diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -7735,39 +7735,43 @@
 //
 multiclass avx2_pmovmask<string OpcodeStr,
                          Intrinsic IntLd128, Intrinsic IntLd256,
-                         Intrinsic IntSt128, Intrinsic IntSt256> {
+                         Intrinsic IntSt128, Intrinsic IntSt256,
+                         X86SchedWriteMaskMove schedX,
+                         X86SchedWriteMaskMove schedY> {
   def rm  : AVX28I<0x8c, MRMSrcMem, (outs VR128:$dst),
              (ins VR128:$src1, i128mem:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(set VR128:$dst, (IntLd128 addr:$src2, VR128:$src1))]>,
-             VEX_4V, Sched<[WriteVecMaskedLoad]>;
+             VEX_4V, Sched<[schedX.RM]>;
   def Yrm : AVX28I<0x8c, MRMSrcMem, (outs VR256:$dst),
              (ins VR256:$src1, i256mem:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(set VR256:$dst, (IntLd256 addr:$src2, VR256:$src1))]>,
-             VEX_4V, VEX_L, Sched<[WriteVecMaskedLoadY]>;
+             VEX_4V, VEX_L, Sched<[schedY.RM]>;
   def mr  : AVX28I<0x8e, MRMDestMem, (outs),
              (ins i128mem:$dst, VR128:$src1, VR128:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(IntSt128 addr:$dst, VR128:$src1, VR128:$src2)]>,
-             VEX_4V, Sched<[WriteVecMaskedStore]>;
+             VEX_4V, Sched<[schedX.MR]>;
   def Ymr : AVX28I<0x8e, MRMDestMem, (outs),
              (ins i256mem:$dst, VR256:$src1, VR256:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(IntSt256 addr:$dst, VR256:$src1, VR256:$src2)]>,
-             VEX_4V, VEX_L, Sched<[WriteVecMaskedStoreY]>;
+             VEX_4V, VEX_L, Sched<[schedY.MR]>;
 }
 
 defm VPMASKMOVD : avx2_pmovmask<"vpmaskmovd",
                                 int_x86_avx2_maskload_d,
                                 int_x86_avx2_maskload_d_256,
                                 int_x86_avx2_maskstore_d,
-                                int_x86_avx2_maskstore_d_256>;
+                                int_x86_avx2_maskstore_d_256,
+                                WriteVecMaskMove32, WriteVecMaskMove32Y>;
 defm VPMASKMOVQ : avx2_pmovmask<"vpmaskmovq",
                                 int_x86_avx2_maskload_q,
                                 int_x86_avx2_maskload_q_256,
                                 int_x86_avx2_maskstore_q,
-                                int_x86_avx2_maskstore_q_256>, VEX_W;
+                                int_x86_avx2_maskstore_q_256,
+                                WriteVecMaskMove64, WriteVecMaskMove64Y>, VEX_W;
 
 multiclass maskmov_lowering<string InstrStr, RegisterClass RC, ValueType VT,
                           ValueType MaskVT> {
diff --git a/llvm/lib/Target/X86/X86SchedBroadwell.td b/llvm/lib/Target/X86/X86SchedBroadwell.td
--- a/llvm/lib/Target/X86/X86SchedBroadwell.td
+++ b/llvm/lib/Target/X86/X86SchedBroadwell.td
@@ -352,8 +352,10 @@
 defm : X86WriteRes<WriteVecStoreY,       [BWPort237,BWPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNT,      [BWPort237,BWPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNTY,     [BWPort237,BWPort4], 1, [1,1], 2>;
-defm : X86WriteRes<WriteVecMaskedStore,  [BWPort0,BWPort4,BWPort237,BWPort15], 5, [1,1,1,1], 4>;
-defm : X86WriteRes<WriteVecMaskedStoreY, [BWPort0,BWPort4,BWPort237,BWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore32,  [BWPort0,BWPort4,BWPort237,BWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore32Y, [BWPort0,BWPort4,BWPort237,BWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore64,  [BWPort0,BWPort4,BWPort237,BWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore64Y, [BWPort0,BWPort4,BWPort237,BWPort15], 5, [1,1,1,1], 4>;
 defm : X86WriteRes<WriteVecMove,         [BWPort015], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveX,        [BWPort015], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveY,        [BWPort015], 1, [1], 1>;
diff --git a/llvm/lib/Target/X86/X86SchedHaswell.td b/llvm/lib/Target/X86/X86SchedHaswell.td
--- a/llvm/lib/Target/X86/X86SchedHaswell.td
+++ b/llvm/lib/Target/X86/X86SchedHaswell.td
@@ -392,8 +392,10 @@
 defm : X86WriteRes<WriteVecStoreY,       [HWPort237,HWPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNT,      [HWPort237,HWPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNTY,     [HWPort237,HWPort4], 1, [1,1], 2>;
-defm : X86WriteRes<WriteVecMaskedStore,  [HWPort0,HWPort4,HWPort237,HWPort15], 5, [1,1,1,1], 4>;
-defm : X86WriteRes<WriteVecMaskedStoreY, [HWPort0,HWPort4,HWPort237,HWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore32,  [HWPort0,HWPort4,HWPort237,HWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore32Y, [HWPort0,HWPort4,HWPort237,HWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore64,  [HWPort0,HWPort4,HWPort237,HWPort15], 5, [1,1,1,1], 4>;
+defm : X86WriteRes<WriteVecMaskedStore64Y, [HWPort0,HWPort4,HWPort237,HWPort15], 5, [1,1,1,1], 4>;
 defm : X86WriteRes<WriteVecMove,         [HWPort015], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveX,        [HWPort015], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveY,        [HWPort015], 1, [1], 1>;
diff --git a/llvm/lib/Target/X86/X86SchedSandyBridge.td b/llvm/lib/Target/X86/X86SchedSandyBridge.td
--- a/llvm/lib/Target/X86/X86SchedSandyBridge.td
+++ b/llvm/lib/Target/X86/X86SchedSandyBridge.td
@@ -367,8 +367,10 @@
 defm : X86WriteRes<WriteVecStoreY,       [SBPort23,SBPort4], 1, [1,1], 1>;
 defm : X86WriteRes<WriteVecStoreNT,      [SBPort23,SBPort4], 1, [1,1], 1>;
 defm : X86WriteRes<WriteVecStoreNTY,     [SBPort23,SBPort4], 1, [1,1], 1>;
-defm : X86WriteRes<WriteVecMaskedStore,  [SBPort4,SBPort01,SBPort23], 5, [1,1,1], 3>;
-defm : X86WriteRes<WriteVecMaskedStoreY, [SBPort4,SBPort01,SBPort23], 5, [1,1,1], 3>;
+defm : X86WriteRes<WriteVecMaskedStore32,  [SBPort4,SBPort01,SBPort23], 5, [1,1,1], 3>;
+defm : X86WriteRes<WriteVecMaskedStore32Y, [SBPort4,SBPort01,SBPort23], 5, [1,1,1], 3>;
+defm : X86WriteRes<WriteVecMaskedStore64,  [SBPort4,SBPort01,SBPort23], 5, [1,1,1], 3>;
+defm : X86WriteRes<WriteVecMaskedStore64Y, [SBPort4,SBPort01,SBPort23], 5, [1,1,1], 3>;
 defm : X86WriteRes<WriteVecMove,         [SBPort05], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveX,        [SBPort015], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveY,        [SBPort05], 1, [1], 1>;
diff --git a/llvm/lib/Target/X86/X86SchedSkylakeClient.td b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
--- a/llvm/lib/Target/X86/X86SchedSkylakeClient.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
@@ -343,8 +343,10 @@
 defm : X86WriteRes<WriteVecStoreY,       [SKLPort237,SKLPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNT,      [SKLPort237,SKLPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNTY,     [SKLPort237,SKLPort4], 1, [1,1], 2>;
-defm : X86WriteRes<WriteVecMaskedStore,  [SKLPort237,SKLPort0], 2, [1,1], 2>;
-defm : X86WriteRes<WriteVecMaskedStoreY, [SKLPort237,SKLPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore32,  [SKLPort237,SKLPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore32Y, [SKLPort237,SKLPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore64,  [SKLPort237,SKLPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore64Y, [SKLPort237,SKLPort0], 2, [1,1], 2>;
 defm : X86WriteRes<WriteVecMove,         [SKLPort05],  1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveX,        [SKLPort015], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveY,        [SKLPort015], 1, [1], 1>;
diff --git a/llvm/lib/Target/X86/X86SchedSkylakeServer.td b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
--- a/llvm/lib/Target/X86/X86SchedSkylakeServer.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
@@ -343,8 +343,10 @@
 defm : X86WriteRes<WriteVecStoreY,       [SKXPort237,SKXPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNT,      [SKXPort237,SKXPort4], 1, [1,1], 2>;
 defm : X86WriteRes<WriteVecStoreNTY,     [SKXPort237,SKXPort4], 1, [1,1], 2>;
-defm : X86WriteRes<WriteVecMaskedStore,  [SKXPort237,SKXPort0], 2, [1,1], 2>;
-defm : X86WriteRes<WriteVecMaskedStoreY, [SKXPort237,SKXPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore32,  [SKXPort237,SKXPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore32Y, [SKXPort237,SKXPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore64,  [SKXPort237,SKXPort0], 2, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore64Y, [SKXPort237,SKXPort0], 2, [1,1], 2>;
 defm : X86WriteRes<WriteVecMove,         [SKXPort05],  1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveX,        [SKXPort015], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveY,        [SKXPort015], 1, [1], 1>;
diff --git a/llvm/lib/Target/X86/X86Schedule.td b/llvm/lib/Target/X86/X86Schedule.td
--- a/llvm/lib/Target/X86/X86Schedule.td
+++ b/llvm/lib/Target/X86/X86Schedule.td
@@ -341,8 +341,10 @@
 def  WriteVecStoreY       : SchedWrite;
 def  WriteVecStoreNT      : SchedWrite;
 def  WriteVecStoreNTY     : SchedWrite;
-def  WriteVecMaskedStore  : SchedWrite;
-def  WriteVecMaskedStoreY : SchedWrite;
+def  WriteVecMaskedStore32  : SchedWrite;
+def  WriteVecMaskedStore64  : SchedWrite;
+def  WriteVecMaskedStore32Y : SchedWrite;
+def  WriteVecMaskedStore64Y : SchedWrite;
 def  WriteVecMove         : SchedWrite;
 def  WriteVecMoveX        : SchedWrite;
 def  WriteVecMoveY        : SchedWrite;
@@ -550,6 +552,14 @@
   : X86SchedWriteMaskMove<WriteFMaskedLoadY, WriteFMaskedStore32Y>;
 def WriteFMaskMove64Y
   : X86SchedWriteMaskMove<WriteFMaskedLoadY, WriteFMaskedStore64Y>;
+def WriteVecMaskMove32
+  : X86SchedWriteMaskMove<WriteVecMaskedLoad, WriteVecMaskedStore32>;
+def WriteVecMaskMove64
+  : X86SchedWriteMaskMove<WriteVecMaskedLoad, WriteVecMaskedStore64>;
+def WriteVecMaskMove32Y
+  : X86SchedWriteMaskMove<WriteVecMaskedLoadY, WriteVecMaskedStore32Y>;
+def WriteVecMaskMove64Y
+  : X86SchedWriteMaskMove<WriteVecMaskedLoadY, WriteVecMaskedStore64Y>;
 
 // Vector width wrappers.
 def SchedWriteFAdd
diff --git a/llvm/lib/Target/X86/X86ScheduleAtom.td b/llvm/lib/Target/X86/X86ScheduleAtom.td
--- a/llvm/lib/Target/X86/X86ScheduleAtom.td
+++ b/llvm/lib/Target/X86/X86ScheduleAtom.td
@@ -369,8 +369,10 @@
 defm : X86WriteResUnsupported<WriteVecStoreY>;
 def  : WriteRes<WriteVecStoreNT,      [AtomPort0]>;
 defm : X86WriteResUnsupported<WriteVecStoreNTY>;
-def  : WriteRes<WriteVecMaskedStore,  [AtomPort0]>;
-defm : X86WriteResUnsupported<WriteVecMaskedStoreY>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore32>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore64>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore32Y>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore64Y>;
 
 def  : WriteRes<WriteVecMove,          [AtomPort0]>;
 def  : WriteRes<WriteVecMoveX,        [AtomPort01]>;
diff --git a/llvm/lib/Target/X86/X86ScheduleBdVer2.td b/llvm/lib/Target/X86/X86ScheduleBdVer2.td
--- a/llvm/lib/Target/X86/X86ScheduleBdVer2.td
+++ b/llvm/lib/Target/X86/X86ScheduleBdVer2.td
@@ -1093,8 +1093,10 @@
 defm : PdWriteRes<WriteVecStoreNT,          [PdStore, PdFPU1,   PdFPSTO], 2>;
 defm : PdWriteRes<WriteVecStoreNTY,         [PdStore, PdFPU1,   PdFPSTO], 2, [2, 2, 2], 4>;
 
-defm : PdWriteRes<WriteVecMaskedStore,      [PdStore, PdFPU01, PdFPMAL], 6, [1, 1, 4]>;
-defm : PdWriteRes<WriteVecMaskedStoreY,     [PdStore, PdFPU01, PdFPMAL], 6, [2, 2, 4], 2>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore32>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore32Y>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore64>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore64Y>;
 
 defm : PdWriteRes<WriteVecMove,             [PdFPU01, PdFPMAL], 2>;
 defm : PdWriteRes<WriteVecMoveX,            [PdFPU01, PdFPMAL], 1, [1, 2]>;
diff --git a/llvm/lib/Target/X86/X86ScheduleBtVer2.td b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
--- a/llvm/lib/Target/X86/X86ScheduleBtVer2.td
+++ b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
@@ -670,8 +670,10 @@
 defm : X86WriteRes<WriteVecStoreY,        [JSAGU, JFPU1,   JSTC], 1, [2, 2, 2], 2>;
 defm : X86WriteRes<WriteVecStoreNT,       [JSAGU, JFPU1,   JSTC], 2, [1, 1, 1], 1>;
 defm : X86WriteRes<WriteVecStoreNTY,      [JSAGU, JFPU1,   JSTC], 2, [2, 2, 2], 1>;
-defm : X86WriteRes<WriteVecMaskedStore,   [JSAGU, JFPU01, JVALU], 6, [1, 1, 4], 1>;
-defm : X86WriteRes<WriteVecMaskedStoreY,  [JSAGU, JFPU01, JVALU], 6, [2, 2, 4], 2>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore32>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore64>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore32Y>;
+defm : X86WriteResUnsupported<WriteVecMaskedStore64Y>;
 
 defm : X86WriteRes<WriteVecMove,          [JFPU01, JVALU], 1, [1, 1], 1>;
 defm : X86WriteRes<WriteVecMoveX,         [JFPU01, JVALU], 1, [1, 1], 1>;
diff --git a/llvm/lib/Target/X86/X86ScheduleSLM.td b/llvm/lib/Target/X86/X86ScheduleSLM.td
--- a/llvm/lib/Target/X86/X86ScheduleSLM.td
+++ b/llvm/lib/Target/X86/X86ScheduleSLM.td
@@ -311,8 +311,10 @@
 def  : WriteRes<WriteVecStoreY,       [SLM_MEC_RSV]>;
 def  : WriteRes<WriteVecStoreNT,      [SLM_MEC_RSV]>;
 def  : WriteRes<WriteVecStoreNTY,     [SLM_MEC_RSV]>;
-def  : WriteRes<WriteVecMaskedStore,  [SLM_MEC_RSV]>;
-def  : WriteRes<WriteVecMaskedStoreY, [SLM_MEC_RSV]>;
+def  : WriteRes<WriteVecMaskedStore32,    [SLM_MEC_RSV]>;
+def  : WriteRes<WriteVecMaskedStore32Y,   [SLM_MEC_RSV]>;
+def  : WriteRes<WriteVecMaskedStore64,    [SLM_MEC_RSV]>;
+def  : WriteRes<WriteVecMaskedStore64Y,   [SLM_MEC_RSV]>;
 def  : WriteRes<WriteVecMove,         [SLM_FPC_RSV01]>;
 def  : WriteRes<WriteVecMoveX,        [SLM_FPC_RSV01]>;
 def  : WriteRes<WriteVecMoveY,        [SLM_FPC_RSV01]>;
diff --git a/llvm/lib/Target/X86/X86ScheduleZnver1.td b/llvm/lib/Target/X86/X86ScheduleZnver1.td
--- a/llvm/lib/Target/X86/X86ScheduleZnver1.td
+++ b/llvm/lib/Target/X86/X86ScheduleZnver1.td
@@ -388,8 +388,10 @@
 defm : X86WriteRes<WriteVecStoreY,       [ZnAGU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecStoreNT,      [ZnAGU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecStoreNTY,     [ZnAGU], 1, [1], 1>;
-defm : X86WriteRes<WriteVecMaskedStore,  [ZnAGU,ZnFPU01], 4, [1,1], 1>;
-defm : X86WriteRes<WriteVecMaskedStoreY, [ZnAGU,ZnFPU01], 5, [1,2], 2>;
+defm : X86WriteRes<WriteVecMaskedStore32,  [ZnAGU,ZnFPU01], 4, [1,1], 1>;
+defm : X86WriteRes<WriteVecMaskedStore32Y, [ZnAGU,ZnFPU01], 5, [1,2], 2>;
+defm : X86WriteRes<WriteVecMaskedStore64,  [ZnAGU,ZnFPU01], 4, [1,1], 1>;
+defm : X86WriteRes<WriteVecMaskedStore64Y, [ZnAGU,ZnFPU01], 5, [1,2], 2>;
 defm : X86WriteRes<WriteVecMove,         [ZnFPU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveX,        [ZnFPU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveY,        [ZnFPU], 2, [1], 2>;
diff --git a/llvm/lib/Target/X86/X86ScheduleZnver2.td b/llvm/lib/Target/X86/X86ScheduleZnver2.td
--- a/llvm/lib/Target/X86/X86ScheduleZnver2.td
+++ b/llvm/lib/Target/X86/X86ScheduleZnver2.td
@@ -370,8 +370,10 @@
 defm : X86WriteRes<WriteVecStoreY,       [Zn2AGU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecStoreNT,      [Zn2AGU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecStoreNTY,     [Zn2AGU], 1, [1], 1>;
-defm : X86WriteRes<WriteVecMaskedStore,  [Zn2AGU,Zn2FPU01], 4, [1,1], 1>;
-defm : X86WriteRes<WriteVecMaskedStoreY, [Zn2AGU,Zn2FPU01], 5, [1,1], 2>;
+defm : X86WriteRes<WriteVecMaskedStore32,  [Zn2AGU,Zn2FPU01], 4, [1,1], 1>;
+defm : X86WriteRes<WriteVecMaskedStore32Y, [Zn2AGU,Zn2FPU01], 5, [1,2], 2>;
+defm : X86WriteRes<WriteVecMaskedStore64,  [Zn2AGU,Zn2FPU01], 4, [1,1], 1>;
+defm : X86WriteRes<WriteVecMaskedStore64Y, [Zn2AGU,Zn2FPU01], 5, [1,2], 2>;
 defm : X86WriteRes<WriteVecMove,         [Zn2FPU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveX,        [Zn2FPU], 1, [1], 1>;
 defm : X86WriteRes<WriteVecMoveY,        [Zn2FPU], 2, [1], 2>;