diff --git a/src/cmd/compile/internal/amd64/ssa.go b/src/cmd/compile/internal/amd64/ssa.go
index 9c31b77e703..0fafd69f54b 100644
--- a/src/cmd/compile/internal/amd64/ssa.go
+++ b/src/cmd/compile/internal/amd64/ssa.go
@@ -1461,13 +1461,13 @@ func ssaGenValue(s *ssagen.State, v *ssa.Value) {
 		p.AddRestSourceReg(simdReg(v.Args[1]))
 		p.To.Type = obj.TYPE_REG
 		p.To.Reg = simdReg(v)
-	case ssa.OpAMD64VMOVDQUload128, ssa.OpAMD64VMOVDQUload256, ssa.OpAMD64VMOVDQUload512:
+	case ssa.OpAMD64VMOVDQUload128, ssa.OpAMD64VMOVDQUload256, ssa.OpAMD64VMOVDQUload512, ssa.OpAMD64KMOVQload:
 		p := s.Prog(v.Op.Asm())
 		p.From.Type = obj.TYPE_MEM
 		p.From.Reg = v.Args[0].Reg()
 		ssagen.AddAux(&p.From, v)
 		p.To.Type = obj.TYPE_REG
-		p.To.Reg = simdReg(v)
+		p.To.Reg = simdOrMaskReg(v)
 	case ssa.OpAMD64VMOVDQUstore128, ssa.OpAMD64VMOVDQUstore256, ssa.OpAMD64VMOVDQUstore512:
 		p := s.Prog(v.Op.Asm())
 		p.From.Type = obj.TYPE_REG
diff --git a/src/cmd/compile/internal/ssa/_gen/AMD64.rules b/src/cmd/compile/internal/ssa/_gen/AMD64.rules
index 2972eae87d5..bb7513795d9 100644
--- a/src/cmd/compile/internal/ssa/_gen/AMD64.rules
+++ b/src/cmd/compile/internal/ssa/_gen/AMD64.rules
@@ -1682,6 +1682,22 @@
 (Select0 a:(ADD(Q|L)constflags [c] x)) && a.Uses == 1 => (ADD(Q|L)const [c] x)
 
 // XXX SIMD
+(LoadMask8x16 <t> ptr mem) => (VPMOVMToVec8x16 <types.TypeVec128> (KMOVQload <t> ptr mem))
+(LoadMask8x32 <t> ptr mem) => (VPMOVMToVec8x32 <types.TypeVec256> (KMOVQload <t> ptr mem))
+(LoadMask8x64 <t> ptr mem) => (VPMOVMToVec8x64 <types.TypeVec512> (KMOVQload <t> ptr mem))
+
+(LoadMask16x8 <t> ptr mem) => (VPMOVMToVec16x8 <types.TypeVec128> (KMOVQload <t> ptr mem))
+(LoadMask16x16 <t> ptr mem) => (VPMOVMToVec16x16 <types.TypeVec256> (KMOVQload <t> ptr mem))
+(LoadMask16x32 <t> ptr mem) => (VPMOVMToVec16x32 <types.TypeVec512> (KMOVQload <t> ptr mem))
+
+(LoadMask32x4 <t> ptr mem) => (VPMOVMToVec32x4 <types.TypeVec128> (KMOVQload <t> ptr mem))
+(LoadMask32x8 <t> ptr mem) => (VPMOVMToVec32x8 <types.TypeVec256> (KMOVQload <t> ptr mem))
+(LoadMask32x16 <t> ptr mem) => (VPMOVMToVec32x16 <types.TypeVec512> (KMOVQload <t> ptr mem))
+
+(LoadMask64x2 <t> ptr mem) => (VPMOVMToVec64x2 <types.TypeVec128> (KMOVQload <t> ptr mem))
+(LoadMask64x4 <t> ptr mem) => (VPMOVMToVec64x4 <types.TypeVec256> (KMOVQload <t> ptr mem))
+(LoadMask64x8 <t> ptr mem) => (VPMOVMToVec64x8 <types.TypeVec512> (KMOVQload <t> ptr mem))
+
 (Load <t> ptr mem) && t.Size() == 16 => (VMOVDQUload128 ptr mem)
 
 (Store {t} ptr val mem) && t.Size() == 16 => (VMOVDQUstore128 ptr val mem)
diff --git a/src/cmd/compile/internal/ssa/_gen/AMD64Ops.go b/src/cmd/compile/internal/ssa/_gen/AMD64Ops.go
index 543233f4d83..ec335f67f87 100644
--- a/src/cmd/compile/internal/ssa/_gen/AMD64Ops.go
+++ b/src/cmd/compile/internal/ssa/_gen/AMD64Ops.go
@@ -234,6 +234,8 @@ func init() {
 		wfpw  = regInfo{inputs: []regMask{w, fp}, outputs: wonly}
 		wfpkw = regInfo{inputs: []regMask{w, fp, mask}, outputs: wonly}
 
+		kload = regInfo{inputs: []regMask{gpspsb, 0}, outputs: maskonly}
+
 		prefreg = regInfo{inputs: []regMask{gpspsbg}}
 	)
 
@@ -1314,6 +1316,8 @@ func init() {
 
 		{name: "VZEROUPPER", argLength: 0, asm: "VZEROUPPER"},
 		{name: "VZEROALL", argLength: 0, asm: "VZEROALL"},
+
+		{name: "KMOVQload", argLength: 2, reg: kload, asm: "KMOVQ", aux: "SymOff", faultOnNilArg0: true, symEffect: "Read"},
 	}
 
 	var AMD64blocks = []blockData{
diff --git a/src/cmd/compile/internal/ssa/_gen/genericOps.go b/src/cmd/compile/internal/ssa/_gen/genericOps.go
index 2d44cc85f82..6257396a6f5 100644
--- a/src/cmd/compile/internal/ssa/_gen/genericOps.go
+++ b/src/cmd/compile/internal/ssa/_gen/genericOps.go
@@ -666,6 +666,18 @@ var genericOps = []opData{
 	// XXX SIMD
 	{name: "Add32x4", argLength: 2}, // arg0 + arg1
 	{name: "ZeroSIMD", argLength: 0},
+	{name: "LoadMask8x16", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask8x32", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask8x64", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask16x8", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask16x16", argLength: 2}, // arg0 = ptr, arg1 = mem
+	{name: "LoadMask16x32", argLength: 2}, // arg0 = ptr, arg1 = mem
+	{name: "LoadMask32x4", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask32x8", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask32x16", argLength: 2}, // arg0 = ptr, arg1 = mem
+	{name: "LoadMask64x2", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask64x4", argLength: 2},  // arg0 = ptr, arg1 = mem
+	{name: "LoadMask64x8", argLength: 2},  // arg0 = ptr, arg1 = mem
 }
 
 //     kind          controls          successors   implicit exit
diff --git a/src/cmd/compile/internal/ssa/opGen.go b/src/cmd/compile/internal/ssa/opGen.go
index 29058f0b193..d69e714082b 100644
--- a/src/cmd/compile/internal/ssa/opGen.go
+++ b/src/cmd/compile/internal/ssa/opGen.go
@@ -1198,6 +1198,7 @@ const (
 	OpAMD64Zero512
 	OpAMD64VZEROUPPER
 	OpAMD64VZEROALL
+	OpAMD64KMOVQload
 	OpAMD64VADDPS512
 	OpAMD64VADDPSMasked512
 	OpAMD64VRCP14PS512
@@ -4403,6 +4404,18 @@ const (
 	OpPrefetchCacheStreamed
 	OpAdd32x4
 	OpZeroSIMD
+	OpLoadMask8x16
+	OpLoadMask8x32
+	OpLoadMask8x64
+	OpLoadMask16x8
+	OpLoadMask16x16
+	OpLoadMask16x32
+	OpLoadMask32x4
+	OpLoadMask32x8
+	OpLoadMask32x16
+	OpLoadMask64x2
+	OpLoadMask64x4
+	OpLoadMask64x8
 	OpAddFloat32x16
 	OpAddMaskedFloat32x16
 	OpApproximateReciprocalFloat32x16
@@ -18801,6 +18814,22 @@ var opcodeTable = [...]opInfo{
 		asm:    x86.AVZEROALL,
 		reg:    regInfo{},
 	},
+	{
+		name:           "KMOVQload",
+		auxType:        auxSymOff,
+		argLen:         2,
+		faultOnNilArg0: true,
+		symEffect:      SymRead,
+		asm:            x86.AKMOVQ,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 72057594037977087}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15 SB
+			},
+			outputs: []outputInfo{
+				{0, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+			},
+		},
+	},
 	{
 		name:        "VADDPS512",
 		argLen:      2,
@@ -60727,6 +60756,66 @@ var opcodeTable = [...]opInfo{
 		argLen:  0,
 		generic: true,
 	},
+	{
+		name:    "LoadMask8x16",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask8x32",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask8x64",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask16x8",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask16x16",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask16x32",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask32x4",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask32x8",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask32x16",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask64x2",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask64x4",
+		argLen:  2,
+		generic: true,
+	},
+	{
+		name:    "LoadMask64x8",
+		argLen:  2,
+		generic: true,
+	},
 	{
 		name:        "AddFloat32x16",
 		argLen:      2,
diff --git a/src/cmd/compile/internal/ssa/rewriteAMD64.go b/src/cmd/compile/internal/ssa/rewriteAMD64.go
index 5c7cafd6f23..0ff19a680e4 100644
--- a/src/cmd/compile/internal/ssa/rewriteAMD64.go
+++ b/src/cmd/compile/internal/ssa/rewriteAMD64.go
@@ -2438,6 +2438,30 @@ func rewriteValueAMD64(v *Value) bool {
 		return rewriteValueAMD64_OpLessUint8x64(v)
 	case OpLoad:
 		return rewriteValueAMD64_OpLoad(v)
+	case OpLoadMask16x16:
+		return rewriteValueAMD64_OpLoadMask16x16(v)
+	case OpLoadMask16x32:
+		return rewriteValueAMD64_OpLoadMask16x32(v)
+	case OpLoadMask16x8:
+		return rewriteValueAMD64_OpLoadMask16x8(v)
+	case OpLoadMask32x16:
+		return rewriteValueAMD64_OpLoadMask32x16(v)
+	case OpLoadMask32x4:
+		return rewriteValueAMD64_OpLoadMask32x4(v)
+	case OpLoadMask32x8:
+		return rewriteValueAMD64_OpLoadMask32x8(v)
+	case OpLoadMask64x2:
+		return rewriteValueAMD64_OpLoadMask64x2(v)
+	case OpLoadMask64x4:
+		return rewriteValueAMD64_OpLoadMask64x4(v)
+	case OpLoadMask64x8:
+		return rewriteValueAMD64_OpLoadMask64x8(v)
+	case OpLoadMask8x16:
+		return rewriteValueAMD64_OpLoadMask8x16(v)
+	case OpLoadMask8x32:
+		return rewriteValueAMD64_OpLoadMask8x32(v)
+	case OpLoadMask8x64:
+		return rewriteValueAMD64_OpLoadMask8x64(v)
 	case OpLocalAddr:
 		return rewriteValueAMD64_OpLocalAddr(v)
 	case OpLsh16x16:
@@ -40303,6 +40327,222 @@ func rewriteValueAMD64_OpLoad(v *Value) bool {
 	}
 	return false
 }
+func rewriteValueAMD64_OpLoadMask16x16(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask16x16 <t> ptr mem)
+	// result: (VPMOVMToVec16x16 <types.TypeVec256> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec16x16)
+		v.Type = types.TypeVec256
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask16x32(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask16x32 <t> ptr mem)
+	// result: (VPMOVMToVec16x32 <types.TypeVec512> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec16x32)
+		v.Type = types.TypeVec512
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask16x8(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask16x8 <t> ptr mem)
+	// result: (VPMOVMToVec16x8 <types.TypeVec128> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec16x8)
+		v.Type = types.TypeVec128
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask32x16(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask32x16 <t> ptr mem)
+	// result: (VPMOVMToVec32x16 <types.TypeVec512> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec32x16)
+		v.Type = types.TypeVec512
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask32x4(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask32x4 <t> ptr mem)
+	// result: (VPMOVMToVec32x4 <types.TypeVec128> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec32x4)
+		v.Type = types.TypeVec128
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask32x8(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask32x8 <t> ptr mem)
+	// result: (VPMOVMToVec32x8 <types.TypeVec256> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec32x8)
+		v.Type = types.TypeVec256
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask64x2(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask64x2 <t> ptr mem)
+	// result: (VPMOVMToVec64x2 <types.TypeVec128> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec64x2)
+		v.Type = types.TypeVec128
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask64x4(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask64x4 <t> ptr mem)
+	// result: (VPMOVMToVec64x4 <types.TypeVec256> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec64x4)
+		v.Type = types.TypeVec256
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask64x8(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask64x8 <t> ptr mem)
+	// result: (VPMOVMToVec64x8 <types.TypeVec512> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec64x8)
+		v.Type = types.TypeVec512
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask8x16(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask8x16 <t> ptr mem)
+	// result: (VPMOVMToVec8x16 <types.TypeVec128> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec8x16)
+		v.Type = types.TypeVec128
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask8x32(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask8x32 <t> ptr mem)
+	// result: (VPMOVMToVec8x32 <types.TypeVec256> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec8x32)
+		v.Type = types.TypeVec256
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
+func rewriteValueAMD64_OpLoadMask8x64(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	b := v.Block
+	// match: (LoadMask8x64 <t> ptr mem)
+	// result: (VPMOVMToVec8x64 <types.TypeVec512> (KMOVQload <t> ptr mem))
+	for {
+		t := v.Type
+		ptr := v_0
+		mem := v_1
+		v.reset(OpAMD64VPMOVMToVec8x64)
+		v.Type = types.TypeVec512
+		v0 := b.NewValue0(v.Pos, OpAMD64KMOVQload, t)
+		v0.AddArg2(ptr, mem)
+		v.AddArg(v0)
+		return true
+	}
+}
 func rewriteValueAMD64_OpLocalAddr(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
diff --git a/src/cmd/compile/internal/ssagen/intrinsics.go b/src/cmd/compile/internal/ssagen/intrinsics.go
index 5415143ec31..e012b536b55 100644
--- a/src/cmd/compile/internal/ssagen/intrinsics.go
+++ b/src/cmd/compile/internal/ssagen/intrinsics.go
@@ -1775,6 +1775,22 @@ func simdStore() func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
 	}
 }
 
+func simdLoadMask(elemBits, lanes int) func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
+	return func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value {
+		opCodes := map[int]map[int]ssa.Op{
+			8:  {16: ssa.OpLoadMask8x16, 32: ssa.OpLoadMask8x32, 64: ssa.OpLoadMask8x64},
+			16: {8: ssa.OpLoadMask16x8, 16: ssa.OpLoadMask16x16, 32: ssa.OpLoadMask16x32},
+			32: {4: ssa.OpLoadMask32x4, 8: ssa.OpLoadMask32x8, 16: ssa.OpLoadMask32x16},
+			64: {2: ssa.OpLoadMask64x2, 4: ssa.OpLoadMask64x4, 8: ssa.OpLoadMask64x8},
+		}
+		op := opCodes[elemBits][lanes]
+		if op == 0 {
+			panic(fmt.Sprintf("Unknown mask shape: Mask%dx%d", elemBits, lanes))
+		}
+		return s.newValue2(op, types.TypeMask, args[0], s.mem())
+	}
+}
+
 // findIntrinsic returns a function which builds the SSA equivalent of the
 // function identified by the symbol sym.  If sym is not an intrinsic call, returns nil.
 func findIntrinsic(sym *types.Sym) intrinsicBuilder {
diff --git a/src/cmd/compile/internal/ssagen/simdintrinsics.go b/src/cmd/compile/internal/ssagen/simdintrinsics.go
index 3d929499085..8040a187bda 100644
--- a/src/cmd/compile/internal/ssagen/simdintrinsics.go
+++ b/src/cmd/compile/internal/ssagen/simdintrinsics.go
@@ -2132,76 +2132,64 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Uint64x4.Store", simdStore(), sys.AMD64)
 	addF(simdPackage, "LoadUint64x8", simdLoad(), sys.AMD64)
 	addF(simdPackage, "Uint64x8.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask8x16", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask8x16.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask8x32", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask8x32.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask8x64", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask8x64.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask16x8", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask16x8.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask16x16", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask16x16.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask16x32", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask16x32.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask32x4", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask32x4.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask32x8", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask32x8.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask32x16", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask32x16.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask64x2", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask64x2.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask64x4", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask64x4.Store", simdStore(), sys.AMD64)
-	addF(simdPackage, "LoadMask64x8", simdLoad(), sys.AMD64)
-	addF(simdPackage, "Mask64x8.Store", simdStore(), sys.AMD64)
 	addF(simdPackage, "Mask8x16.AsInt8x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int8x16.AsMask8x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask8x16.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask8x16.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
+	addF(simdPackage, "LoadMask8x16FromBits", simdLoadMask(8, 16), sys.AMD64)
 	addF(simdPackage, "Mask8x32.AsInt8x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int8x32.AsMask8x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask8x32.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask8x32.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
+	addF(simdPackage, "LoadMask8x32FromBits", simdLoadMask(8, 32), sys.AMD64)
 	addF(simdPackage, "Mask8x64.AsInt8x64", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int8x64.AsMask8x64", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask8x64.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask8x64.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
+	addF(simdPackage, "LoadMask8x64FromBits", simdLoadMask(8, 64), sys.AMD64)
 	addF(simdPackage, "Mask16x8.AsInt16x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int16x8.AsMask16x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask16x8.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask16x8.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
+	addF(simdPackage, "LoadMask16x8FromBits", simdLoadMask(16, 8), sys.AMD64)
 	addF(simdPackage, "Mask16x16.AsInt16x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int16x16.AsMask16x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask16x16.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask16x16.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
+	addF(simdPackage, "LoadMask16x16FromBits", simdLoadMask(16, 16), sys.AMD64)
 	addF(simdPackage, "Mask16x32.AsInt16x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int16x32.AsMask16x32", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask16x32.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask16x32.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
+	addF(simdPackage, "LoadMask16x32FromBits", simdLoadMask(16, 32), sys.AMD64)
 	addF(simdPackage, "Mask32x4.AsInt32x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int32x4.AsMask32x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask32x4.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask32x4.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
+	addF(simdPackage, "LoadMask32x4FromBits", simdLoadMask(32, 4), sys.AMD64)
 	addF(simdPackage, "Mask32x8.AsInt32x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int32x8.AsMask32x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask32x8.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask32x8.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
+	addF(simdPackage, "LoadMask32x8FromBits", simdLoadMask(32, 8), sys.AMD64)
 	addF(simdPackage, "Mask32x16.AsInt32x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int32x16.AsMask32x16", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask32x16.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask32x16.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
+	addF(simdPackage, "LoadMask32x16FromBits", simdLoadMask(32, 16), sys.AMD64)
 	addF(simdPackage, "Mask64x2.AsInt64x2", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int64x2.AsMask64x2", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask64x2.And", opLen2(ssa.OpAndInt32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Mask64x2.Or", opLen2(ssa.OpOrInt32x4, types.TypeVec128), sys.AMD64)
+	addF(simdPackage, "LoadMask64x2FromBits", simdLoadMask(64, 2), sys.AMD64)
 	addF(simdPackage, "Mask64x4.AsInt64x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int64x4.AsMask64x4", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask64x4.And", opLen2(ssa.OpAndInt32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Mask64x4.Or", opLen2(ssa.OpOrInt32x8, types.TypeVec256), sys.AMD64)
+	addF(simdPackage, "LoadMask64x4FromBits", simdLoadMask(64, 4), sys.AMD64)
 	addF(simdPackage, "Mask64x8.AsInt64x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Int64x8.AsMask64x8", func(s *state, n *ir.CallExpr, args []*ssa.Value) *ssa.Value { return args[0] }, sys.AMD64)
 	addF(simdPackage, "Mask64x8.And", opLen2(ssa.OpAndInt32x16, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Mask64x8.Or", opLen2(ssa.OpOrInt32x16, types.TypeVec512), sys.AMD64)
+	addF(simdPackage, "LoadMask64x8FromBits", simdLoadMask(64, 8), sys.AMD64)
 }
diff --git a/src/simd/simd_test.go b/src/simd/simd_test.go
index 14e5fe31794..276ae9ed5d6 100644
--- a/src/simd/simd_test.go
+++ b/src/simd/simd_test.go
@@ -460,3 +460,20 @@ func testMergeLocalswrapper(t *testing.T, op func(simd.Int64x4, simd.Int64x4) si
 		}
 	}
 }
+
+func TestBitMask(t *testing.T) {
+	if !simd.HasAVX512() {
+		t.Skip("Test requires HasAVX512, not available on this hardware")
+		return
+	}
+	var bits uint64 = 0b10
+	results := [2]int64{}
+	want := [2]int64{0, 6}
+	m := simd.LoadMask64x2FromBits(&bits)
+	simd.LoadInt64x2Slice([]int64{1, 2}).AddMasked(simd.LoadInt64x2Slice([]int64{3, 4}), m).Store(&results)
+	for i := range 2 {
+		if results[i] != want[i] {
+			t.Errorf("Result at %d incorrect: want %v, got %v", i, want[i], results[i])
+		}
+	}
+}
diff --git a/src/simd/types_amd64.go b/src/simd/types_amd64.go
index 6cc79275767..ccc8427bb3e 100644
--- a/src/simd/types_amd64.go
+++ b/src/simd/types_amd64.go
@@ -205,24 +205,48 @@ type Mask8x16 struct {
 	vals    [16]int8
 }
 
+// Mask8x16FromBits constructs a Mask8x16 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 16 bits of y are used.
+//
+//go:noescape
+func LoadMask8x16FromBits(y *uint64) Mask8x16
+
 // Mask16x8 is a 128-bit SIMD vector of 8 int16
 type Mask16x8 struct {
 	int16x8 v128
 	vals    [8]int16
 }
 
+// Mask16x8FromBits constructs a Mask16x8 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 8 bits of y are used.
+//
+//go:noescape
+func LoadMask16x8FromBits(y *uint64) Mask16x8
+
 // Mask32x4 is a 128-bit SIMD vector of 4 int32
 type Mask32x4 struct {
 	int32x4 v128
 	vals    [4]int32
 }
 
+// Mask32x4FromBits constructs a Mask32x4 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 4 bits of y are used.
+//
+//go:noescape
+func LoadMask32x4FromBits(y *uint64) Mask32x4
+
 // Mask64x2 is a 128-bit SIMD vector of 2 int64
 type Mask64x2 struct {
 	int64x2 v128
 	vals    [2]int64
 }
 
+// Mask64x2FromBits constructs a Mask64x2 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 2 bits of y are used.
+//
+//go:noescape
+func LoadMask64x2FromBits(y *uint64) Mask64x2
+
 // v256 is a tag type that tells the compiler that this is really 256-bit SIMD
 type v256 struct {
 	_256 struct{}
@@ -424,24 +448,48 @@ type Mask8x32 struct {
 	vals    [32]int8
 }
 
+// Mask8x32FromBits constructs a Mask8x32 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 32 bits of y are used.
+//
+//go:noescape
+func LoadMask8x32FromBits(y *uint64) Mask8x32
+
 // Mask16x16 is a 256-bit SIMD vector of 16 int16
 type Mask16x16 struct {
 	int16x16 v256
 	vals     [16]int16
 }
 
+// Mask16x16FromBits constructs a Mask16x16 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 16 bits of y are used.
+//
+//go:noescape
+func LoadMask16x16FromBits(y *uint64) Mask16x16
+
 // Mask32x8 is a 256-bit SIMD vector of 8 int32
 type Mask32x8 struct {
 	int32x8 v256
 	vals    [8]int32
 }
 
+// Mask32x8FromBits constructs a Mask32x8 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 8 bits of y are used.
+//
+//go:noescape
+func LoadMask32x8FromBits(y *uint64) Mask32x8
+
 // Mask64x4 is a 256-bit SIMD vector of 4 int64
 type Mask64x4 struct {
 	int64x4 v256
 	vals    [4]int64
 }
 
+// Mask64x4FromBits constructs a Mask64x4 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 4 bits of y are used.
+//
+//go:noescape
+func LoadMask64x4FromBits(y *uint64) Mask64x4
+
 // v512 is a tag type that tells the compiler that this is really 512-bit SIMD
 type v512 struct {
 	_512 struct{}
@@ -643,20 +691,44 @@ type Mask8x64 struct {
 	vals    [64]int8
 }
 
+// Mask8x64FromBits constructs a Mask8x64 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 64 bits of y are used.
+//
+//go:noescape
+func LoadMask8x64FromBits(y *uint64) Mask8x64
+
 // Mask16x32 is a 512-bit SIMD vector of 32 int16
 type Mask16x32 struct {
 	int16x32 v512
 	vals     [32]int16
 }
 
+// Mask16x32FromBits constructs a Mask16x32 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 32 bits of y are used.
+//
+//go:noescape
+func LoadMask16x32FromBits(y *uint64) Mask16x32
+
 // Mask32x16 is a 512-bit SIMD vector of 16 int32
 type Mask32x16 struct {
 	int32x16 v512
 	vals     [16]int32
 }
 
+// Mask32x16FromBits constructs a Mask32x16 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 16 bits of y are used.
+//
+//go:noescape
+func LoadMask32x16FromBits(y *uint64) Mask32x16
+
 // Mask64x8 is a 512-bit SIMD vector of 8 int64
 type Mask64x8 struct {
 	int64x8 v512
 	vals    [8]int64
 }
+
+// Mask64x8FromBits constructs a Mask64x8 from an a bitmap, where 1 means set for the indexed element, 0 means unset.
+// Only the lower 8 bits of y are used.
+//
+//go:noescape
+func LoadMask64x8FromBits(y *uint64) Mask64x8