Index: lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- lib/Target/X86/X86ISelLowering.cpp
+++ lib/Target/X86/X86ISelLowering.cpp
@@ -19021,8 +19021,10 @@
                                     SDValue PreservedSrc,
                                     const X86Subtarget &Subtarget,
                                     SelectionDAG &DAG) {
-  if (isAllOnesConstant(Mask))
-    return Op;
+
+  if (auto *MaskConst = dyn_cast<ConstantSDNode>(Mask))
+    if (MaskConst->getZExtValue() & 0x1)
+      return Op;
 
   MVT VT = Op.getSimpleValueType();
   SDLoc dl(Op);
Index: test/CodeGen/X86/avx512-scalar_mask.ll
===================================================================
--- /dev/null
+++ test/CodeGen/X86/avx512-scalar_mask.ll
@@ -0,0 +1,107 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
+
+declare <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32)
+declare <4 x float> @llvm.x86.avx512.maskz.vfmadd.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32)
+
+define <4 x float>@test_var_mask(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2, i8 %mask) {
+; CHECK-LABEL: test_var_mask:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    andl $1, %edi
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 {%k1}
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 %mask, i32 4)
+  ret < 4 x float> %res
+}
+
+define <4 x float>@test_var_maskz(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2, i8 %mask) {
+; CHECK-LABEL: test_var_maskz:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    andl $1, %edi
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 {%k1} {z}
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 %mask, i32 4)
+  ret < 4 x float> %res
+}
+
+; FIXME: we should just return %xmm0 here.
+define <4 x float>@test_const0_mask(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const0_mask:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kxorw %k0, %k0, %k1
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 {%k1}
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 0, i32 4)
+  ret < 4 x float> %res
+}
+
+; FIXME: we should zero the lower element of xmm0 and return it.
+define <4 x float>@test_const0_maskz(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const0_maskz:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kxorw %k0, %k0, %k1
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 {%k1} {z}
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 0, i32 4)
+  ret < 4 x float> %res
+}
+
+; FIXME: we should just return %xmm0 here.
+define <4 x float>@test_const2_mask(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const2_mask:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kxorw %k0, %k0, %k1
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 {%k1}
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 2, i32 4)
+  ret < 4 x float> %res
+}
+
+; FIXME: we should zero the lower element of xmm0 and return it.
+define <4 x float>@test_const2_maskz(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const2_maskz:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kxorw %k0, %k0, %k1
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 {%k1} {z}
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 2, i32 4)
+  ret < 4 x float> %res
+}
+
+define <4 x float>@test_const_allone_mask(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const_allone_mask:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 -1, i32 4)
+  ret < 4 x float> %res
+}
+
+define <4 x float>@test_const_allone_maskz(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const_allone_maskz:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 -1, i32 4)
+  ret < 4 x float> %res
+}
+
+define <4 x float>@test_const_3_mask(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const_3_mask:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 3, i32 4)
+  ret < 4 x float> %res
+}
+
+define <4 x float>@test_const_3_maskz(<4 x float> %v0, <4 x float> %v1, <4 x float> %v2) {
+; CHECK-LABEL: test_const_3_maskz:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0
+; CHECK-NEXT:    retq
+  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ss(<4 x float> %v0,<4 x float> %v1, <4 x float> %v2,  i8 3, i32 4)
+  ret < 4 x float> %res
+}
Index: test/CodeGen/X86/avx512er-intrinsics.ll
===================================================================
--- test/CodeGen/X86/avx512er-intrinsics.ll
+++ test/CodeGen/X86/avx512er-intrinsics.ll
@@ -121,9 +121,7 @@
 define <4 x float> @test_rsqrt28_ss_maskz(<4 x float> %a0) {
 ; CHECK-LABEL: test_rsqrt28_ss_maskz:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 # encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kshiftrw $15, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc8,0x0f]
-; CHECK-NEXT:    vrsqrt28ss {sae}, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0xcd,0xc0]
+; CHECK-NEXT:    vrsqrt28ss {sae}, %xmm0, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0xcd,0xc0]
 ; CHECK-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx512.rsqrt28.ss(<4 x float> %a0, <4 x float> %a0, <4 x float> zeroinitializer, i8 7, i32 8) ;
   ret <4 x float> %res
@@ -132,10 +130,7 @@
 define <4 x float> @test_rsqrt28_ss_mask(<4 x float> %a0, <4 x float> %b0, <4 x float> %c0) {
 ; CHECK-LABEL: test_rsqrt28_ss_mask:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 # encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kshiftrw $15, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc8,0x0f]
-; CHECK-NEXT:    vrsqrt28ss {sae}, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0xcd,0xd1]
-; CHECK-NEXT:    vmovaps %xmm2, %xmm0 # encoding: [0xc5,0xf8,0x28,0xc2]
+; CHECK-NEXT:    vrsqrt28ss {sae}, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0xcd,0xc1]
 ; CHECK-NEXT:    retq # encoding: [0xc3]
   %res = call <4 x float> @llvm.x86.avx512.rsqrt28.ss(<4 x float> %a0, <4 x float> %b0, <4 x float> %c0, i8 7, i32 8) ;
   ret <4 x float> %res
@@ -144,9 +139,7 @@
 define <2 x double> @test_rsqrt28_sd_maskz(<2 x double> %a0) {
 ; CHECK-LABEL: test_rsqrt28_sd_maskz:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 # encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kshiftrw $15, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc8,0x0f]
-; CHECK-NEXT:    vrsqrt28sd {sae}, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x99,0xcd,0xc0]
+; CHECK-NEXT:    vrsqrt28sd {sae}, %xmm0, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x18,0xcd,0xc0]
 ; CHECK-NEXT:    retq # encoding: [0xc3]
   %res = call <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double> %a0, <2 x double> %a0, <2 x double> zeroinitializer, i8 7, i32 8) ;
   ret <2 x double> %res
@@ -155,10 +148,7 @@
 define <2 x double> @test_rsqrt28_sd_mask(<2 x double> %a0, <2 x double> %b0, <2 x double> %c0) {
 ; CHECK-LABEL: test_rsqrt28_sd_mask:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 # encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kshiftrw $15, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc8,0x0f]
-; CHECK-NEXT:    vrsqrt28sd {sae}, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x19,0xcd,0xd1]
-; CHECK-NEXT:    vmovapd %xmm2, %xmm0 # encoding: [0xc5,0xf9,0x28,0xc2]
+; CHECK-NEXT:    vrsqrt28sd {sae}, %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x18,0xcd,0xc1]
 ; CHECK-NEXT:    retq # encoding: [0xc3]
   %res = call <2 x double> @llvm.x86.avx512.rsqrt28.sd(<2 x double> %a0, <2 x double> %b0, <2 x double> %c0, i8 7, i32 8) ;
   ret <2 x double> %res
@@ -169,9 +159,7 @@
 define <2 x double> @test_rsqrt28_sd_maskz_mem(<2 x double> %a0, double* %ptr ) {
 ; CHECK-LABEL: test_rsqrt28_sd_maskz_mem:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 # encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kshiftrw $15, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc8,0x0f]
-; CHECK-NEXT:    vrsqrt28sd (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0x07]
+; CHECK-NEXT:    vrsqrt28sd (%rdi), %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0xcd,0x07]
 ; CHECK-NEXT:    retq # encoding: [0xc3]
   %mem = load double , double * %ptr, align 8
   %mem_v = insertelement <2 x double> undef, double %mem, i32 0
@@ -182,9 +170,7 @@
 define <2 x double> @test_rsqrt28_sd_maskz_mem_offset(<2 x double> %a0, double* %ptr ) {
 ; CHECK-LABEL: test_rsqrt28_sd_maskz_mem_offset:
 ; CHECK:       # BB#0:
-; CHECK-NEXT:    kxnorw %k0, %k0, %k0 # encoding: [0xc5,0xfc,0x46,0xc0]
-; CHECK-NEXT:    kshiftrw $15, %k0, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc8,0x0f]
-; CHECK-NEXT:    vrsqrt28sd 144(%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0xcd,0x47,0x12]
+; CHECK-NEXT:    vrsqrt28sd 144(%rdi), %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0xcd,0x47,0x12]
 ; CHECK-NEXT:    retq # encoding: [0xc3]
   %ptr1 = getelementptr double, double* %ptr, i32 18
   %mem = load double , double * %ptr1, align 8