq
+; X32-LABEL: I:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; X32-NEXT:    retl
+;
+; X64-LABEL: I:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; X64-NEXT:    retq
 entry:
   %q = load double, double* %ptr, align 4
   %vecinit.i = insertelement <2 x double> undef, double %q, i32 0
@@ -284,12 +436,21 @@ entry:
 }
 
 define <4 x float> @_RR(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _RR:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %xmm0
-; CHECK-NEXT:    movl (%rsi), %eax
-; CHECK-NEXT:    movl %eax, (%rax)
-; CHECK-NEXT:    retq
+; X32-LABEL: _RR:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X32-NEXT:    vbroadcastss (%ecx), %xmm0
+; X32-NEXT:    movl (%eax), %eax
+; X32-NEXT:    movl %eax, (%eax)
+; X32-NEXT:    retl
+;
+; X64-LABEL: _RR:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %xmm0
+; X64-NEXT:    movl (%rsi), %eax
+; X64-NEXT:    movl %eax, (%rax)
+; X64-NEXT:    retq
 entry:
   %q = load float, float* %ptr, align 4
   %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
@@ -303,10 +464,16 @@ entry:
 }
 
 define <4 x float> @_RR2(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _RR2:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _RR2:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _RR2:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %q = load float, float* %ptr, align 4
   %v = insertelement <4 x float> undef, float %q, i32 0
@@ -319,10 +486,16 @@ entry:
 ; (via the insertelements).
 
 define <8 x float> @splat_concat1(float* %p) {
-; CHECK-LABEL: splat_concat1:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat1:
+; X32:       ## BB#0:
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat1:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss (%rdi), %ymm0
+; X64-NEXT:    retq
   %1 = load float, float* %p, align 4
   %2 = insertelement <4 x float> undef, float %1, i32 0
   %3 = insertelement <4 x float> %2, float %1, i32 1
@@ -333,10 +506,16 @@ define <8 x float> @splat_concat1(float* %p) {
 }
 
 define <8 x float> @splat_concat2(float* %p) {
-; CHECK-LABEL: splat_concat2:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat2:
+; X32:       ## BB#0:
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat2:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss (%rdi), %ymm0
+; X64-NEXT:    retq
   %1 = load float, float* %p, align 4
   %2 = insertelement <4 x float> undef, float %1, i32 0
   %3 = insertelement <4 x float> %2, float %1, i32 1
@@ -351,10 +530,16 @@ define <8 x float> @splat_concat2(float* %p) {
 }
 
 define <4 x double> @splat_concat3(double* %p) {
-; CHECK-LABEL: splat_concat3:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastsd (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat3:
+; X32:       ## BB#0:
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat3:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastsd (%rdi), %ymm0
+; X64-NEXT:    retq
   %1 = load double, double* %p, align 8
   %2 = insertelement <2 x double> undef, double %1, i32 0
   %3 = insertelement <2 x double> %2, double %1, i32 1
@@ -363,10 +548,16 @@ define <4 x double> @splat_concat3(double* %p) {
 }
 
 define <4 x double> @splat_concat4(double* %p) {
-; CHECK-LABEL: splat_concat4:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastsd (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat4:
+; X32:       ## BB#0:
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat4:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastsd (%rdi), %ymm0
+; X64-NEXT:    retq
   %1 = load double, double* %p, align 8
   %2 = insertelement <2 x double> undef, double %1, i32 0
   %3 = insertelement <2 x double> %2, double %1, i32 1
diff --git a/test/CodeGen/X86/avx2-vbroadcast.ll b/test/CodeGen/X86/avx2-vbroadcast.ll
index 6b77edb155a..8fd50ae3015 100644
--- a/test/CodeGen/X86/avx2-vbroadcast.ll
+++ b/test/CodeGen/X86/avx2-vbroadcast.ll
@@ -1,11 +1,18 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx2 | FileCheck %s
+; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=X32
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=X64
 
 define <16 x i8> @BB16(i8* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: BB16:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastb (%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: BB16:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastb (%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: BB16:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastb (%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %q = load i8, i8* %ptr, align 4
   %q0 = insertelement <16 x i8> undef, i8 %q, i32 0
@@ -28,10 +35,16 @@ entry:
 }
 
 define <32 x i8> @BB32(i8* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: BB32:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastb (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: BB32:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastb (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: BB32:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastb (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load i8, i8* %ptr, align 4
   %q0 = insertelement <32 x i8> undef, i8 %q, i32 0
@@ -71,10 +84,16 @@ entry:
 }
 
 define <8 x i16> @W16(i16* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: W16:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastw (%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: W16:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastw (%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: W16:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastw (%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %q = load i16, i16* %ptr, align 4
   %q0 = insertelement <8 x i16> undef, i16 %q, i32 0
@@ -89,10 +108,16 @@ entry:
 }
 
 define <16 x i16> @WW16(i16* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: WW16:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastw (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: WW16:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastw (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: WW16:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastw (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load i16, i16* %ptr, align 4
   %q0 = insertelement <16 x i16> undef, i16 %q, i32 0
@@ -115,10 +140,16 @@ entry:
 }
 
 define <4 x i32> @D32(i32* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: D32:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: D32:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: D32:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %q = load i32, i32* %ptr, align 4
   %q0 = insertelement <4 x i32> undef, i32 %q, i32 0
@@ -129,10 +160,16 @@ entry:
 }
 
 define <8 x i32> @DD32(i32* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: DD32:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: DD32:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss (%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: DD32:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load i32, i32* %ptr, align 4
   %q0 = insertelement <8 x i32> undef, i32 %q, i32 0
@@ -147,10 +184,21 @@ entry:
 }
 
 define <2 x i64> @Q64(i64* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: Q64:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastq (%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: Q64:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl (%eax), %ecx
+; X32-NEXT:    movl 4(%eax), %eax
+; X32-NEXT:    vmovd %ecx, %xmm0
+; X32-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: Q64:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastq (%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %q = load i64, i64* %ptr, align 4
   %q0 = insertelement <2 x i64> undef, i64 %q, i32 0
@@ -159,10 +207,22 @@ entry:
 }
 
 define <4 x i64> @QQ64(i64* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: QQ64:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd (%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: QQ64:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    movl (%eax), %ecx
+; X32-NEXT:    movl 4(%eax), %eax
+; X32-NEXT:    vmovd %ecx, %xmm0
+; X32-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0
+; X32-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0
+; X32-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: QQ64:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd (%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %q = load i64, i64* %ptr, align 4
   %q0 = insertelement <4 x i64> undef, i64 %q, i32 0
@@ -175,10 +235,16 @@ entry:
 ; FIXME: Pointer adjusted broadcasts
 
 define <16 x i8> @load_splat_16i8_16i8_1111111111111111(<16 x i8>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_16i8_16i8_1111111111111111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastb 1(%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_16i8_16i8_1111111111111111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastb 1(%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_16i8_16i8_1111111111111111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastb 1(%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %ld = load <16 x i8>, <16 x i8>* %ptr
   %ret = shufflevector <16 x i8> %ld, <16 x i8> undef, <16 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -186,10 +252,16 @@ entry:
 }
 
 define <32 x i8> @load_splat_32i8_16i8_11111111111111111111111111111111(<16 x i8>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_32i8_16i8_11111111111111111111111111111111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastb 1(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_32i8_16i8_11111111111111111111111111111111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastb 1(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_32i8_16i8_11111111111111111111111111111111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastb 1(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <16 x i8>, <16 x i8>* %ptr
   %ret = shufflevector <16 x i8> %ld, <16 x i8> undef, <32 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -197,10 +269,16 @@ entry:
 }
 
 define <32 x i8> @load_splat_32i8_32i8_11111111111111111111111111111111(<32 x i8>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_32i8_32i8_11111111111111111111111111111111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastb 1(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_32i8_32i8_11111111111111111111111111111111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastb 1(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_32i8_32i8_11111111111111111111111111111111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastb 1(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <32 x i8>, <32 x i8>* %ptr
   %ret = shufflevector <32 x i8> %ld, <32 x i8> undef, <32 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -208,10 +286,16 @@ entry:
 }
 
 define <8 x i16> @load_splat_8i16_8i16_11111111(<8 x i16>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8i16_8i16_11111111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastw 2(%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8i16_8i16_11111111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastw 2(%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8i16_8i16_11111111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastw 2(%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %ld = load <8 x i16>, <8 x i16>* %ptr
   %ret = shufflevector <8 x i16> %ld, <8 x i16> undef, <8 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -219,10 +303,16 @@ entry:
 }
 
 define <16 x i16> @load_splat_16i16_8i16_1111111111111111(<8 x i16>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_16i16_8i16_1111111111111111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastw 2(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_16i16_8i16_1111111111111111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastw 2(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_16i16_8i16_1111111111111111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastw 2(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <8 x i16>, <8 x i16>* %ptr
   %ret = shufflevector <8 x i16> %ld, <8 x i16> undef, <16 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -230,10 +320,16 @@ entry:
 }
 
 define <16 x i16> @load_splat_16i16_16i16_1111111111111111(<16 x i16>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_16i16_16i16_1111111111111111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastw 2(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_16i16_16i16_1111111111111111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vpbroadcastw 2(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_16i16_16i16_1111111111111111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastw 2(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <16 x i16>, <16 x i16>* %ptr
   %ret = shufflevector <16 x i16> %ld, <16 x i16> undef, <16 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
@@ -241,10 +337,16 @@ entry:
 }
 
 define <4 x i32> @load_splat_4i32_4i32_1111(<4 x i32>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4i32_4i32_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 4(%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4i32_4i32_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 4(%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4i32_4i32_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 4(%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x i32>, <4 x i32>* %ptr
   %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -252,10 +354,16 @@ entry:
 }
 
 define <8 x i32> @load_splat_8i32_4i32_33333333(<4 x i32>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8i32_4i32_33333333:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 12(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8i32_4i32_33333333:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 12(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8i32_4i32_33333333:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 12(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x i32>, <4 x i32>* %ptr
   %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
@@ -263,10 +371,16 @@ entry:
 }
 
 define <8 x i32> @load_splat_8i32_8i32_55555555(<8 x i32>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8i32_8i32_55555555:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 20(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8i32_8i32_55555555:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 20(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8i32_8i32_55555555:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 20(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <8 x i32>, <8 x i32>* %ptr
   %ret = shufflevector <8 x i32> %ld, <8 x i32> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
@@ -274,10 +388,16 @@ entry:
 }
 
 define <4 x float> @load_splat_4f32_4f32_1111(<4 x float>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4f32_4f32_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 4(%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4f32_4f32_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 4(%eax), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4f32_4f32_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 4(%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x float>, <4 x float>* %ptr
   %ret = shufflevector <4 x float> %ld, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -285,10 +405,16 @@ entry:
 }
 
 define <8 x float> @load_splat_8f32_4f32_33333333(<4 x float>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8f32_4f32_33333333:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 12(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8f32_4f32_33333333:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 12(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8f32_4f32_33333333:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 12(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x float>, <4 x float>* %ptr
   %ret = shufflevector <4 x float> %ld, <4 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
@@ -296,10 +422,16 @@ entry:
 }
 
 define <8 x float> @load_splat_8f32_8f32_55555555(<8 x float>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_8f32_8f32_55555555:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss 20(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_8f32_8f32_55555555:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastss 20(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_8f32_8f32_55555555:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss 20(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <8 x float>, <8 x float>* %ptr
   %ret = shufflevector <8 x float> %ld, <8 x float> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
@@ -307,10 +439,17 @@ entry:
 }
 
 define <2 x i64> @load_splat_2i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_2i64_2i64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastq 8(%rdi), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_2i64_2i64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; X32-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_2i64_2i64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastq 8(%rdi), %xmm0
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x i64>, <2 x i64>* %ptr
   %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
@@ -318,10 +457,16 @@ entry:
 }
 
 define <4 x i64> @load_splat_4i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4i64_2i64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd 8(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4i64_2i64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd 8(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4i64_2i64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd 8(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x i64>, <2 x i64>* %ptr
   %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -329,10 +474,16 @@ entry:
 }
 
 define <4 x i64> @load_splat_4i64_4i64_2222(<4 x i64>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4i64_4i64_2222:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd 16(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4i64_4i64_2222:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd 16(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4i64_4i64_2222:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd 16(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x i64>, <4 x i64>* %ptr
   %ret = shufflevector <4 x i64> %ld, <4 x i64> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
@@ -340,11 +491,18 @@ entry:
 }
 
 define <2 x double> @load_splat_2f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_2f64_2f64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovaps (%rdi), %xmm0
-; CHECK-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_2f64_2f64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovaps (%eax), %xmm0
+; X32-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_2f64_2f64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovaps (%rdi), %xmm0
+; X64-NEXT:    vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x double>, <2 x double>* %ptr
   %ret = shufflevector <2 x double> %ld, <2 x double> undef, <2 x i32> <i32 1, i32 1>
@@ -352,10 +510,16 @@ entry:
 }
 
 define <4 x double> @load_splat_4f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4f64_2f64_1111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd 8(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4f64_2f64_1111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd 8(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4f64_2f64_1111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd 8(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <2 x double>, <2 x double>* %ptr
   %ret = shufflevector <2 x double> %ld, <2 x double> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
@@ -363,10 +527,16 @@ entry:
 }
 
 define <4 x double> @load_splat_4f64_4f64_2222(<4 x double>* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: load_splat_4f64_4f64_2222:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastsd 16(%rdi), %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: load_splat_4f64_4f64_2222:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vbroadcastsd 16(%eax), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: load_splat_4f64_4f64_2222:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastsd 16(%rdi), %ymm0
+; X64-NEXT:    retq
 entry:
   %ld = load <4 x double>, <4 x double>* %ptr
   %ret = shufflevector <4 x double> %ld, <4 x double> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
@@ -376,10 +546,16 @@ entry:
 ; make sure that we still don't support broadcast double into 128-bit vector
 ; this used to crash
 define <2 x double> @I(double* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: I:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; CHECK-NEXT:    retq
+; X32-LABEL: I:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; X32-NEXT:    retl
+;
+; X64-LABEL: I:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; X64-NEXT:    retq
 entry:
   %q = load double, double* %ptr, align 4
   %vecinit.i = insertelement <2 x double> undef, double %q, i32 0
@@ -388,32 +564,49 @@ entry:
 }
 
 define <8 x i32> @V111(<8 x i32> %in) nounwind uwtable readnone ssp {
-; CHECK-LABEL: V111:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
-; CHECK-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: V111:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    vpbroadcastd LCPI27_0, %ymm1
+; X32-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: V111:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vpbroadcastd {{.*}}(%rip), %ymm1
+; X64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
+; X64-NEXT:    retq
 entry:
   %g = add <8 x i32> %in, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <8 x i32> %g
 }
 
 define <8 x float> @V113(<8 x float> %in) nounwind uwtable readnone ssp {
-; CHECK-LABEL: V113:
-; CHECK:       ## BB#0: ## %entry
-; CHECK-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1
-; CHECK-NEXT:    vaddps %ymm1, %ymm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: V113:
+; X32:       ## BB#0: ## %entry
+; X32-NEXT:    vbroadcastss LCPI28_0, %ymm1
+; X32-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: V113:
+; X64:       ## BB#0: ## %entry
+; X64-NEXT:    vbroadcastss {{.*}}(%rip), %ymm1
+; X64-NEXT:    vaddps %ymm1, %ymm0, %ymm0
+; X64-NEXT:    retq
 entry:
   %g = fadd <8 x float> %in, <float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000, float 0xbf80000000000000>
   ret <8 x float> %g
 }
 
 define <4 x float> @_e2(float* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _e2:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss {{.*}}(%rip), %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _e2:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss LCPI29_0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _e2:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss {{.*}}(%rip), %xmm0
+; X64-NEXT:    retq
   %vecinit.i = insertelement <4 x float> undef, float        0xbf80000000000000, i32 0
   %vecinit2.i = insertelement <4 x float> %vecinit.i, float  0xbf80000000000000, i32 1
   %vecinit4.i = insertelement <4 x float> %vecinit2.i, float 0xbf80000000000000, i32 2
@@ -422,10 +615,15 @@ define <4 x float> @_e2(float* %ptr) nounwind uwtable readnone ssp {
 }
 
 define <8 x i8> @_e4(i8* %ptr) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _e4:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vmovaps {{.*#+}} xmm0 = [52,52,52,52,52,52,52,52]
-; CHECK-NEXT:    retq
+; X32-LABEL: _e4:
+; X32:       ## BB#0:
+; X32-NEXT:    vmovaps {{.*#+}} xmm0 = [52,52,52,52,52,52,52,52]
+; X32-NEXT:    retl
+;
+; X64-LABEL: _e4:
+; X64:       ## BB#0:
+; X64-NEXT:    vmovaps {{.*#+}} xmm0 = [52,52,52,52,52,52,52,52]
+; X64-NEXT:    retq
   %vecinit0.i = insertelement <8 x i8> undef, i8       52, i32 0
   %vecinit1.i = insertelement <8 x i8> %vecinit0.i, i8 52, i32 1
   %vecinit2.i = insertelement <8 x i8> %vecinit1.i, i8 52, i32 2
@@ -437,19 +635,30 @@ define <8 x i8> @_e4(i8* %ptr) nounwind uwtable readnone ssp {
   ret <8 x i8> %vecinit7.i
 }
 
-
 define void @crash() nounwind alwaysinline {
-; CHECK-LABEL: crash:
-; CHECK:       ## BB#0: ## %WGLoopsEntry
-; CHECK-NEXT:    xorl %eax, %eax
-; CHECK-NEXT:    testb %al, %al
-; CHECK-NEXT:    je LBB31_1
-; CHECK-NEXT:  ## BB#2: ## %ret
-; CHECK-NEXT:    retq
-; CHECK-NEXT:    .align 4, 0x90
-; CHECK-NEXT:  LBB31_1: ## %footer349VF
-; CHECK-NEXT:    ## =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    jmp LBB31_1
+; X32-LABEL: crash:
+; X32:       ## BB#0: ## %WGLoopsEntry
+; X32-NEXT:    xorl %eax, %eax
+; X32-NEXT:    testb %al, %al
+; X32-NEXT:    je LBB31_1
+; X32-NEXT:  ## BB#2: ## %ret
+; X32-NEXT:    retl
+; X32-NEXT:    .align 4, 0x90
+; X32-NEXT:  LBB31_1: ## %footer349VF
+; X32-NEXT:    ## =>This Inner Loop Header: Depth=1
+; X32-NEXT:    jmp LBB31_1
+;
+; X64-LABEL: crash:
+; X64:       ## BB#0: ## %WGLoopsEntry
+; X64-NEXT:    xorl %eax, %eax
+; X64-NEXT:    testb %al, %al
+; X64-NEXT:    je LBB31_1
+; X64-NEXT:  ## BB#2: ## %ret
+; X64-NEXT:    retq
+; X64-NEXT:    .align 4, 0x90
+; X64-NEXT:  LBB31_1: ## %footer349VF
+; X64-NEXT:    ## =>This Inner Loop Header: Depth=1
+; X64-NEXT:    jmp LBB31_1
 WGLoopsEntry:
   br i1 undef, label %ret, label %footer329VF
 
@@ -477,150 +686,230 @@ ret:
 }
 
 define <8 x i32> @_inreg0(i32 %scalar) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _inreg0:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vmovd %edi, %xmm0
-; CHECK-NEXT:    vbroadcastss %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg0:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss {{[0-9]+}}(%esp), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg0:
+; X64:       ## BB#0:
+; X64-NEXT:    vmovd %edi, %xmm0
+; X64-NEXT:    vbroadcastss %xmm0, %ymm0
+; X64-NEXT:    retq
   %in = insertelement <8 x i32> undef, i32 %scalar, i32 0
   %wide = shufflevector <8 x i32> %in, <8 x i32> undef, <8 x i32> zeroinitializer
   ret <8 x i32> %wide
 }
 
 define <8 x float> @_inreg1(float %scalar) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _inreg1:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg1:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss {{[0-9]+}}(%esp), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg1:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %ymm0
+; X64-NEXT:    retq
   %in = insertelement <8 x float> undef, float %scalar, i32 0
   %wide = shufflevector <8 x float> %in, <8 x float> undef, <8 x i32> zeroinitializer
   ret <8 x float> %wide
 }
 
 define <4 x float> @_inreg2(float %scalar) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _inreg2:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg2:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss {{[0-9]+}}(%esp), %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg2:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %xmm0
+; X64-NEXT:    retq
   %in = insertelement <4 x float> undef, float %scalar, i32 0
   %wide = shufflevector <4 x float> %in, <4 x float> undef, <4 x i32> zeroinitializer
   ret <4 x float> %wide
 }
 
 define <4 x double> @_inreg3(double %scalar) nounwind uwtable readnone ssp {
-; CHECK-LABEL: _inreg3:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastsd %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg3:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastsd {{[0-9]+}}(%esp), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg3:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X64-NEXT:    retq
   %in = insertelement <4 x double> undef, double %scalar, i32 0
   %wide = shufflevector <4 x double> %in, <4 x double> undef, <4 x i32> zeroinitializer
   ret <4 x double> %wide
 }
 
 define   <8 x float> @_inreg8xfloat(<8 x float> %a) {
-; CHECK-LABEL: _inreg8xfloat:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg8xfloat:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss %xmm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg8xfloat:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %ymm0
+; X64-NEXT:    retq
   %b = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> zeroinitializer
   ret <8 x float> %b
 }
 
 define   <4 x float> @_inreg4xfloat(<4 x float> %a) {
-; CHECK-LABEL: _inreg4xfloat:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg4xfloat:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg4xfloat:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %xmm0
+; X64-NEXT:    retq
   %b = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> zeroinitializer
   ret <4 x float> %b
 }
 
 define   <16 x i16> @_inreg16xi16(<16 x i16> %a) {
-; CHECK-LABEL: _inreg16xi16:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpbroadcastw %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg16xi16:
+; X32:       ## BB#0:
+; X32-NEXT:    vpbroadcastw %xmm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg16xi16:
+; X64:       ## BB#0:
+; X64-NEXT:    vpbroadcastw %xmm0, %ymm0
+; X64-NEXT:    retq
   %b = shufflevector <16 x i16> %a, <16 x i16> undef, <16 x i32> zeroinitializer
   ret <16 x i16> %b
 }
 
 define   <8 x i16> @_inreg8xi16(<8 x i16> %a) {
-; CHECK-LABEL: _inreg8xi16:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpbroadcastw %xmm0, %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg8xi16:
+; X32:       ## BB#0:
+; X32-NEXT:    vpbroadcastw %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg8xi16:
+; X64:       ## BB#0:
+; X64-NEXT:    vpbroadcastw %xmm0, %xmm0
+; X64-NEXT:    retq
   %b = shufflevector <8 x i16> %a, <8 x i16> undef, <8 x i32> zeroinitializer
   ret <8 x i16> %b
 }
 
 define   <4 x i64> @_inreg4xi64(<4 x i64> %a) {
-; CHECK-LABEL: _inreg4xi64:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastsd %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg4xi64:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg4xi64:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X64-NEXT:    retq
   %b = shufflevector <4 x i64> %a, <4 x i64> undef, <4 x i32> zeroinitializer
   ret <4 x i64> %b
 }
 
 define   <2 x i64> @_inreg2xi64(<2 x i64> %a) {
-; CHECK-LABEL: _inreg2xi64:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpbroadcastq %xmm0, %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg2xi64:
+; X32:       ## BB#0:
+; X32-NEXT:    vpbroadcastq %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg2xi64:
+; X64:       ## BB#0:
+; X64-NEXT:    vpbroadcastq %xmm0, %xmm0
+; X64-NEXT:    retq
   %b = shufflevector <2 x i64> %a, <2 x i64> undef, <2 x i32> zeroinitializer
   ret <2 x i64> %b
 }
 
 define   <4 x double> @_inreg4xdouble(<4 x double> %a) {
-; CHECK-LABEL: _inreg4xdouble:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastsd %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg4xdouble:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg4xdouble:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X64-NEXT:    retq
   %b = shufflevector <4 x double> %a, <4 x double> undef, <4 x i32> zeroinitializer
   ret <4 x double> %b
 }
 
 define   <2 x double> @_inreg2xdouble(<2 x double> %a) {
-; CHECK-LABEL: _inreg2xdouble:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg2xdouble:
+; X32:       ## BB#0:
+; X32-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg2xdouble:
+; X64:       ## BB#0:
+; X64-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
+; X64-NEXT:    retq
   %b = shufflevector <2 x double> %a, <2 x double> undef, <2 x i32> zeroinitializer
   ret <2 x double> %b
 }
 
 define   <8 x i32> @_inreg8xi32(<8 x i32> %a) {
-; CHECK-LABEL: _inreg8xi32:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg8xi32:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss %xmm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg8xi32:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %ymm0
+; X64-NEXT:    retq
   %b = shufflevector <8 x i32> %a, <8 x i32> undef, <8 x i32> zeroinitializer
   ret <8 x i32> %b
 }
 
 define   <4 x i32> @_inreg4xi32(<4 x i32> %a) {
-; CHECK-LABEL: _inreg4xi32:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg4xi32:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg4xi32:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %xmm0
+; X64-NEXT:    retq
   %b = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> zeroinitializer
   ret <4 x i32> %b
 }
 
 define   <32 x i8> @_inreg32xi8(<32 x i8> %a) {
-; CHECK-LABEL: _inreg32xi8:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpbroadcastb %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg32xi8:
+; X32:       ## BB#0:
+; X32-NEXT:    vpbroadcastb %xmm0, %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg32xi8:
+; X64:       ## BB#0:
+; X64-NEXT:    vpbroadcastb %xmm0, %ymm0
+; X64-NEXT:    retq
   %b = shufflevector <32 x i8> %a, <32 x i8> undef, <32 x i32> zeroinitializer
   ret <32 x i8> %b
 }
 
 define   <16 x i8> @_inreg16xi8(<16 x i8> %a) {
-; CHECK-LABEL: _inreg16xi8:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vpbroadcastb %xmm0, %xmm0
-; CHECK-NEXT:    retq
+; X32-LABEL: _inreg16xi8:
+; X32:       ## BB#0:
+; X32-NEXT:    vpbroadcastb %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: _inreg16xi8:
+; X64:       ## BB#0:
+; X64-NEXT:    vpbroadcastb %xmm0, %xmm0
+; X64-NEXT:    retq
   %b = shufflevector <16 x i8> %a, <16 x i8> undef, <16 x i32> zeroinitializer
   ret <16 x i8> %b
 }
@@ -630,10 +919,15 @@ define   <16 x i8> @_inreg16xi8(<16 x i8> %a) {
 ; (via the insertelements).
 
 define <8 x float> @splat_concat1(float %f) {
-; CHECK-LABEL: splat_concat1:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat1:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss {{[0-9]+}}(%esp), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat1:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %ymm0
+; X64-NEXT:    retq
   %1 = insertelement <4 x float> undef, float %f, i32 0
   %2 = insertelement <4 x float> %1, float %f, i32 1
   %3 = insertelement <4 x float> %2, float %f, i32 2
@@ -643,10 +937,15 @@ define <8 x float> @splat_concat1(float %f) {
 }
 
 define <8 x float> @splat_concat2(float %f) {
-; CHECK-LABEL: splat_concat2:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastss %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat2:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastss {{[0-9]+}}(%esp), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat2:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastss %xmm0, %ymm0
+; X64-NEXT:    retq
   %1 = insertelement <4 x float> undef, float %f, i32 0
   %2 = insertelement <4 x float> %1, float %f, i32 1
   %3 = insertelement <4 x float> %2, float %f, i32 2
@@ -660,10 +959,15 @@ define <8 x float> @splat_concat2(float %f) {
 }
 
 define <4 x double> @splat_concat3(double %d) {
-; CHECK-LABEL: splat_concat3:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastsd %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat3:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastsd {{[0-9]+}}(%esp), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat3:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X64-NEXT:    retq
   %1 = insertelement <2 x double> undef, double %d, i32 0
   %2 = insertelement <2 x double> %1, double %d, i32 1
   %3 = shufflevector <2 x double> %2, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
@@ -671,10 +975,15 @@ define <4 x double> @splat_concat3(double %d) {
 }
 
 define <4 x double> @splat_concat4(double %d) {
-; CHECK-LABEL: splat_concat4:
-; CHECK:       ## BB#0:
-; CHECK-NEXT:    vbroadcastsd %xmm0, %ymm0
-; CHECK-NEXT:    retq
+; X32-LABEL: splat_concat4:
+; X32:       ## BB#0:
+; X32-NEXT:    vbroadcastsd {{[0-9]+}}(%esp), %ymm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: splat_concat4:
+; X64:       ## BB#0:
+; X64-NEXT:    vbroadcastsd %xmm0, %ymm0
+; X64-NEXT:    retq
   %1 = insertelement <2 x double> undef, double %d, i32 0
   %2 = insertelement <2 x double> %1, double %d, i32 1
   %3 = insertelement <2 x double> undef, double %d, i32 0
@@ -805,9 +1114,9 @@ eintry:
   ret void
 }
 
-; CHECK-LABEL: isel_crash_2q
-; CHECK: vpbroadcastq {{[^,]+}}, %xmm{{[0-9]+}}
-; CHECK: ret
+; X64-LABEL: isel_crash_2q
+; X64: vpbroadcastq {{[^,]+}}, %xmm{{[0-9]+}}
+; X64: ret
 define void @isel_crash_2q(i64* %cV_R.addr) {
 entry:
   %__a.addr.i = alloca <2 x i64>, align 16
@@ -823,9 +1132,9 @@ entry:
   ret void
 }
 
-; CHECK-LABEL: isel_crash_4q
-; CHECK: vbroadcastsd {{[^,]+}}, %ymm{{[0-9]+}}
-; CHECK: ret
+; X64-LABEL: isel_crash_4q
+; X64: vbroadcastsd {{[^,]+}}, %ymm{{[0-9]+}}
+; X64: ret
 define void @isel_crash_4q(i64* %cV_R.addr) {
 eintry:
   %__a.addr.i = alloca <4 x i64>, align 16