diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -19426,6 +19426,41 @@
       Ops.append(NumElts, DAG.getUNDEF(InVal.getValueType()));
       return UpdateBuildVector(Ops);
     }
+
+    // If we're inserting into the end of a vector as part of an sequence, see
+    // if we can create a BUILD_VECTOR be following the sequence back up the
+    // chain.
+    if (Elt == (NumElts - 1)) {
+      SmallVector<SDValue> ReverseInsertions;
+      ReverseInsertions.push_back(InVal);
+
+      EVT MaxEltVT = InVal.getValueType();
+      SDValue CurVec = InVec;
+      for (unsigned I = 1; I != NumElts; ++I) {
+        if (CurVec.getOpcode() != ISD::INSERT_VECTOR_ELT || !CurVec.hasOneUse())
+          break;
+
+        auto *CurIdx = dyn_cast<ConstantSDNode>(CurVec.getOperand(2));
+        if (!CurIdx || CurIdx->getAPIntValue() != ((NumElts - 1) - I))
+          break;
+        SDValue CurVal = CurVec.getOperand(1);
+        ReverseInsertions.push_back(CurVal);
+        if (VT.isInteger()) {
+          EVT CurValVT = CurVal.getValueType();
+          MaxEltVT = MaxEltVT.bitsGE(CurValVT) ? MaxEltVT : CurValVT;
+        }
+        CurVec = CurVec.getOperand(0);
+      }
+
+      if (ReverseInsertions.size() == NumElts) {
+        for (unsigned I = 0; I != NumElts; ++I) {
+          SDValue Val = ReverseInsertions[(NumElts - 1) - I];
+          Val = VT.isInteger() ? DAG.getAnyExtOrTrunc(Val, DL, MaxEltVT) : Val;
+          Ops.push_back(Val);
+        }
+        return DAG.getBuildVector(VT, DL, Ops);
+      }
+    }
   }
 
   return SDValue();
diff --git a/llvm/test/CodeGen/AArch64/arm64-neon-copy.ll b/llvm/test/CodeGen/AArch64/arm64-neon-copy.ll
--- a/llvm/test/CodeGen/AArch64/arm64-neon-copy.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-neon-copy.ll
@@ -258,9 +258,7 @@
 define <1 x i64> @ins2d1(<2 x i64> %tmp1, <1 x i64> %tmp2) {
 ; CHECK-LABEL: ins2d1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    // kill: def $d1 killed $d1 def $q1
-; CHECK-NEXT:    mov v1.d[0], v0.d[0]
-; CHECK-NEXT:    fmov d0, d1
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    ret
   %tmp3 = extractelement <2 x i64> %tmp1, i32 0
   %tmp4 = insertelement <1 x i64> %tmp2, i64 %tmp3, i32 0
@@ -282,7 +280,7 @@
 define <1 x double> @ins2f1(<2 x double> %tmp1, <1 x double> %tmp2) {
 ; CHECK-LABEL: ins2f1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    dup v0.2d, v0.d[1]
+; CHECK-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    ret
   %tmp3 = extractelement <2 x double> %tmp1, i32 1
diff --git a/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll b/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll
--- a/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fixed-length-insert-vector-elt.ll
@@ -228,7 +228,8 @@
 define <1 x double> @insertelement_v1f64(<1 x double> %op1) #0 {
 ; VBITS_GE_256-LABEL: insertelement_v1f64:
 ; VBITS_GE_256:       // %bb.0:
-; VBITS_GE_256-NEXT:    fmov d0, #5.00000000
+; VBITS_GE_256-NEXT:    mov x8, #4617315517961601024
+; VBITS_GE_256-NEXT:    fmov d0, x8
 ; VBITS_GE_256-NEXT:    ret
     %r = insertelement <1 x double> %op1, double 5.0, i64 0
     ret <1 x double> %r
diff --git a/llvm/test/CodeGen/AArch64/vecreduce-propagate-sd-flags.ll b/llvm/test/CodeGen/AArch64/vecreduce-propagate-sd-flags.ll
--- a/llvm/test/CodeGen/AArch64/vecreduce-propagate-sd-flags.ll
+++ b/llvm/test/CodeGen/AArch64/vecreduce-propagate-sd-flags.ll
@@ -13,7 +13,7 @@
 ; CHECK: Legally typed node: [[VTWOA]]: v2f64 = BUILD_VECTOR
 ; CHECK: Legalizing node: [[VTWOB:t.*]]: v2f64 = BUILD_VECTOR
 ; CHECK: Legally typed node: [[VTWOB]]: v2f64 = BUILD_VECTOR
-; CHECK: Legalizing node: t34: v2f64 = fmaxnum nnan reassoc [[VTWOB]], [[VTWOA]]
+; CHECK: Legalizing node: t30: v2f64 = fmaxnum nnan reassoc [[VTWOB]], [[VTWOA]]
 
 target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
 target triple = "aarch64--linux-gnu"
diff --git a/llvm/test/CodeGen/ARM/neon-copy.ll b/llvm/test/CodeGen/ARM/neon-copy.ll
--- a/llvm/test/CodeGen/ARM/neon-copy.ll
+++ b/llvm/test/CodeGen/ARM/neon-copy.ll
@@ -257,10 +257,7 @@
 define <1 x i64> @ins2d1(<2 x i64> %tmp1, <1 x i64> %tmp2) {
 ; CHECK-LABEL: ins2d1:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vmov r0, r1, d0
-; CHECK-NEXT:    vmov.32 d2[0], r0
-; CHECK-NEXT:    vmov.32 d2[1], r1
-; CHECK-NEXT:    vorr d0, d2, d2
+; CHECK-NEXT:    @ kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bx lr
   %tmp3 = extractelement <2 x i64> %tmp1, i32 0
   %tmp4 = insertelement <1 x i64> %tmp2, i64 %tmp3, i32 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-i1.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-i1.ll
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-i1.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-insert-i1.ll
@@ -7,13 +7,9 @@
 define <1 x i1> @insertelt_v1i1(<1 x i1> %x, i1 %elt) nounwind {
 ; CHECK-LABEL: insertelt_v1i1:
 ; CHECK:       # %bb.0:
+; CHECK-NEXT:    andi a0, a0, 1
 ; CHECK-NEXT:    vsetivli zero, 1, e8, mf8, ta, mu
-; CHECK-NEXT:    vmv.v.i v8, 0
-; CHECK-NEXT:    vmerge.vim v8, v8, 1, v0
-; CHECK-NEXT:    vsetvli zero, zero, e8, mf8, tu, mu
-; CHECK-NEXT:    vmv.s.x v8, a0
-; CHECK-NEXT:    vsetvli zero, zero, e8, mf8, ta, mu
-; CHECK-NEXT:    vand.vi v8, v8, 1
+; CHECK-NEXT:    vmv.v.x v8, a0
 ; CHECK-NEXT:    vmsne.vi v0, v8, 0
 ; CHECK-NEXT:    ret
   %y = insertelement <1 x i1> %x, i1 %elt, i64 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll
@@ -41,9 +41,8 @@
 ; RV64ZVE32F-NEXT:    andi a1, a1, 1
 ; RV64ZVE32F-NEXT:    beqz a1, .LBB0_2
 ; RV64ZVE32F-NEXT:  # %bb.1: # %cond.load
-; RV64ZVE32F-NEXT:    lb a0, 0(a0)
-; RV64ZVE32F-NEXT:    vsetivli zero, 1, e8, mf4, tu, mu
-; RV64ZVE32F-NEXT:    vmv.s.x v8, a0
+; RV64ZVE32F-NEXT:    vsetivli zero, 1, e8, mf4, ta, mu
+; RV64ZVE32F-NEXT:    vlse8.v v8, (a0), zero
 ; RV64ZVE32F-NEXT:  .LBB0_2: # %else
 ; RV64ZVE32F-NEXT:    ret
   %v = call <1 x i8> @llvm.masked.gather.v1i8.v1p0i8(<1 x i8*> %ptrs, i32 1, <1 x i1> %m, <1 x i8> %passthru)
@@ -1012,9 +1011,8 @@
 ; RV64ZVE32F-NEXT:    andi a1, a1, 1
 ; RV64ZVE32F-NEXT:    beqz a1, .LBB13_2
 ; RV64ZVE32F-NEXT:  # %bb.1: # %cond.load
-; RV64ZVE32F-NEXT:    lh a0, 0(a0)
-; RV64ZVE32F-NEXT:    vsetivli zero, 1, e16, mf2, tu, mu
-; RV64ZVE32F-NEXT:    vmv.s.x v8, a0
+; RV64ZVE32F-NEXT:    vsetivli zero, 1, e16, mf2, ta, mu
+; RV64ZVE32F-NEXT:    vlse16.v v8, (a0), zero
 ; RV64ZVE32F-NEXT:  .LBB13_2: # %else
 ; RV64ZVE32F-NEXT:    ret
   %v = call <1 x i16> @llvm.masked.gather.v1i16.v1p0i16(<1 x i16*> %ptrs, i32 2, <1 x i1> %m, <1 x i16> %passthru)
@@ -2325,9 +2323,8 @@
 ; RV64ZVE32F-NEXT:    andi a1, a1, 1
 ; RV64ZVE32F-NEXT:    beqz a1, .LBB27_2
 ; RV64ZVE32F-NEXT:  # %bb.1: # %cond.load
-; RV64ZVE32F-NEXT:    lw a0, 0(a0)
-; RV64ZVE32F-NEXT:    vsetivli zero, 1, e32, m1, tu, mu
-; RV64ZVE32F-NEXT:    vmv.s.x v8, a0
+; RV64ZVE32F-NEXT:    vsetivli zero, 1, e32, m1, ta, mu
+; RV64ZVE32F-NEXT:    vlse32.v v8, (a0), zero
 ; RV64ZVE32F-NEXT:  .LBB27_2: # %else
 ; RV64ZVE32F-NEXT:    ret
   %v = call <1 x i32> @llvm.masked.gather.v1i32.v1p0i32(<1 x i32*> %ptrs, i32 4, <1 x i1> %m, <1 x i32> %passthru)
@@ -7574,9 +7571,8 @@
 ; RV64ZVE32F-NEXT:    andi a1, a1, 1
 ; RV64ZVE32F-NEXT:    beqz a1, .LBB58_2
 ; RV64ZVE32F-NEXT:  # %bb.1: # %cond.load
-; RV64ZVE32F-NEXT:    flh ft0, 0(a0)
-; RV64ZVE32F-NEXT:    vsetivli zero, 1, e16, mf2, tu, mu
-; RV64ZVE32F-NEXT:    vfmv.s.f v8, ft0
+; RV64ZVE32F-NEXT:    vsetivli zero, 1, e16, mf2, ta, mu
+; RV64ZVE32F-NEXT:    vlse16.v v8, (a0), zero
 ; RV64ZVE32F-NEXT:  .LBB58_2: # %else
 ; RV64ZVE32F-NEXT:    ret
   %v = call <1 x half> @llvm.masked.gather.v1f16.v1p0f16(<1 x half*> %ptrs, i32 2, <1 x i1> %m, <1 x half> %passthru)
@@ -8594,9 +8590,8 @@
 ; RV64ZVE32F-NEXT:    andi a1, a1, 1
 ; RV64ZVE32F-NEXT:    beqz a1, .LBB68_2
 ; RV64ZVE32F-NEXT:  # %bb.1: # %cond.load
-; RV64ZVE32F-NEXT:    flw ft0, 0(a0)
-; RV64ZVE32F-NEXT:    vsetivli zero, 1, e32, m1, tu, mu
-; RV64ZVE32F-NEXT:    vfmv.s.f v8, ft0
+; RV64ZVE32F-NEXT:    vsetivli zero, 1, e32, m1, ta, mu
+; RV64ZVE32F-NEXT:    vlse32.v v8, (a0), zero
 ; RV64ZVE32F-NEXT:  .LBB68_2: # %else
 ; RV64ZVE32F-NEXT:    ret
   %v = call <1 x float> @llvm.masked.gather.v1f32.v1p0f32(<1 x float*> %ptrs, i32 4, <1 x i1> %m, <1 x float> %passthru)
diff --git a/llvm/test/CodeGen/X86/avx512-insert-extract.ll b/llvm/test/CodeGen/X86/avx512-insert-extract.ll
--- a/llvm/test/CodeGen/X86/avx512-insert-extract.ll
+++ b/llvm/test/CodeGen/X86/avx512-insert-extract.ll
@@ -593,8 +593,9 @@
 define <2 x i64> @insert_v2i64(<2 x i64> %x, i64 %y , i64* %ptr) {
 ; CHECK-LABEL: insert_v2i64:
 ; CHECK:       ## %bb.0:
-; CHECK-NEXT:    vpinsrq $0, %rdi, %xmm0, %xmm0
-; CHECK-NEXT:    vpinsrq $1, (%rsi), %xmm0, %xmm0
+; CHECK-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
+; CHECK-NEXT:    vmovq %rdi, %xmm1
+; CHECK-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; CHECK-NEXT:    retq
   %val = load i64, i64* %ptr
   %r1 = insertelement <2 x i64> %x, i64 %val, i32 1
diff --git a/llvm/test/CodeGen/X86/masked_gather_scatter.ll b/llvm/test/CodeGen/X86/masked_gather_scatter.ll
--- a/llvm/test/CodeGen/X86/masked_gather_scatter.ll
+++ b/llvm/test/CodeGen/X86/masked_gather_scatter.ll
@@ -2284,38 +2284,40 @@
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vpmovsxdq %xmm0, %xmm0
 ; KNL_64-NEXT:    vpsllq $3, %xmm0, %xmm0
-; KNL_64-NEXT:    vmovq %rdi, %xmm2
-; KNL_64-NEXT:    vpbroadcastq %xmm2, %xmm2
-; KNL_64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; KNL_64-NEXT:    vmovq %rdi, %xmm1
+; KNL_64-NEXT:    vpbroadcastq %xmm1, %xmm1
+; KNL_64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
 ; KNL_64-NEXT:    vmovq %xmm0, %rax
-; KNL_64-NEXT:    vpinsrq $0, (%rax), %xmm1, %xmm1
-; KNL_64-NEXT:    vpextrq $1, %xmm0, %rax
-; KNL_64-NEXT:    vpinsrq $1, (%rax), %xmm1, %xmm0
+; KNL_64-NEXT:    vpextrq $1, %xmm0, %rcx
+; KNL_64-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; KNL_64-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; KNL_64-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test26:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    vpslld $3, %xmm0, %xmm0
-; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm2
-; KNL_32-NEXT:    vpaddd %xmm0, %xmm2, %xmm0
+; KNL_32-NEXT:    vpbroadcastd {{[0-9]+}}(%esp), %xmm1
+; KNL_32-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
 ; KNL_32-NEXT:    vmovd %xmm0, %eax
-; KNL_32-NEXT:    vpinsrd $0, (%eax), %xmm1, %xmm1
-; KNL_32-NEXT:    vpinsrd $1, 4(%eax), %xmm1, %xmm1
-; KNL_32-NEXT:    vpextrd $1, %xmm0, %eax
-; KNL_32-NEXT:    vpinsrd $2, (%eax), %xmm1, %xmm0
-; KNL_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm0
+; KNL_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; KNL_32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; KNL_32-NEXT:    vpinsrd $1, 4(%eax), %xmm0, %xmm0
+; KNL_32-NEXT:    vpinsrd $2, (%ecx), %xmm0, %xmm0
+; KNL_32-NEXT:    vpinsrd $3, 4(%ecx), %xmm0, %xmm0
 ; KNL_32-NEXT:    retl
 ;
 ; SKX-LABEL: test26:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpmovsxdq %xmm0, %xmm0
-; SKX-NEXT:    vpbroadcastq %rdi, %xmm2
+; SKX-NEXT:    vpbroadcastq %rdi, %xmm1
 ; SKX-NEXT:    vpsllq $3, %xmm0, %xmm0
-; SKX-NEXT:    vpaddq %xmm0, %xmm2, %xmm0
+; SKX-NEXT:    vpaddq %xmm0, %xmm1, %xmm0
 ; SKX-NEXT:    vmovq %xmm0, %rax
-; SKX-NEXT:    vpinsrq $0, (%rax), %xmm1, %xmm1
-; SKX-NEXT:    vpextrq $1, %xmm0, %rax
-; SKX-NEXT:    vpinsrq $1, (%rax), %xmm1, %xmm0
+; SKX-NEXT:    vpextrq $1, %xmm0, %rcx
+; SKX-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; SKX-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; SKX-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test26:
@@ -2323,11 +2325,11 @@
 ; SKX_32-NEXT:    vpslld $3, %xmm0, %xmm0
 ; SKX_32-NEXT:    vpaddd {{[0-9]+}}(%esp){1to4}, %xmm0, %xmm0
 ; SKX_32-NEXT:    vmovd %xmm0, %eax
-; SKX_32-NEXT:    vpinsrd $0, (%eax), %xmm1, %xmm1
-; SKX_32-NEXT:    vpinsrd $1, 4(%eax), %xmm1, %xmm1
-; SKX_32-NEXT:    vpextrd $1, %xmm0, %eax
-; SKX_32-NEXT:    vpinsrd $2, (%eax), %xmm1, %xmm0
-; SKX_32-NEXT:    vpinsrd $3, 4(%eax), %xmm0, %xmm0
+; SKX_32-NEXT:    vpextrd $1, %xmm0, %ecx
+; SKX_32-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; SKX_32-NEXT:    vpinsrd $1, 4(%eax), %xmm0, %xmm0
+; SKX_32-NEXT:    vpinsrd $2, (%ecx), %xmm0, %xmm0
+; SKX_32-NEXT:    vpinsrd $3, 4(%ecx), %xmm0, %xmm0
 ; SKX_32-NEXT:    retl
   %sext_ind = sext <2 x i32> %ind to <2 x i64>
   %gep.random = getelementptr i64, i64* %base, <2 x i64> %sext_ind
diff --git a/llvm/test/CodeGen/X86/sse-insertelt-from-mem.ll b/llvm/test/CodeGen/X86/sse-insertelt-from-mem.ll
--- a/llvm/test/CodeGen/X86/sse-insertelt-from-mem.ll
+++ b/llvm/test/CodeGen/X86/sse-insertelt-from-mem.ll
@@ -379,26 +379,15 @@
 }
 
 define <2 x i64> @insert_i64_two_elts(<2 x i64> %x, i64* %s.addr) {
-; SSE2-LABEL: insert_i64_two_elts:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    movq (%rdi), %rax
-; SSE2-NEXT:    movq %rax, %xmm0
-; SSE2-NEXT:    movq %rax, %xmm1
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE2-NEXT:    retq
-;
-; SSE41-LABEL: insert_i64_two_elts:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    movq (%rdi), %rax
-; SSE41-NEXT:    pinsrq $0, %rax, %xmm0
-; SSE41-NEXT:    pinsrq $1, %rax, %xmm0
-; SSE41-NEXT:    retq
+; SSE-LABEL: insert_i64_two_elts:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: insert_i64_two_elts:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    movq (%rdi), %rax
-; AVX-NEXT:    vpinsrq $0, %rax, %xmm0, %xmm0
-; AVX-NEXT:    vpinsrq $1, %rax, %xmm0, %xmm0
+; AVX-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX-NEXT:    retq
   %s = load i64, i64* %s.addr
   %i0 = insertelement <2 x i64> %x, i64 %s, i32 0
diff --git a/llvm/test/CodeGen/X86/sse-insertelt.ll b/llvm/test/CodeGen/X86/sse-insertelt.ll
--- a/llvm/test/CodeGen/X86/sse-insertelt.ll
+++ b/llvm/test/CodeGen/X86/sse-insertelt.ll
@@ -352,24 +352,11 @@
 }
 
 define <2 x i64> @insert_i64_two_elts(<2 x i64> %x, i64 %s) {
-; SSE2-LABEL: insert_i64_two_elts:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    movq %rdi, %xmm0
-; SSE2-NEXT:    movq %rdi, %xmm1
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE2-NEXT:    retq
-;
-; SSE41-LABEL: insert_i64_two_elts:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pinsrq $0, %rdi, %xmm0
-; SSE41-NEXT:    pinsrq $1, %rdi, %xmm0
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: insert_i64_two_elts:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpinsrq $0, %rdi, %xmm0, %xmm0
-; AVX-NEXT:    vpinsrq $1, %rdi, %xmm0, %xmm0
-; AVX-NEXT:    retq
+; SSE-LABEL: insert_i64_two_elts:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movq %rdi, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; SSE-NEXT:    retq
   %i0 = insertelement <2 x i64> %x, i64 %s, i32 0
   %i1 = insertelement <2 x i64> %i0, i64 %s, i32 1
   ret <2 x i64> %i1
diff --git a/llvm/test/CodeGen/X86/vec_insert-7.ll b/llvm/test/CodeGen/X86/vec_insert-7.ll
--- a/llvm/test/CodeGen/X86/vec_insert-7.ll
+++ b/llvm/test/CodeGen/X86/vec_insert-7.ll
@@ -8,7 +8,8 @@
 define x86_mmx @mmx_movzl(x86_mmx %x) nounwind {
 ; X86-LABEL: mmx_movzl:
 ; X86:       ## %bb.0:
-; X86-NEXT:    movq {{\.?LCPI[0-9]+_[0-9]+}}, %mm0
+; X86-NEXT:    movl $32, %eax
+; X86-NEXT:    movd %eax, %mm0
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: mmx_movzl: