diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -31103,6 +31103,10 @@
     Mask = ExtendToType(Mask, MaskVT, DAG, true);
   }
 
+  // Break dependency on the data register.
+  if (PassThru.isUndef())
+    PassThru = getZeroVector(VT, Subtarget, DAG, dl);
+
   SDValue Ops[] = { N->getChain(), PassThru, Mask, N->getBasePtr(), Index,
                     N->getScale() };
   SDValue NewGather = DAG.getMemIntrinsicNode(
diff --git a/llvm/test/CodeGen/X86/masked_gather.ll b/llvm/test/CodeGen/X86/masked_gather.ll
--- a/llvm/test/CodeGen/X86/masked_gather.ll
+++ b/llvm/test/CodeGen/X86/masked_gather.ll
@@ -1748,24 +1748,28 @@
 ; AVX512F-NEXT:    kshiftlw $8, %k0, %k0
 ; AVX512F-NEXT:    kshiftrw $8, %k0, %k1
 ; AVX512F-NEXT:    vpbroadcastd {{.*#+}} zmm0 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX512F-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512F-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX512F-NEXT:    kmovw %k1, %k2
-; AVX512F-NEXT:    vpgatherdd c(,%zmm0), %zmm1 {%k2}
+; AVX512F-NEXT:    vpgatherdd c(,%zmm0), %zmm2 {%k2}
 ; AVX512F-NEXT:    vpbroadcastd {{.*#+}} zmm0 = [28,28,28,28,28,28,28,28,28,28,28,28,28,28,28,28]
-; AVX512F-NEXT:    vpgatherdd c(,%zmm0), %zmm2 {%k1}
-; AVX512F-NEXT:    vpaddd %ymm2, %ymm2, %ymm0
-; AVX512F-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpgatherdd c(,%zmm0), %zmm1 {%k1}
+; AVX512F-NEXT:    vpaddd %ymm1, %ymm1, %ymm0
+; AVX512F-NEXT:    vpaddd %ymm0, %ymm2, %ymm0
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512VL-LABEL: gather_v8i32_v8i32:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vptestnmd %ymm0, %ymm0, %k1
-; AVX512VL-NEXT:    vpbroadcastd {{.*#+}} ymm0 = [12,12,12,12,12,12,12,12]
+; AVX512VL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
+; AVX512VL-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12]
 ; AVX512VL-NEXT:    kmovw %k1, %k2
-; AVX512VL-NEXT:    vpgatherdd c(,%ymm0), %ymm1 {%k2}
-; AVX512VL-NEXT:    vpbroadcastd {{.*#+}} ymm0 = [28,28,28,28,28,28,28,28]
-; AVX512VL-NEXT:    vpgatherdd c(,%ymm0), %ymm2 {%k1}
-; AVX512VL-NEXT:    vpaddd %ymm2, %ymm2, %ymm0
-; AVX512VL-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
+; AVX512VL-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512VL-NEXT:    vpgatherdd c(,%ymm1), %ymm2 {%k2}
+; AVX512VL-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [28,28,28,28,28,28,28,28]
+; AVX512VL-NEXT:    vpgatherdd c(,%ymm1), %ymm0 {%k1}
+; AVX512VL-NEXT:    vpaddd %ymm0, %ymm0, %ymm0
+; AVX512VL-NEXT:    vpaddd %ymm0, %ymm2, %ymm0
 ; AVX512VL-NEXT:    retq
   %1 = icmp eq <8 x i32> %trigger, zeroinitializer
   %2 = call <8 x i32> @llvm.masked.gather.v8i32.v8p0i32(<8 x i32*> getelementptr (%struct.a, <8 x %struct.a*> <%struct.a* @c, %struct.a* @c, %struct.a* @c, %struct.a* @c, %struct.a* @c, %struct.a* @c, %struct.a* @c, %struct.a* @c>, <8 x i64> zeroinitializer, i32 0, <8 x i64> <i64 3, i64 3, i64 3, i64 3, i64 3, i64 3, i64 3, i64 3>), i32 4, <8 x i1> %1, <8 x i32> undef)
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
@@ -21,6 +21,7 @@
 ; KNL_64-LABEL: test1:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; KNL_64-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_64-NEXT:    retq
@@ -29,6 +30,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; KNL_32-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_32-NEXT:    retl
@@ -36,6 +38,7 @@
 ; SKX-LABEL: test1:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; SKX-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX-NEXT:    retq
@@ -44,6 +47,7 @@
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; SKX_32-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX_32-NEXT:    retl
@@ -78,6 +82,7 @@
 ; KNL_64-LABEL: test2:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kmovw %esi, %k1
+; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; KNL_64-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_64-NEXT:    retq
@@ -86,6 +91,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; KNL_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; KNL_32-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_32-NEXT:    retl
@@ -93,6 +99,7 @@
 ; SKX-LABEL: test2:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovw %esi, %k1
+; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; SKX-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX-NEXT:    retq
@@ -101,6 +108,7 @@
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; SKX_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; SKX_32-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX_32-NEXT:    retl
@@ -119,6 +127,7 @@
 ; KNL_64-LABEL: test3:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kmovw %esi, %k1
+; KNL_64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm1 {%k1}
 ; KNL_64-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; KNL_64-NEXT:    retq
@@ -127,6 +136,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; KNL_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm1 {%k1}
 ; KNL_32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; KNL_32-NEXT:    retl
@@ -134,6 +144,7 @@
 ; SKX-LABEL: test3:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovw %esi, %k1
+; SKX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm1 {%k1}
 ; SKX-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; SKX-NEXT:    retq
@@ -142,6 +153,7 @@
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm1 {%k1}
 ; SKX_32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; SKX_32-NEXT:    retl
@@ -161,6 +173,7 @@
 ; KNL_64-LABEL: test4:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kmovw %esi, %k1
+; KNL_64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    kmovw %k1, %k2
 ; KNL_64-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm1 {%k2}
 ; KNL_64-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -172,6 +185,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; KNL_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    kmovw %k1, %k2
 ; KNL_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm1 {%k2}
 ; KNL_32-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -182,6 +196,7 @@
 ; SKX-LABEL: test4:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovw %esi, %k1
+; SKX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    kmovw %k1, %k2
 ; SKX-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm1 {%k2}
 ; SKX-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -193,6 +208,7 @@
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    kmovw %k1, %k2
 ; SKX_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm1 {%k2}
 ; SKX_32-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -292,6 +308,7 @@
 ; KNL_64-LABEL: test6:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k2
 ; KNL_64-NEXT:    vpgatherqd (,%zmm1), %ymm2 {%k2}
 ; KNL_64-NEXT:    vpscatterqd %ymm0, (,%zmm1) {%k1}
@@ -302,6 +319,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    # kill: def $ymm1 killed $ymm1 def $zmm1
 ; KNL_32-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; KNL_32-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; KNL_32-NEXT:    movw $255, %ax
 ; KNL_32-NEXT:    kmovw %eax, %k1
 ; KNL_32-NEXT:    kmovw %k1, %k2
@@ -313,6 +331,7 @@
 ; SKX-LABEL: test6:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    kxnorw %k0, %k0, %k2
 ; SKX-NEXT:    vpgatherqd (,%zmm1), %ymm2 {%k2}
 ; SKX-NEXT:    vpscatterqd %ymm0, (,%zmm1) {%k1}
@@ -322,6 +341,7 @@
 ; SKX_32-LABEL: test6:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k2
 ; SKX_32-NEXT:    vpgatherdd (,%ymm1), %ymm2 {%k2}
 ; SKX_32-NEXT:    vpscatterdd %ymm0, (,%ymm1) {%k1}
@@ -342,6 +362,7 @@
 ; KNL_64-NEXT:    kmovw %esi, %k0
 ; KNL_64-NEXT:    kshiftlw $8, %k0, %k0
 ; KNL_64-NEXT:    kshiftrw $8, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    kmovw %k1, %k2
 ; KNL_64-NEXT:    vpgatherdd (%rdi,%zmm0,4), %zmm1 {%k2}
 ; KNL_64-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -357,6 +378,7 @@
 ; KNL_32-NEXT:    kmovw %ecx, %k0
 ; KNL_32-NEXT:    kshiftlw $8, %k0, %k0
 ; KNL_32-NEXT:    kshiftrw $8, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    kmovw %k1, %k2
 ; KNL_32-NEXT:    vpgatherdd (%eax,%zmm0,4), %zmm1 {%k2}
 ; KNL_32-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -367,6 +389,7 @@
 ; SKX-LABEL: test7:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovw %esi, %k1
+; SKX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    kmovw %k1, %k2
 ; SKX-NEXT:    vpgatherdd (%rdi,%ymm0,4), %ymm1 {%k2}
 ; SKX-NEXT:    vmovdqa %ymm1, %ymm2
@@ -378,6 +401,7 @@
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    kmovb {{[0-9]+}}(%esp), %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    kmovw %k1, %k2
 ; SKX_32-NEXT:    vpgatherdd (%eax,%ymm0,4), %ymm1 {%k2}
 ; SKX_32-NEXT:    vmovdqa %ymm1, %ymm2
@@ -403,20 +427,23 @@
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kmovw %edi, %k1
 ; KNL_64-NEXT:    kshiftrw $8, %k1, %k2
+; KNL_64-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; KNL_64-NEXT:    kmovw %k2, %k3
-; KNL_64-NEXT:    vpgatherqd (,%zmm1), %ymm2 {%k3}
+; KNL_64-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; KNL_64-NEXT:    vpgatherqd (,%zmm1), %ymm3 {%k3}
 ; KNL_64-NEXT:    kmovw %k1, %k3
-; KNL_64-NEXT:    vpgatherqd (,%zmm0), %ymm3 {%k3}
-; KNL_64-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm4
-; KNL_64-NEXT:    vpgatherqd (,%zmm1), %ymm2 {%k2}
-; KNL_64-NEXT:    vpgatherqd (,%zmm0), %ymm3 {%k1}
-; KNL_64-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm0
+; KNL_64-NEXT:    vpgatherqd (,%zmm0), %ymm2 {%k3}
+; KNL_64-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm4
+; KNL_64-NEXT:    vpgatherqd (,%zmm1), %ymm3 {%k2}
+; KNL_64-NEXT:    vpgatherqd (,%zmm0), %ymm2 {%k1}
+; KNL_64-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm0
 ; KNL_64-NEXT:    vpaddd %zmm0, %zmm4, %zmm0
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test8:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; KNL_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    kmovw %k1, %k2
 ; KNL_32-NEXT:    vpgatherdd (,%zmm0), %zmm1 {%k2}
 ; KNL_32-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -428,20 +455,23 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kmovw %edi, %k1
 ; SKX-NEXT:    kshiftrw $8, %k1, %k2
+; SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; SKX-NEXT:    kmovw %k2, %k3
-; SKX-NEXT:    vpgatherqd (,%zmm1), %ymm2 {%k3}
+; SKX-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; SKX-NEXT:    vpgatherqd (,%zmm1), %ymm3 {%k3}
 ; SKX-NEXT:    kmovw %k1, %k3
-; SKX-NEXT:    vpgatherqd (,%zmm0), %ymm3 {%k3}
-; SKX-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm4
-; SKX-NEXT:    vpgatherqd (,%zmm1), %ymm2 {%k2}
-; SKX-NEXT:    vpgatherqd (,%zmm0), %ymm3 {%k1}
-; SKX-NEXT:    vinserti64x4 $1, %ymm2, %zmm3, %zmm0
+; SKX-NEXT:    vpgatherqd (,%zmm0), %ymm2 {%k3}
+; SKX-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm4
+; SKX-NEXT:    vpgatherqd (,%zmm1), %ymm3 {%k2}
+; SKX-NEXT:    vpgatherqd (,%zmm0), %ymm2 {%k1}
+; SKX-NEXT:    vinserti64x4 $1, %ymm3, %zmm2, %zmm0
 ; SKX-NEXT:    vpaddd %zmm0, %zmm4, %zmm0
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test8:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    kmovw %k1, %k2
 ; SKX_32-NEXT:    vpgatherdd (,%zmm0), %zmm1 {%k2}
 ; SKX_32-NEXT:    vmovdqa64 %zmm1, %zmm2
@@ -479,6 +509,7 @@
 ; KNL_64-NEXT:    vpaddq %zmm0, %zmm4, %zmm0
 ; KNL_64-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm1
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    vpgatherqd (,%zmm1), %ymm0 {%k1}
 ; KNL_64-NEXT:    retq
 ;
@@ -494,6 +525,7 @@
 ; KNL_32-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [68,68,68,68,68,68,68,68]
 ; KNL_32-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; KNL_32-NEXT:    vpaddd %ymm0, %ymm2, %ymm1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    movw $255, %ax
 ; KNL_32-NEXT:    kmovw %eax, %k1
 ; KNL_32-NEXT:    vpgatherdd (,%zmm1), %zmm0 {%k1}
@@ -510,6 +542,7 @@
 ; SKX_SMALL-NEXT:    vpaddq %zmm0, %zmm2, %zmm0
 ; SKX_SMALL-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm1
 ; SKX_SMALL-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_SMALL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_SMALL-NEXT:    vpgatherqd (,%zmm1), %ymm0 {%k1}
 ; SKX_SMALL-NEXT:    retq
 ;
@@ -526,6 +559,7 @@
 ; SKX_LARGE-NEXT:    movabsq ${{\.?LCPI[0-9]+_[0-9]+}}, %rax
 ; SKX_LARGE-NEXT:    vpaddq (%rax){1to8}, %zmm0, %zmm1
 ; SKX_LARGE-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_LARGE-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_LARGE-NEXT:    vpgatherqd (,%zmm1), %ymm0 {%k1}
 ; SKX_LARGE-NEXT:    retq
 ;
@@ -538,6 +572,7 @@
 ; SKX_32-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; SKX_32-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %ymm0, %ymm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vpgatherdd (,%ymm1), %ymm0 {%k1}
 ; SKX_32-NEXT:    retl
 entry:
@@ -565,6 +600,7 @@
 ; KNL_64-NEXT:    vpaddq %zmm0, %zmm4, %zmm0
 ; KNL_64-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm1
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    vpgatherqd (,%zmm1), %ymm0 {%k1}
 ; KNL_64-NEXT:    retq
 ;
@@ -580,6 +616,7 @@
 ; KNL_32-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [68,68,68,68,68,68,68,68]
 ; KNL_32-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; KNL_32-NEXT:    vpaddd %ymm0, %ymm2, %ymm1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    movw $255, %ax
 ; KNL_32-NEXT:    kmovw %eax, %k1
 ; KNL_32-NEXT:    vpgatherdd (,%zmm1), %zmm0 {%k1}
@@ -596,6 +633,7 @@
 ; SKX_SMALL-NEXT:    vpaddq %zmm0, %zmm2, %zmm0
 ; SKX_SMALL-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm1
 ; SKX_SMALL-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_SMALL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_SMALL-NEXT:    vpgatherqd (,%zmm1), %ymm0 {%k1}
 ; SKX_SMALL-NEXT:    retq
 ;
@@ -612,6 +650,7 @@
 ; SKX_LARGE-NEXT:    movabsq ${{\.?LCPI[0-9]+_[0-9]+}}, %rax
 ; SKX_LARGE-NEXT:    vpaddq (%rax){1to8}, %zmm0, %zmm1
 ; SKX_LARGE-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_LARGE-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_LARGE-NEXT:    vpgatherqd (,%zmm1), %ymm0 {%k1}
 ; SKX_LARGE-NEXT:    retq
 ;
@@ -624,6 +663,7 @@
 ; SKX_32-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
 ; SKX_32-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %ymm0, %ymm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vpgatherdd (,%ymm1), %ymm0 {%k1}
 ; SKX_32-NEXT:    retl
 entry:
@@ -643,6 +683,7 @@
 ; KNL_64-NEXT:    leaq (%rdi,%rax,4), %rax
 ; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    vgatherdps (%rax,%zmm1,4), %zmm0 {%k1}
 ; KNL_64-NEXT:    retq
 ;
@@ -653,6 +694,7 @@
 ; KNL_32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; KNL_32-NEXT:    retl
 ;
@@ -662,6 +704,7 @@
 ; SKX-NEXT:    leaq (%rdi,%rax,4), %rax
 ; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; SKX-NEXT:    vgatherdps (%rax,%zmm1,4), %zmm0 {%k1}
 ; SKX-NEXT:    retq
 ;
@@ -672,6 +715,7 @@
 ; SKX_32-NEXT:    addl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; SKX_32-NEXT:    retl
 
@@ -689,6 +733,7 @@
 ; KNL_64-LABEL: test12:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; KNL_64-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_64-NEXT:    retq
@@ -697,6 +742,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; KNL_32-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_32-NEXT:    retl
@@ -704,6 +750,7 @@
 ; SKX-LABEL: test12:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; SKX-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX-NEXT:    retq
@@ -712,6 +759,7 @@
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; SKX_32-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX_32-NEXT:    retl
@@ -728,6 +776,7 @@
 ; KNL_64-LABEL: test13:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; KNL_64-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_64-NEXT:    retq
@@ -736,6 +785,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; KNL_32-NEXT:    vmovaps %zmm1, %zmm0
 ; KNL_32-NEXT:    retl
@@ -743,6 +793,7 @@
 ; SKX-LABEL: test13:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; SKX-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX-NEXT:    retq
@@ -751,6 +802,7 @@
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; SKX_32-NEXT:    vmovaps %zmm1, %zmm0
 ; SKX_32-NEXT:    retl
@@ -772,6 +824,7 @@
 ; KNL_64-NEXT:    vpmovsxdq %ymm0, %zmm0
 ; KNL_64-NEXT:    vpsllq $2, %zmm0, %zmm0
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vgatherqps (%rax,%zmm0), %ymm1 {%k1}
 ; KNL_64-NEXT:    vinsertf64x4 $1, %ymm1, %zmm1, %zmm0
 ; KNL_64-NEXT:    retq
@@ -781,6 +834,7 @@
 ; KNL_32-NEXT:    vmovd %xmm0, %eax
 ; KNL_32-NEXT:    vpslld $2, {{[0-9]+}}(%esp){1to16}, %zmm1
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm1), %zmm0 {%k1}
 ; KNL_32-NEXT:    retl
 ;
@@ -791,6 +845,7 @@
 ; SKX-NEXT:    vpmovsxdq %ymm0, %zmm0
 ; SKX-NEXT:    vpsllq $2, %zmm0, %zmm0
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vgatherqps (%rax,%zmm0), %ymm1 {%k1}
 ; SKX-NEXT:    vinsertf64x4 $1, %ymm1, %zmm1, %zmm0
 ; SKX-NEXT:    retq
@@ -800,6 +855,7 @@
 ; SKX_32-NEXT:    vmovd %xmm0, %eax
 ; SKX_32-NEXT:    vpslld $2, {{[0-9]+}}(%esp){1to16}, %zmm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm1), %zmm0 {%k1}
 ; SKX_32-NEXT:    retl
 
@@ -909,6 +965,7 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX-NEXT:    vpmovd2m %xmm1, %k1
+; SKX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vgatherdps (%rdi,%xmm0,4), %xmm1 {%k1}
 ; SKX-NEXT:    vmovaps %xmm1, %xmm0
 ; SKX-NEXT:    retq
@@ -918,6 +975,7 @@
 ; SKX_32-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX_32-NEXT:    vpmovd2m %xmm1, %k1
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vgatherdps (%eax,%xmm0,4), %xmm1 {%k1}
 ; SKX_32-NEXT:    vmovaps %xmm1, %xmm0
 ; SKX_32-NEXT:    retl
@@ -2393,6 +2451,7 @@
 define <16 x float> @test29(float* %base, <16 x i32> %ind) {
 ; KNL_64-LABEL: test29:
 ; KNL_64:       # %bb.0:
+; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    movw $44, %ax
 ; KNL_64-NEXT:    kmovw %eax, %k1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
@@ -2402,6 +2461,7 @@
 ; KNL_32-LABEL: test29:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; KNL_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    movw $44, %cx
 ; KNL_32-NEXT:    kmovw %ecx, %k1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
@@ -2410,6 +2470,7 @@
 ;
 ; SKX-LABEL: test29:
 ; SKX:       # %bb.0:
+; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    movw $44, %ax
 ; SKX-NEXT:    kmovw %eax, %k1
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
@@ -2419,6 +2480,7 @@
 ; SKX_32-LABEL: test29:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; SKX_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    movw $44, %cx
 ; SKX_32-NEXT:    kmovw %ecx, %k1
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
@@ -2682,16 +2744,19 @@
 ; KNL_64-LABEL: test31:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; KNL_64-NEXT:    vpxor %xmm3, %xmm3, %xmm3
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k2
-; KNL_64-NEXT:    vpgatherqq (,%zmm0), %zmm2 {%k2}
-; KNL_64-NEXT:    vpgatherqq (,%zmm1), %zmm3 {%k1}
-; KNL_64-NEXT:    vmovdqa64 %zmm2, %zmm0
-; KNL_64-NEXT:    vmovdqa64 %zmm3, %zmm1
+; KNL_64-NEXT:    vpgatherqq (,%zmm0), %zmm3 {%k2}
+; KNL_64-NEXT:    vpgatherqq (,%zmm1), %zmm2 {%k1}
+; KNL_64-NEXT:    vmovdqa64 %zmm3, %zmm0
+; KNL_64-NEXT:    vmovdqa64 %zmm2, %zmm1
 ; KNL_64-NEXT:    retq
 ;
 ; KNL_32-LABEL: test31:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vpgatherdd (,%zmm0), %zmm1 {%k1}
 ; KNL_32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; KNL_32-NEXT:    retl
@@ -2699,16 +2764,19 @@
 ; SKX-LABEL: test31:
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; SKX-NEXT:    vpxor %xmm3, %xmm3, %xmm3
 ; SKX-NEXT:    kxnorw %k0, %k0, %k2
-; SKX-NEXT:    vpgatherqq (,%zmm0), %zmm2 {%k2}
-; SKX-NEXT:    vpgatherqq (,%zmm1), %zmm3 {%k1}
-; SKX-NEXT:    vmovdqa64 %zmm2, %zmm0
-; SKX-NEXT:    vmovdqa64 %zmm3, %zmm1
+; SKX-NEXT:    vpgatherqq (,%zmm0), %zmm3 {%k2}
+; SKX-NEXT:    vpgatherqq (,%zmm1), %zmm2 {%k1}
+; SKX-NEXT:    vmovdqa64 %zmm3, %zmm0
+; SKX-NEXT:    vmovdqa64 %zmm2, %zmm1
 ; SKX-NEXT:    retq
 ;
 ; SKX_32-LABEL: test31:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vpgatherdd (,%zmm0), %zmm1 {%k1}
 ; SKX_32-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; SKX_32-NEXT:    retl
@@ -3439,6 +3507,7 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX-NEXT:    vpmovd2m %xmm1, %k1
+; SKX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vpgatherqq (,%ymm0), %ymm1 {%k1}
 ; SKX-NEXT:    vpaddq %ymm1, %ymm1, %ymm0
 ; SKX-NEXT:    vpaddq %ymm0, %ymm1, %ymm0
@@ -3455,6 +3524,7 @@
 ; SKX_32-NEXT:    subl $32, %esp
 ; SKX_32-NEXT:    vpslld $31, %xmm1, %xmm1
 ; SKX_32-NEXT:    vpmovd2m %xmm1, %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vpgatherdq (,%xmm0), %ymm1 {%k1}
 ; SKX_32-NEXT:    vpaddq %ymm1, %ymm1, %ymm0
 ; SKX_32-NEXT:    vpaddq %ymm0, %ymm1, %ymm0
@@ -3475,6 +3545,7 @@
 ; KNL_64-LABEL: test_global_array:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; KNL_64-NEXT:    vmovdqa %ymm1, %ymm0
 ; KNL_64-NEXT:    retq
@@ -3482,6 +3553,7 @@
 ; KNL_32-LABEL: test_global_array:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; KNL_32-NEXT:    vmovdqa %ymm1, %ymm0
 ; KNL_32-NEXT:    retl
@@ -3489,6 +3561,7 @@
 ; SKX_SMALL-LABEL: test_global_array:
 ; SKX_SMALL:       # %bb.0:
 ; SKX_SMALL-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_SMALL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_SMALL-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; SKX_SMALL-NEXT:    vmovdqa %ymm1, %ymm0
 ; SKX_SMALL-NEXT:    retq
@@ -3497,6 +3570,7 @@
 ; SKX_LARGE:       # %bb.0:
 ; SKX_LARGE-NEXT:    movabsq $glob_array, %rax
 ; SKX_LARGE-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_LARGE-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_LARGE-NEXT:    vpgatherqd (%rax,%zmm0,4), %ymm1 {%k1}
 ; SKX_LARGE-NEXT:    vmovdqa %ymm1, %ymm0
 ; SKX_LARGE-NEXT:    retq
@@ -3504,6 +3578,7 @@
 ; SKX_32-LABEL: test_global_array:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; SKX_32-NEXT:    vmovdqa %ymm1, %ymm0
 ; SKX_32-NEXT:    retl
@@ -3516,6 +3591,7 @@
 ; KNL_64-LABEL: test_global_array_zeroinitializer_index:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; KNL_64-NEXT:    vmovdqa %ymm1, %ymm0
 ; KNL_64-NEXT:    retq
@@ -3523,6 +3599,7 @@
 ; KNL_32-LABEL: test_global_array_zeroinitializer_index:
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; KNL_32-NEXT:    vmovdqa %ymm1, %ymm0
 ; KNL_32-NEXT:    retl
@@ -3530,6 +3607,7 @@
 ; SKX_SMALL-LABEL: test_global_array_zeroinitializer_index:
 ; SKX_SMALL:       # %bb.0:
 ; SKX_SMALL-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_SMALL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_SMALL-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; SKX_SMALL-NEXT:    vmovdqa %ymm1, %ymm0
 ; SKX_SMALL-NEXT:    retq
@@ -3538,6 +3616,7 @@
 ; SKX_LARGE:       # %bb.0:
 ; SKX_LARGE-NEXT:    movabsq $glob_array, %rax
 ; SKX_LARGE-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_LARGE-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_LARGE-NEXT:    vpgatherqd (%rax,%zmm0,4), %ymm1 {%k1}
 ; SKX_LARGE-NEXT:    vmovdqa %ymm1, %ymm0
 ; SKX_LARGE-NEXT:    retq
@@ -3545,6 +3624,7 @@
 ; SKX_32-LABEL: test_global_array_zeroinitializer_index:
 ; SKX_32:       # %bb.0:
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vpgatherqd glob_array(,%zmm0,4), %ymm1 {%k1}
 ; SKX_32-NEXT:    vmovdqa %ymm1, %ymm0
 ; SKX_32-NEXT:    retl
@@ -3764,6 +3844,7 @@
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vpmovsxbd %xmm0, %zmm1
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
 ; KNL_64-NEXT:    retq
 ;
@@ -3772,6 +3853,7 @@
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    vpmovsxbd %xmm0, %zmm1
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; KNL_32-NEXT:    retl
 ;
@@ -3779,6 +3861,7 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpmovsxbd %xmm0, %zmm1
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
 ; SKX-NEXT:    retq
 ;
@@ -3787,6 +3870,7 @@
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    vpmovsxbd %xmm0, %zmm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; SKX_32-NEXT:    retl
 
@@ -3802,6 +3886,7 @@
 ; KNL_64-LABEL: sext_v8i8_index:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vpmovsxbd %xmm0, %ymm1
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    movw $255, %ax
 ; KNL_64-NEXT:    kmovw %eax, %k1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
@@ -3812,6 +3897,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    vpmovsxbd %xmm0, %ymm1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    movw $255, %cx
 ; KNL_32-NEXT:    kmovw %ecx, %k1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
@@ -3822,6 +3908,7 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpmovsxbd %xmm0, %ymm1
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX-NEXT:    vgatherdps (%rdi,%ymm1,4), %ymm0 {%k1}
 ; SKX-NEXT:    retq
 ;
@@ -3830,6 +3917,7 @@
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    vpmovsxbd %xmm0, %ymm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vgatherdps (%eax,%ymm1,4), %ymm0 {%k1}
 ; SKX_32-NEXT:    retl
 
@@ -3847,6 +3935,7 @@
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
 ; KNL_64-NEXT:    retq
 ;
@@ -3855,6 +3944,7 @@
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; KNL_32-NEXT:    retl
 ;
@@ -3862,6 +3952,7 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
 ; SKX-NEXT:    retq
 ;
@@ -3870,6 +3961,7 @@
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    vpmovzxbd {{.*#+}} zmm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; SKX_32-NEXT:    retl
 
@@ -3885,6 +3977,7 @@
 ; KNL_64-LABEL: zext_v8i8_index:
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vpmovzxbd {{.*#+}} ymm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    movw $255, %ax
 ; KNL_64-NEXT:    kmovw %eax, %k1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
@@ -3895,6 +3988,7 @@
 ; KNL_32:       # %bb.0:
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    vpmovzxbd {{.*#+}} ymm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    movw $255, %cx
 ; KNL_32-NEXT:    kmovw %ecx, %k1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
@@ -3905,6 +3999,7 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vpmovzxbd {{.*#+}} ymm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX-NEXT:    vgatherdps (%rdi,%ymm1,4), %ymm0 {%k1}
 ; SKX-NEXT:    retq
 ;
@@ -3913,6 +4008,7 @@
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    vpmovzxbd {{.*#+}} ymm1 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vgatherdps (%eax,%ymm1,4), %ymm0 {%k1}
 ; SKX_32-NEXT:    retl
 
@@ -4041,6 +4137,7 @@
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm1
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
 ; KNL_64-NEXT:    retq
 ;
@@ -4049,6 +4146,7 @@
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; KNL_32-NEXT:    vpandd {{\.?LCPI[0-9]+_[0-9]+}}{1to16}, %zmm0, %zmm1
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; KNL_32-NEXT:    retl
 ;
@@ -4056,6 +4154,7 @@
 ; SKX_SMALL:       # %bb.0:
 ; SKX_SMALL-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm1
 ; SKX_SMALL-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_SMALL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; SKX_SMALL-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
 ; SKX_SMALL-NEXT:    retq
 ;
@@ -4064,6 +4163,7 @@
 ; SKX_LARGE-NEXT:    movabsq ${{\.?LCPI[0-9]+_[0-9]+}}, %rax
 ; SKX_LARGE-NEXT:    vandps (%rax){1to16}, %zmm0, %zmm1
 ; SKX_LARGE-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_LARGE-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; SKX_LARGE-NEXT:    vgatherdps (%rdi,%zmm1,4), %zmm0 {%k1}
 ; SKX_LARGE-NEXT:    retq
 ;
@@ -4072,6 +4172,7 @@
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; SKX_32-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to16}, %zmm0, %zmm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm1,4), %zmm0 {%k1}
 ; SKX_32-NEXT:    retl
   %ind_masked = and <16 x i32> %ind, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
@@ -4244,6 +4345,7 @@
 ; KNL_64:       # %bb.0:
 ; KNL_64-NEXT:    vmovaps %zmm0, (%rsi)
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_64-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; KNL_64-NEXT:    vaddps %zmm1, %zmm1, %zmm0
 ; KNL_64-NEXT:    retq
@@ -4254,6 +4356,7 @@
 ; KNL_32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; KNL_32-NEXT:    vmovaps %zmm0, (%ecx)
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; KNL_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; KNL_32-NEXT:    vaddps %zmm1, %zmm1, %zmm0
 ; KNL_32-NEXT:    retl
@@ -4262,6 +4365,7 @@
 ; SKX:       # %bb.0:
 ; SKX-NEXT:    vmovaps %zmm0, (%rsi)
 ; SKX-NEXT:    kxnorw %k0, %k0, %k1
+; SKX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX-NEXT:    vgatherdps (%rdi,%zmm0,4), %zmm1 {%k1}
 ; SKX-NEXT:    vaddps %zmm1, %zmm1, %zmm0
 ; SKX-NEXT:    retq
@@ -4272,6 +4376,7 @@
 ; SKX_32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; SKX_32-NEXT:    vmovaps %zmm0, (%ecx)
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; SKX_32-NEXT:    vgatherdps (%eax,%zmm0,4), %zmm1 {%k1}
 ; SKX_32-NEXT:    vaddps %zmm1, %zmm1, %zmm0
 ; SKX_32-NEXT:    retl
@@ -5127,6 +5232,7 @@
 ; KNL_64:       # %bb.0: # %bb
 ; KNL_64-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm1
 ; KNL_64-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_64-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_64-NEXT:    vpgatherqq (,%zmm1), %zmm0 {%k1}
 ; KNL_64-NEXT:    retq
 ;
@@ -5135,6 +5241,7 @@
 ; KNL_32-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
 ; KNL_32-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
 ; KNL_32-NEXT:    kxnorw %k0, %k0, %k1
+; KNL_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; KNL_32-NEXT:    vpgatherdq (,%ymm1), %zmm0 {%k1}
 ; KNL_32-NEXT:    retl
 ;
@@ -5142,6 +5249,7 @@
 ; SKX_SMALL:       # %bb.0: # %bb
 ; SKX_SMALL-NEXT:    vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm1
 ; SKX_SMALL-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_SMALL-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_SMALL-NEXT:    vpgatherqq (,%zmm1), %zmm0 {%k1}
 ; SKX_SMALL-NEXT:    retq
 ;
@@ -5150,6 +5258,7 @@
 ; SKX_LARGE-NEXT:    movabsq ${{\.?LCPI[0-9]+_[0-9]+}}, %rax
 ; SKX_LARGE-NEXT:    vpaddq (%rax){1to8}, %zmm0, %zmm1
 ; SKX_LARGE-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_LARGE-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_LARGE-NEXT:    vpgatherqq (,%zmm1), %zmm0 {%k1}
 ; SKX_LARGE-NEXT:    retq
 ;
@@ -5157,6 +5266,7 @@
 ; SKX_32:       # %bb.0: # %bb
 ; SKX_32-NEXT:    vpaddd {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %ymm0, %ymm1
 ; SKX_32-NEXT:    kxnorw %k0, %k0, %k1
+; SKX_32-NEXT:    vpxor %xmm0, %xmm0, %xmm0
 ; SKX_32-NEXT:    vpgatherdq (,%ymm1), %zmm0 {%k1}
 ; SKX_32-NEXT:    retl
 bb:
diff --git a/llvm/test/CodeGen/X86/masked_gather_scatter_widen.ll b/llvm/test/CodeGen/X86/masked_gather_scatter_widen.ll
--- a/llvm/test/CodeGen/X86/masked_gather_scatter_widen.ll
+++ b/llvm/test/CodeGen/X86/masked_gather_scatter_widen.ll
@@ -664,12 +664,14 @@
 ; WIDEN_SKX-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; WIDEN_SKX-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; WIDEN_SKX-NEXT:    kxnorw %k0, %k0, %k1
-; WIDEN_SKX-NEXT:    vgatherdps (%rsi,%zmm0,4), %zmm2 {%k1}
+; WIDEN_SKX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; WIDEN_SKX-NEXT:    vxorps %xmm3, %xmm3, %xmm3
+; WIDEN_SKX-NEXT:    vgatherdps (%rsi,%zmm0,4), %zmm3 {%k1}
 ; WIDEN_SKX-NEXT:    movw $1, %cx
 ; WIDEN_SKX-NEXT:    kmovw %ecx, %k1
-; WIDEN_SKX-NEXT:    vgatherdps (%rsi,%zmm1,4), %zmm0 {%k1}
-; WIDEN_SKX-NEXT:    vmovss %xmm0, 64(%rdi)
-; WIDEN_SKX-NEXT:    vmovaps %zmm2, (%rdi)
+; WIDEN_SKX-NEXT:    vgatherdps (%rsi,%zmm1,4), %zmm2 {%k1}
+; WIDEN_SKX-NEXT:    vmovss %xmm2, 64(%rdi)
+; WIDEN_SKX-NEXT:    vmovaps %zmm3, (%rdi)
 ; WIDEN_SKX-NEXT:    vzeroupper
 ; WIDEN_SKX-NEXT:    retq
 ;
@@ -697,12 +699,14 @@
 ; WIDEN_KNL-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
 ; WIDEN_KNL-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; WIDEN_KNL-NEXT:    kxnorw %k0, %k0, %k1
-; WIDEN_KNL-NEXT:    vgatherdps (%rsi,%zmm0,4), %zmm2 {%k1}
+; WIDEN_KNL-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; WIDEN_KNL-NEXT:    vxorps %xmm3, %xmm3, %xmm3
+; WIDEN_KNL-NEXT:    vgatherdps (%rsi,%zmm0,4), %zmm3 {%k1}
 ; WIDEN_KNL-NEXT:    movw $1, %cx
 ; WIDEN_KNL-NEXT:    kmovw %ecx, %k1
-; WIDEN_KNL-NEXT:    vgatherdps (%rsi,%zmm1,4), %zmm0 {%k1}
-; WIDEN_KNL-NEXT:    vmovss %xmm0, 64(%rdi)
-; WIDEN_KNL-NEXT:    vmovaps %zmm2, (%rdi)
+; WIDEN_KNL-NEXT:    vgatherdps (%rsi,%zmm1,4), %zmm2 {%k1}
+; WIDEN_KNL-NEXT:    vmovss %xmm2, 64(%rdi)
+; WIDEN_KNL-NEXT:    vmovaps %zmm3, (%rdi)
 ; WIDEN_KNL-NEXT:    vzeroupper
 ; WIDEN_KNL-NEXT:    retq
 ;
@@ -711,32 +715,35 @@
 ; WIDEN_AVX2-NEXT:    vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; WIDEN_AVX2-NEXT:    vpinsrd $1, {{[0-9]+}}(%rsp), %xmm0, %xmm0
 ; WIDEN_AVX2-NEXT:    vpinsrd $2, {{[0-9]+}}(%rsp), %xmm0, %xmm0
+; WIDEN_AVX2-NEXT:    movq %rdi, %rax
 ; WIDEN_AVX2-NEXT:    vpinsrd $3, {{[0-9]+}}(%rsp), %xmm0, %xmm0
 ; WIDEN_AVX2-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
 ; WIDEN_AVX2-NEXT:    vpinsrd $1, {{[0-9]+}}(%rsp), %xmm1, %xmm1
+; WIDEN_AVX2-NEXT:    vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
 ; WIDEN_AVX2-NEXT:    vpinsrd $2, {{[0-9]+}}(%rsp), %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vpinsrd $3, {{[0-9]+}}(%rsp), %xmm1, %xmm1
-; WIDEN_AVX2-NEXT:    movq %rdi, %rax
-; WIDEN_AVX2-NEXT:    vmovd %edx, %xmm2
-; WIDEN_AVX2-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm2
-; WIDEN_AVX2-NEXT:    vpinsrd $2, %r8d, %xmm2, %xmm2
-; WIDEN_AVX2-NEXT:    vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
-; WIDEN_AVX2-NEXT:    vpinsrd $3, %r9d, %xmm2, %xmm2
-; WIDEN_AVX2-NEXT:    vmovd {{.*#+}} xmm4 = mem[0],zero,zero,zero
-; WIDEN_AVX2-NEXT:    vpinsrd $1, {{[0-9]+}}(%rsp), %xmm4, %xmm4
+; WIDEN_AVX2-NEXT:    vmovd %edx, %xmm3
+; WIDEN_AVX2-NEXT:    vpinsrd $1, %ecx, %xmm3, %xmm3
+; WIDEN_AVX2-NEXT:    vpinsrd $2, %r8d, %xmm3, %xmm3
+; WIDEN_AVX2-NEXT:    vpinsrd $3, %r9d, %xmm3, %xmm3
 ; WIDEN_AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
-; WIDEN_AVX2-NEXT:    vpinsrd $2, {{[0-9]+}}(%rsp), %xmm4, %xmm1
+; WIDEN_AVX2-NEXT:    vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; WIDEN_AVX2-NEXT:    vpinsrd $1, {{[0-9]+}}(%rsp), %xmm1, %xmm1
+; WIDEN_AVX2-NEXT:    vpinsrd $2, {{[0-9]+}}(%rsp), %xmm1, %xmm1
 ; WIDEN_AVX2-NEXT:    vpinsrd $3, {{[0-9]+}}(%rsp), %xmm1, %xmm1
-; WIDEN_AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
-; WIDEN_AVX2-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; WIDEN_AVX2-NEXT:    vgatherdps %ymm2, (%rsi,%ymm1,4), %ymm4
-; WIDEN_AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; WIDEN_AVX2-NEXT:    vgatherdps %ymm1, (%rsi,%ymm0,4), %ymm2
+; WIDEN_AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm3, %ymm1
+; WIDEN_AVX2-NEXT:    vpcmpeqd %ymm3, %ymm3, %ymm3
+; WIDEN_AVX2-NEXT:    vxorps %xmm4, %xmm4, %xmm4
+; WIDEN_AVX2-NEXT:    vpcmpeqd %ymm5, %ymm5, %ymm5
+; WIDEN_AVX2-NEXT:    vxorps %xmm6, %xmm6, %xmm6
+; WIDEN_AVX2-NEXT:    vgatherdps %ymm5, (%rsi,%ymm1,4), %ymm6
+; WIDEN_AVX2-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; WIDEN_AVX2-NEXT:    vgatherdps %ymm3, (%rsi,%ymm0,4), %ymm1
 ; WIDEN_AVX2-NEXT:    vmovaps {{.*#+}} xmm0 = [4294967295,0,0,0]
-; WIDEN_AVX2-NEXT:    vgatherdps %ymm0, (%rsi,%ymm3,4), %ymm1
-; WIDEN_AVX2-NEXT:    vmovss %xmm1, 64(%rdi)
-; WIDEN_AVX2-NEXT:    vmovaps %ymm2, 32(%rdi)
-; WIDEN_AVX2-NEXT:    vmovaps %ymm4, (%rdi)
+; WIDEN_AVX2-NEXT:    vgatherdps %ymm0, (%rsi,%ymm2,4), %ymm4
+; WIDEN_AVX2-NEXT:    vmovss %xmm4, 64(%rdi)
+; WIDEN_AVX2-NEXT:    vmovaps %ymm1, 32(%rdi)
+; WIDEN_AVX2-NEXT:    vmovaps %ymm6, (%rdi)
 ; WIDEN_AVX2-NEXT:    vzeroupper
 ; WIDEN_AVX2-NEXT:    retq
 {
diff --git a/llvm/test/CodeGen/X86/pr45067.ll b/llvm/test/CodeGen/X86/pr45067.ll
--- a/llvm/test/CodeGen/X86/pr45067.ll
+++ b/llvm/test/CodeGen/X86/pr45067.ll
@@ -7,8 +7,9 @@
 ; CHECK-LABEL: foo:
 ; CHECK:       ## %bb.0:
 ; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; CHECK-NEXT:    movq _global@GOTPCREL(%rip), %rax
+; CHECK-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; CHECK-NEXT:    vpxor %xmm3, %xmm3, %xmm3
 ; CHECK-NEXT:    vpgatherdd %ymm1, (%rax,%ymm2), %ymm3
 ; CHECK-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
 ; CHECK-NEXT:    vpslld $31, %ymm0, %ymm0