[dev.simd] simd, cmd/compile: add float -> float conversions

This should mark the end of the conversion table, except for float16
which does not exist on Go yet.

The rounding logic documentation of float64 -> float32 is based on
abi-internal default MXCSR:
| RC | 14/13 | 0 (RN) | Round to nearest |

Change-Id: I27a86560e8d74d20f21350bf78314b4eada20ec0
Reviewed-on: https://go-review.googlesource.com/c/go/+/724440
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: David Chase <drchase@google.com>

7 files changed, 826 insertions, 0 deletions

diff --git a/src/cmd/compile/internal/amd64/simdssa.go b/src/cmd/compile/internal/amd64/simdssa.go
index 71b98139d5..465fb980a5 100644
--- a/src/cmd/compile/internal/amd64/simdssa.go
+++ b/src/cmd/compile/internal/amd64/simdssa.go
@@ -42,6 +42,9 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VPBROADCASTW512,
 		ssa.OpAMD64VPBROADCASTD512,
 		ssa.OpAMD64VPBROADCASTQ512,
+		ssa.OpAMD64VCVTPD2PSX128,
+		ssa.OpAMD64VCVTPD2PSY128,
+		ssa.OpAMD64VCVTPD2PS256,
 		ssa.OpAMD64VCVTDQ2PS128,
 		ssa.OpAMD64VCVTDQ2PS256,
 		ssa.OpAMD64VCVTDQ2PS512,
@@ -54,6 +57,8 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VCVTUQQ2PSX128,
 		ssa.OpAMD64VCVTUQQ2PSY128,
 		ssa.OpAMD64VCVTUQQ2PS256,
+		ssa.OpAMD64VCVTPS2PD256,
+		ssa.OpAMD64VCVTPS2PD512,
 		ssa.OpAMD64VCVTDQ2PD256,
 		ssa.OpAMD64VCVTDQ2PD512,
 		ssa.OpAMD64VCVTQQ2PD128,
@@ -854,6 +859,9 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VPCOMPRESSQMasked128,
 		ssa.OpAMD64VPCOMPRESSQMasked256,
 		ssa.OpAMD64VPCOMPRESSQMasked512,
+		ssa.OpAMD64VCVTPD2PSXMasked128,
+		ssa.OpAMD64VCVTPD2PSYMasked128,
+		ssa.OpAMD64VCVTPD2PSMasked256,
 		ssa.OpAMD64VCVTDQ2PSMasked128,
 		ssa.OpAMD64VCVTDQ2PSMasked256,
 		ssa.OpAMD64VCVTDQ2PSMasked512,
@@ -866,6 +874,8 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VCVTUQQ2PSXMasked128,
 		ssa.OpAMD64VCVTUQQ2PSYMasked128,
 		ssa.OpAMD64VCVTUQQ2PSMasked256,
+		ssa.OpAMD64VCVTPS2PDMasked256,
+		ssa.OpAMD64VCVTPS2PDMasked512,
 		ssa.OpAMD64VCVTDQ2PDMasked256,
 		ssa.OpAMD64VCVTDQ2PDMasked512,
 		ssa.OpAMD64VCVTQQ2PDMasked128,
@@ -1764,6 +1774,9 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VPABSQMasked128load,
 		ssa.OpAMD64VPABSQMasked256load,
 		ssa.OpAMD64VPABSQMasked512load,
+		ssa.OpAMD64VCVTPD2PSXMasked128load,
+		ssa.OpAMD64VCVTPD2PSYMasked128load,
+		ssa.OpAMD64VCVTPD2PSMasked256load,
 		ssa.OpAMD64VCVTDQ2PSMasked128load,
 		ssa.OpAMD64VCVTDQ2PSMasked256load,
 		ssa.OpAMD64VCVTDQ2PSMasked512load,
@@ -1776,6 +1789,8 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VCVTUQQ2PSXMasked128load,
 		ssa.OpAMD64VCVTUQQ2PSYMasked128load,
 		ssa.OpAMD64VCVTUQQ2PSMasked256load,
+		ssa.OpAMD64VCVTPS2PDMasked256load,
+		ssa.OpAMD64VCVTPS2PDMasked512load,
 		ssa.OpAMD64VCVTDQ2PDMasked256load,
 		ssa.OpAMD64VCVTDQ2PDMasked512load,
 		ssa.OpAMD64VCVTQQ2PDMasked128load,
@@ -2190,6 +2205,7 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VPABSQ128load,
 		ssa.OpAMD64VPABSQ256load,
 		ssa.OpAMD64VPABSQ512load,
+		ssa.OpAMD64VCVTPD2PS256load,
 		ssa.OpAMD64VCVTDQ2PS512load,
 		ssa.OpAMD64VCVTQQ2PSX128load,
 		ssa.OpAMD64VCVTQQ2PSY128load,
@@ -2200,6 +2216,7 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VCVTUQQ2PSX128load,
 		ssa.OpAMD64VCVTUQQ2PSY128load,
 		ssa.OpAMD64VCVTUQQ2PS256load,
+		ssa.OpAMD64VCVTPS2PD512load,
 		ssa.OpAMD64VCVTDQ2PD512load,
 		ssa.OpAMD64VCVTQQ2PD128load,
 		ssa.OpAMD64VCVTQQ2PD256load,
@@ -2473,6 +2490,9 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VREDUCEPDMasked128Merging,
 		ssa.OpAMD64VREDUCEPDMasked256Merging,
 		ssa.OpAMD64VREDUCEPDMasked512Merging,
+		ssa.OpAMD64VCVTPD2PSXMasked128Merging,
+		ssa.OpAMD64VCVTPD2PSYMasked128Merging,
+		ssa.OpAMD64VCVTPD2PSMasked256Merging,
 		ssa.OpAMD64VCVTDQ2PSMasked128Merging,
 		ssa.OpAMD64VCVTDQ2PSMasked256Merging,
 		ssa.OpAMD64VCVTDQ2PSMasked512Merging,
@@ -2485,6 +2505,8 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VCVTUQQ2PSXMasked128Merging,
 		ssa.OpAMD64VCVTUQQ2PSYMasked128Merging,
 		ssa.OpAMD64VCVTUQQ2PSMasked256Merging,
+		ssa.OpAMD64VCVTPS2PDMasked256Merging,
+		ssa.OpAMD64VCVTPS2PDMasked512Merging,
 		ssa.OpAMD64VCVTDQ2PDMasked256Merging,
 		ssa.OpAMD64VCVTDQ2PDMasked512Merging,
 		ssa.OpAMD64VCVTQQ2PDMasked128Merging,
@@ -2880,6 +2902,12 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VPALIGNRMasked256,
 		ssa.OpAMD64VPALIGNRMasked512,
 		ssa.OpAMD64VPALIGNRMasked128,
+		ssa.OpAMD64VCVTPD2PSXMasked128,
+		ssa.OpAMD64VCVTPD2PSXMasked128load,
+		ssa.OpAMD64VCVTPD2PSYMasked128,
+		ssa.OpAMD64VCVTPD2PSYMasked128load,
+		ssa.OpAMD64VCVTPD2PSMasked256,
+		ssa.OpAMD64VCVTPD2PSMasked256load,
 		ssa.OpAMD64VCVTDQ2PSMasked128,
 		ssa.OpAMD64VCVTDQ2PSMasked128load,
 		ssa.OpAMD64VCVTDQ2PSMasked256,
@@ -2904,6 +2932,10 @@ func ssaGenSIMDValue(s *ssagen.State, v *ssa.Value) bool {
 		ssa.OpAMD64VCVTUQQ2PSYMasked128load,
 		ssa.OpAMD64VCVTUQQ2PSMasked256,
 		ssa.OpAMD64VCVTUQQ2PSMasked256load,
+		ssa.OpAMD64VCVTPS2PDMasked256,
+		ssa.OpAMD64VCVTPS2PDMasked256load,
+		ssa.OpAMD64VCVTPS2PDMasked512,
+		ssa.OpAMD64VCVTPS2PDMasked512load,
 		ssa.OpAMD64VCVTDQ2PDMasked256,
 		ssa.OpAMD64VCVTDQ2PDMasked256load,
 		ssa.OpAMD64VCVTDQ2PDMasked512,
diff --git a/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules b/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules
index 80a52969cb..752fb9d9b0 100644
--- a/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules
+++ b/src/cmd/compile/internal/ssa/_gen/simdAMD64.rules
@@ -249,6 +249,9 @@
 (ConcatShiftBytesRightUint8x16 ...) => (VPALIGNR128 ...)
 (ConcatShiftBytesRightGroupedUint8x32 ...) => (VPALIGNR256 ...)
 (ConcatShiftBytesRightGroupedUint8x64 ...) => (VPALIGNR512 ...)
+(ConvertToFloat32Float64x2 ...) => (VCVTPD2PSX128 ...)
+(ConvertToFloat32Float64x4 ...) => (VCVTPD2PSY128 ...)
+(ConvertToFloat32Float64x8 ...) => (VCVTPD2PS256 ...)
 (ConvertToFloat32Int32x4 ...) => (VCVTDQ2PS128 ...)
 (ConvertToFloat32Int32x8 ...) => (VCVTDQ2PS256 ...)
 (ConvertToFloat32Int32x16 ...) => (VCVTDQ2PS512 ...)
@@ -261,6 +264,8 @@
 (ConvertToFloat32Uint64x2 ...) => (VCVTUQQ2PSX128 ...)
 (ConvertToFloat32Uint64x4 ...) => (VCVTUQQ2PSY128 ...)
 (ConvertToFloat32Uint64x8 ...) => (VCVTUQQ2PS256 ...)
+(ConvertToFloat64Float32x4 ...) => (VCVTPS2PD256 ...)
+(ConvertToFloat64Float32x8 ...) => (VCVTPS2PD512 ...)
 (ConvertToFloat64Int32x4 ...) => (VCVTDQ2PD256 ...)
 (ConvertToFloat64Int32x8 ...) => (VCVTDQ2PD512 ...)
 (ConvertToFloat64Int64x2 ...) => (VCVTQQ2PD128 ...)
@@ -1478,6 +1483,9 @@
 (VMOVDQU8Masked256 (VPALIGNR256 [a] x y) mask) => (VPALIGNRMasked256 [a] x y mask)
 (VMOVDQU8Masked512 (VPALIGNR512 [a] x y) mask) => (VPALIGNRMasked512 [a] x y mask)
 (VMOVDQU8Masked128 (VPALIGNR128 [a] x y) mask) => (VPALIGNRMasked128 [a] x y mask)
+(VMOVDQU64Masked128 (VCVTPD2PSX128 x) mask) => (VCVTPD2PSXMasked128 x mask)
+(VMOVDQU64Masked128 (VCVTPD2PSY128 x) mask) => (VCVTPD2PSYMasked128 x mask)
+(VMOVDQU64Masked256 (VCVTPD2PS256 x) mask) => (VCVTPD2PSMasked256 x mask)
 (VMOVDQU32Masked128 (VCVTDQ2PS128 x) mask) => (VCVTDQ2PSMasked128 x mask)
 (VMOVDQU32Masked256 (VCVTDQ2PS256 x) mask) => (VCVTDQ2PSMasked256 x mask)
 (VMOVDQU32Masked512 (VCVTDQ2PS512 x) mask) => (VCVTDQ2PSMasked512 x mask)
@@ -1490,6 +1498,8 @@
 (VMOVDQU64Masked128 (VCVTUQQ2PSX128 x) mask) => (VCVTUQQ2PSXMasked128 x mask)
 (VMOVDQU64Masked128 (VCVTUQQ2PSY128 x) mask) => (VCVTUQQ2PSYMasked128 x mask)
 (VMOVDQU64Masked256 (VCVTUQQ2PS256 x) mask) => (VCVTUQQ2PSMasked256 x mask)
+(VMOVDQU32Masked256 (VCVTPS2PD256 x) mask) => (VCVTPS2PDMasked256 x mask)
+(VMOVDQU32Masked512 (VCVTPS2PD512 x) mask) => (VCVTPS2PDMasked512 x mask)
 (VMOVDQU32Masked256 (VCVTDQ2PD256 x) mask) => (VCVTDQ2PDMasked256 x mask)
 (VMOVDQU32Masked512 (VCVTDQ2PD512 x) mask) => (VCVTDQ2PDMasked512 x mask)
 (VMOVDQU64Masked128 (VCVTQQ2PD128 x) mask) => (VCVTQQ2PDMasked128 x mask)
@@ -2031,6 +2041,7 @@
 (VPBLENDMDMasked512 dst (VSQRTPS512 x) mask) => (VSQRTPSMasked512Merging dst x mask)
 (VPBLENDMDMasked512 dst (VSUBPS512 x y) mask) => (VSUBPSMasked512Merging dst x y mask)
 (VPBLENDMQMasked512 dst (VADDPD512 x y) mask) => (VADDPDMasked512Merging dst x y mask)
+(VPBLENDMQMasked512 dst (VCVTPD2PS256 x) mask) => (VCVTPD2PSMasked256Merging dst x mask)
 (VPBLENDMQMasked512 dst (VCVTQQ2PD512 x) mask) => (VCVTQQ2PDMasked512Merging dst x mask)
 (VPBLENDMQMasked512 dst (VCVTQQ2PS256 x) mask) => (VCVTQQ2PSMasked256Merging dst x mask)
 (VPBLENDMQMasked512 dst (VCVTTPD2DQ256 x) mask) => (VCVTTPD2DQMasked256Merging dst x mask)
@@ -2121,6 +2132,8 @@
 (VPBLENDVB128 dst (VBROADCASTSS512 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VBROADCASTSSMasked512Merging dst x (VPMOVVec32x4ToM <types.TypeMask> mask))
 (VPBLENDVB128 dst (VCVTDQ2PD256 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTDQ2PDMasked256Merging dst x (VPMOVVec32x4ToM <types.TypeMask> mask))
 (VPBLENDVB128 dst (VCVTDQ2PS128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTDQ2PSMasked128Merging dst x (VPMOVVec32x4ToM <types.TypeMask> mask))
+(VPBLENDVB128 dst (VCVTPD2PSX128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTPD2PSXMasked128Merging dst x (VPMOVVec64x2ToM <types.TypeMask> mask))
+(VPBLENDVB128 dst (VCVTPS2PD256 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTPS2PDMasked256Merging dst x (VPMOVVec32x4ToM <types.TypeMask> mask))
 (VPBLENDVB128 dst (VCVTQQ2PD128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTQQ2PDMasked128Merging dst x (VPMOVVec64x2ToM <types.TypeMask> mask))
 (VPBLENDVB128 dst (VCVTQQ2PSX128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTQQ2PSXMasked128Merging dst x (VPMOVVec64x2ToM <types.TypeMask> mask))
 (VPBLENDVB128 dst (VCVTTPD2DQX128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTTPD2DQXMasked128Merging dst x (VPMOVVec64x2ToM <types.TypeMask> mask))
@@ -2304,6 +2317,8 @@
 (VPBLENDVB256 dst (VADDPS256 x y) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VADDPSMasked256Merging dst x y (VPMOVVec32x8ToM <types.TypeMask> mask))
 (VPBLENDVB256 dst (VCVTDQ2PD512 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTDQ2PDMasked512Merging dst x (VPMOVVec32x8ToM <types.TypeMask> mask))
 (VPBLENDVB256 dst (VCVTDQ2PS256 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTDQ2PSMasked256Merging dst x (VPMOVVec32x8ToM <types.TypeMask> mask))
+(VPBLENDVB256 dst (VCVTPD2PSY128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTPD2PSYMasked128Merging dst x (VPMOVVec64x4ToM <types.TypeMask> mask))
+(VPBLENDVB256 dst (VCVTPS2PD512 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTPS2PDMasked512Merging dst x (VPMOVVec32x8ToM <types.TypeMask> mask))
 (VPBLENDVB256 dst (VCVTQQ2PD256 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTQQ2PDMasked256Merging dst x (VPMOVVec64x4ToM <types.TypeMask> mask))
 (VPBLENDVB256 dst (VCVTQQ2PSY128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTQQ2PSYMasked128Merging dst x (VPMOVVec64x4ToM <types.TypeMask> mask))
 (VPBLENDVB256 dst (VCVTTPD2DQY128 x) mask) && v.Block.CPUfeatures.hasFeature(CPUavx512) => (VCVTTPD2DQYMasked128Merging dst x (VPMOVVec64x4ToM <types.TypeMask> mask))
@@ -2541,6 +2556,7 @@
 (VPERMI2QMasked256 x y l:(VMOVDQUload256 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VPERMI2QMasked256load {sym} [off] x y ptr mask mem)
 (VPERMI2PDMasked512 x y l:(VMOVDQUload512 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VPERMI2PDMasked512load {sym} [off] x y ptr mask mem)
 (VPERMI2QMasked512 x y l:(VMOVDQUload512 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VPERMI2QMasked512load {sym} [off] x y ptr mask mem)
+(VCVTPD2PS256 l:(VMOVDQUload512 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTPD2PS256load {sym} [off] ptr mem)
 (VCVTDQ2PS512 l:(VMOVDQUload512 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTDQ2PS512load {sym} [off] ptr mem)
 (VCVTQQ2PSX128 l:(VMOVDQUload128 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTQQ2PSX128load {sym} [off] ptr mem)
 (VCVTQQ2PSY128 l:(VMOVDQUload256 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTQQ2PSY128load {sym} [off] ptr mem)
@@ -2551,6 +2567,9 @@
 (VCVTUQQ2PSX128 l:(VMOVDQUload128 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PSX128load {sym} [off] ptr mem)
 (VCVTUQQ2PSY128 l:(VMOVDQUload256 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PSY128load {sym} [off] ptr mem)
 (VCVTUQQ2PS256 l:(VMOVDQUload512 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PS256load {sym} [off] ptr mem)
+(VCVTPD2PSXMasked128 l:(VMOVDQUload128 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTPD2PSXMasked128load {sym} [off] ptr mask mem)
+(VCVTPD2PSYMasked128 l:(VMOVDQUload256 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTPD2PSYMasked128load {sym} [off] ptr mask mem)
+(VCVTPD2PSMasked256 l:(VMOVDQUload512 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTPD2PSMasked256load {sym} [off] ptr mask mem)
 (VCVTDQ2PSMasked128 l:(VMOVDQUload128 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTDQ2PSMasked128load {sym} [off] ptr mask mem)
 (VCVTDQ2PSMasked256 l:(VMOVDQUload256 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTDQ2PSMasked256load {sym} [off] ptr mask mem)
 (VCVTDQ2PSMasked512 l:(VMOVDQUload512 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTDQ2PSMasked512load {sym} [off] ptr mask mem)
@@ -2563,6 +2582,7 @@
 (VCVTUQQ2PSXMasked128 l:(VMOVDQUload128 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PSXMasked128load {sym} [off] ptr mask mem)
 (VCVTUQQ2PSYMasked128 l:(VMOVDQUload256 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PSYMasked128load {sym} [off] ptr mask mem)
 (VCVTUQQ2PSMasked256 l:(VMOVDQUload512 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PSMasked256load {sym} [off] ptr mask mem)
+(VCVTPS2PD512 l:(VMOVDQUload256 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTPS2PD512load {sym} [off] ptr mem)
 (VCVTDQ2PD512 l:(VMOVDQUload256 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTDQ2PD512load {sym} [off] ptr mem)
 (VCVTQQ2PD128 l:(VMOVDQUload128 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTQQ2PD128load {sym} [off] ptr mem)
 (VCVTQQ2PD256 l:(VMOVDQUload256 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTQQ2PD256load {sym} [off] ptr mem)
@@ -2572,6 +2592,8 @@
 (VCVTUQQ2PD128 l:(VMOVDQUload128 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PD128load {sym} [off] ptr mem)
 (VCVTUQQ2PD256 l:(VMOVDQUload256 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PD256load {sym} [off] ptr mem)
 (VCVTUQQ2PD512 l:(VMOVDQUload512 {sym} [off] ptr mem)) && canMergeLoad(v, l) && clobber(l) => (VCVTUQQ2PD512load {sym} [off] ptr mem)
+(VCVTPS2PDMasked256 l:(VMOVDQUload128 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTPS2PDMasked256load {sym} [off] ptr mask mem)
+(VCVTPS2PDMasked512 l:(VMOVDQUload256 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTPS2PDMasked512load {sym} [off] ptr mask mem)
 (VCVTDQ2PDMasked256 l:(VMOVDQUload128 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTDQ2PDMasked256load {sym} [off] ptr mask mem)
 (VCVTDQ2PDMasked512 l:(VMOVDQUload256 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTDQ2PDMasked512load {sym} [off] ptr mask mem)
 (VCVTQQ2PDMasked128 l:(VMOVDQUload128 {sym} [off] ptr mem) mask) && canMergeLoad(v, l) && clobber(l) => (VCVTQQ2PDMasked128load {sym} [off] ptr mask mem)
diff --git a/src/cmd/compile/internal/ssa/_gen/simdAMD64ops.go b/src/cmd/compile/internal/ssa/_gen/simdAMD64ops.go
index 0bf39a7981..82f5bfdb3f 100644
--- a/src/cmd/compile/internal/ssa/_gen/simdAMD64ops.go
+++ b/src/cmd/compile/internal/ssa/_gen/simdAMD64ops.go
@@ -66,6 +66,16 @@ func simdAMD64Ops(v11, v21, v2k, vkv, v2kv, v2kk, v31, v3kv, vgpv, vgp, vfpv, vf
 		{name: "VCVTDQ2PSMasked128", argLength: 2, reg: wkw, asm: "VCVTDQ2PS", commutative: false, typ: "Vec128", resultInArg0: false},
 		{name: "VCVTDQ2PSMasked256", argLength: 2, reg: wkw, asm: "VCVTDQ2PS", commutative: false, typ: "Vec256", resultInArg0: false},
 		{name: "VCVTDQ2PSMasked512", argLength: 2, reg: wkw, asm: "VCVTDQ2PS", commutative: false, typ: "Vec512", resultInArg0: false},
+		{name: "VCVTPD2PS256", argLength: 1, reg: w11, asm: "VCVTPD2PS", commutative: false, typ: "Vec256", resultInArg0: false},
+		{name: "VCVTPD2PSMasked256", argLength: 2, reg: wkw, asm: "VCVTPD2PS", commutative: false, typ: "Vec256", resultInArg0: false},
+		{name: "VCVTPD2PSX128", argLength: 1, reg: v11, asm: "VCVTPD2PSX", commutative: false, typ: "Vec128", resultInArg0: false},
+		{name: "VCVTPD2PSXMasked128", argLength: 2, reg: wkw, asm: "VCVTPD2PSX", commutative: false, typ: "Vec128", resultInArg0: false},
+		{name: "VCVTPD2PSY128", argLength: 1, reg: v11, asm: "VCVTPD2PSY", commutative: false, typ: "Vec128", resultInArg0: false},
+		{name: "VCVTPD2PSYMasked128", argLength: 2, reg: wkw, asm: "VCVTPD2PSY", commutative: false, typ: "Vec128", resultInArg0: false},
+		{name: "VCVTPS2PD256", argLength: 1, reg: v11, asm: "VCVTPS2PD", commutative: false, typ: "Vec256", resultInArg0: false},
+		{name: "VCVTPS2PD512", argLength: 1, reg: w11, asm: "VCVTPS2PD", commutative: false, typ: "Vec512", resultInArg0: false},
+		{name: "VCVTPS2PDMasked256", argLength: 2, reg: wkw, asm: "VCVTPS2PD", commutative: false, typ: "Vec256", resultInArg0: false},
+		{name: "VCVTPS2PDMasked512", argLength: 2, reg: wkw, asm: "VCVTPS2PD", commutative: false, typ: "Vec512", resultInArg0: false},
 		{name: "VCVTQQ2PD128", argLength: 1, reg: w11, asm: "VCVTQQ2PD", commutative: false, typ: "Vec128", resultInArg0: false},
 		{name: "VCVTQQ2PD256", argLength: 1, reg: w11, asm: "VCVTQQ2PD", commutative: false, typ: "Vec256", resultInArg0: false},
 		{name: "VCVTQQ2PD512", argLength: 1, reg: w11, asm: "VCVTQQ2PD", commutative: false, typ: "Vec512", resultInArg0: false},
@@ -1488,6 +1498,13 @@ func simdAMD64Ops(v11, v21, v2k, vkv, v2kv, v2kk, v31, v3kv, vgpv, vgp, vfpv, vf
 		{name: "VCVTDQ2PSMasked128load", argLength: 3, reg: wkwload, asm: "VCVTDQ2PS", commutative: false, typ: "Vec128", aux: "SymOff", symEffect: "Read", resultInArg0: false},
 		{name: "VCVTDQ2PSMasked256load", argLength: 3, reg: wkwload, asm: "VCVTDQ2PS", commutative: false, typ: "Vec256", aux: "SymOff", symEffect: "Read", resultInArg0: false},
 		{name: "VCVTDQ2PSMasked512load", argLength: 3, reg: wkwload, asm: "VCVTDQ2PS", commutative: false, typ: "Vec512", aux: "SymOff", symEffect: "Read", resultInArg0: false},
+		{name: "VCVTPD2PS256load", argLength: 2, reg: w11load, asm: "VCVTPD2PS", commutative: false, typ: "Vec256", aux: "SymOff", symEffect: "Read", resultInArg0: false},
+		{name: "VCVTPD2PSMasked256load", argLength: 3, reg: wkwload, asm: "VCVTPD2PS", commutative: false, typ: "Vec256", aux: "SymOff", symEffect: "Read", resultInArg0: false},
+		{name: "VCVTPD2PSXMasked128load", argLength: 3, reg: wkwload, asm: "VCVTPD2PSX", commutative: false, typ: "Vec128", aux: "SymOff", symEffect: "Read", resultInArg0: false},
+		{name: "VCVTPD2PSYMasked128load", argLength: 3, reg: wkwload, asm: "VCVTPD2PSY", commutative: false, typ: "Vec128", aux: "SymOff", symEffect: "Read", resultInArg0: false},
+		{name: "VCVTPS2PD512load", argLength: 2, reg: w11load, asm: "VCVTPS2PD", commutative: false, typ: "Vec512", aux: "SymOff", symEffect: "Read", resultInArg0: false},
+		{name: "VCVTPS2PDMasked256load", argLength: 3, reg: wkwload, asm: "VCVTPS2PD", commutative: false, typ: "Vec256", aux: "SymOff", symEffect: "Read", resultInArg0: false},
+		{name: "VCVTPS2PDMasked512load", argLength: 3, reg: wkwload, asm: "VCVTPS2PD", commutative: false, typ: "Vec512", aux: "SymOff", symEffect: "Read", resultInArg0: false},
 		{name: "VCVTQQ2PD128load", argLength: 2, reg: w11load, asm: "VCVTQQ2PD", commutative: false, typ: "Vec128", aux: "SymOff", symEffect: "Read", resultInArg0: false},
 		{name: "VCVTQQ2PD256load", argLength: 2, reg: w11load, asm: "VCVTQQ2PD", commutative: false, typ: "Vec256", aux: "SymOff", symEffect: "Read", resultInArg0: false},
 		{name: "VCVTQQ2PD512load", argLength: 2, reg: w11load, asm: "VCVTQQ2PD", commutative: false, typ: "Vec512", aux: "SymOff", symEffect: "Read", resultInArg0: false},
@@ -2114,6 +2131,11 @@ func simdAMD64Ops(v11, v21, v2k, vkv, v2kv, v2kk, v31, v3kv, vgpv, vgp, vfpv, vf
 		{name: "VCVTDQ2PSMasked128Merging", argLength: 3, reg: w2kw, asm: "VCVTDQ2PS", commutative: false, typ: "Vec128", resultInArg0: true},
 		{name: "VCVTDQ2PSMasked256Merging", argLength: 3, reg: w2kw, asm: "VCVTDQ2PS", commutative: false, typ: "Vec256", resultInArg0: true},
 		{name: "VCVTDQ2PSMasked512Merging", argLength: 3, reg: w2kw, asm: "VCVTDQ2PS", commutative: false, typ: "Vec512", resultInArg0: true},
+		{name: "VCVTPD2PSMasked256Merging", argLength: 3, reg: w2kw, asm: "VCVTPD2PS", commutative: false, typ: "Vec256", resultInArg0: true},
+		{name: "VCVTPD2PSXMasked128Merging", argLength: 3, reg: w2kw, asm: "VCVTPD2PSX", commutative: false, typ: "Vec128", resultInArg0: true},
+		{name: "VCVTPD2PSYMasked128Merging", argLength: 3, reg: w2kw, asm: "VCVTPD2PSY", commutative: false, typ: "Vec128", resultInArg0: true},
+		{name: "VCVTPS2PDMasked256Merging", argLength: 3, reg: w2kw, asm: "VCVTPS2PD", commutative: false, typ: "Vec256", resultInArg0: true},
+		{name: "VCVTPS2PDMasked512Merging", argLength: 3, reg: w2kw, asm: "VCVTPS2PD", commutative: false, typ: "Vec512", resultInArg0: true},
 		{name: "VCVTQQ2PDMasked128Merging", argLength: 3, reg: w2kw, asm: "VCVTQQ2PD", commutative: false, typ: "Vec128", resultInArg0: true},
 		{name: "VCVTQQ2PDMasked256Merging", argLength: 3, reg: w2kw, asm: "VCVTQQ2PD", commutative: false, typ: "Vec256", resultInArg0: true},
 		{name: "VCVTQQ2PDMasked512Merging", argLength: 3, reg: w2kw, asm: "VCVTQQ2PD", commutative: false, typ: "Vec512", resultInArg0: true},
diff --git a/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go b/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go
index 4f0183f9d3..53bd208e34 100644
--- a/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go
+++ b/src/cmd/compile/internal/ssa/_gen/simdgenericOps.go
@@ -237,6 +237,9 @@ func simdGenericOps() []opData {
 		{name: "ConcatPermuteUint64x2", argLength: 3, commutative: false},
 		{name: "ConcatPermuteUint64x4", argLength: 3, commutative: false},
 		{name: "ConcatPermuteUint64x8", argLength: 3, commutative: false},
+		{name: "ConvertToFloat32Float64x2", argLength: 1, commutative: false},
+		{name: "ConvertToFloat32Float64x4", argLength: 1, commutative: false},
+		{name: "ConvertToFloat32Float64x8", argLength: 1, commutative: false},
 		{name: "ConvertToFloat32Int32x4", argLength: 1, commutative: false},
 		{name: "ConvertToFloat32Int32x8", argLength: 1, commutative: false},
 		{name: "ConvertToFloat32Int32x16", argLength: 1, commutative: false},
@@ -249,6 +252,8 @@ func simdGenericOps() []opData {
 		{name: "ConvertToFloat32Uint64x2", argLength: 1, commutative: false},
 		{name: "ConvertToFloat32Uint64x4", argLength: 1, commutative: false},
 		{name: "ConvertToFloat32Uint64x8", argLength: 1, commutative: false},
+		{name: "ConvertToFloat64Float32x4", argLength: 1, commutative: false},
+		{name: "ConvertToFloat64Float32x8", argLength: 1, commutative: false},
 		{name: "ConvertToFloat64Int32x4", argLength: 1, commutative: false},
 		{name: "ConvertToFloat64Int32x8", argLength: 1, commutative: false},
 		{name: "ConvertToFloat64Int64x2", argLength: 1, commutative: false},
diff --git a/src/cmd/compile/internal/ssa/opGen.go b/src/cmd/compile/internal/ssa/opGen.go
index 634872c6d4..3e589f8738 100644
--- a/src/cmd/compile/internal/ssa/opGen.go
+++ b/src/cmd/compile/internal/ssa/opGen.go
@@ -1307,6 +1307,16 @@ const (
 	OpAMD64VCVTDQ2PSMasked128
 	OpAMD64VCVTDQ2PSMasked256
 	OpAMD64VCVTDQ2PSMasked512
+	OpAMD64VCVTPD2PS256
+	OpAMD64VCVTPD2PSMasked256
+	OpAMD64VCVTPD2PSX128
+	OpAMD64VCVTPD2PSXMasked128
+	OpAMD64VCVTPD2PSY128
+	OpAMD64VCVTPD2PSYMasked128
+	OpAMD64VCVTPS2PD256
+	OpAMD64VCVTPS2PD512
+	OpAMD64VCVTPS2PDMasked256
+	OpAMD64VCVTPS2PDMasked512
 	OpAMD64VCVTQQ2PD128
 	OpAMD64VCVTQQ2PD256
 	OpAMD64VCVTQQ2PD512
@@ -2729,6 +2739,13 @@ const (
 	OpAMD64VCVTDQ2PSMasked128load
 	OpAMD64VCVTDQ2PSMasked256load
 	OpAMD64VCVTDQ2PSMasked512load
+	OpAMD64VCVTPD2PS256load
+	OpAMD64VCVTPD2PSMasked256load
+	OpAMD64VCVTPD2PSXMasked128load
+	OpAMD64VCVTPD2PSYMasked128load
+	OpAMD64VCVTPS2PD512load
+	OpAMD64VCVTPS2PDMasked256load
+	OpAMD64VCVTPS2PDMasked512load
 	OpAMD64VCVTQQ2PD128load
 	OpAMD64VCVTQQ2PD256load
 	OpAMD64VCVTQQ2PD512load
@@ -3355,6 +3372,11 @@ const (
 	OpAMD64VCVTDQ2PSMasked128Merging
 	OpAMD64VCVTDQ2PSMasked256Merging
 	OpAMD64VCVTDQ2PSMasked512Merging
+	OpAMD64VCVTPD2PSMasked256Merging
+	OpAMD64VCVTPD2PSXMasked128Merging
+	OpAMD64VCVTPD2PSYMasked128Merging
+	OpAMD64VCVTPS2PDMasked256Merging
+	OpAMD64VCVTPS2PDMasked512Merging
 	OpAMD64VCVTQQ2PDMasked128Merging
 	OpAMD64VCVTQQ2PDMasked256Merging
 	OpAMD64VCVTQQ2PDMasked512Merging
@@ -6360,6 +6382,9 @@ const (
 	OpConcatPermuteUint64x2
 	OpConcatPermuteUint64x4
 	OpConcatPermuteUint64x8
+	OpConvertToFloat32Float64x2
+	OpConvertToFloat32Float64x4
+	OpConvertToFloat32Float64x8
 	OpConvertToFloat32Int32x4
 	OpConvertToFloat32Int32x8
 	OpConvertToFloat32Int32x16
@@ -6372,6 +6397,8 @@ const (
 	OpConvertToFloat32Uint64x2
 	OpConvertToFloat32Uint64x4
 	OpConvertToFloat32Uint64x8
+	OpConvertToFloat64Float32x4
+	OpConvertToFloat64Float32x8
 	OpConvertToFloat64Int32x4
 	OpConvertToFloat64Int32x8
 	OpConvertToFloat64Int64x2
@@ -21612,6 +21639,141 @@ var opcodeTable = [...]opInfo{
 		},
 	},
 	{
+		name:   "VCVTPD2PS256",
+		argLen: 1,
+		asm:    x86.AVCVTPD2PS,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:   "VCVTPD2PSMasked256",
+		argLen: 2,
+		asm:    x86.AVCVTPD2PS,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:   "VCVTPD2PSX128",
+		argLen: 1,
+		asm:    x86.AVCVTPD2PSX,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 2147418112}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14
+			},
+			outputs: []outputInfo{
+				{0, 2147418112}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14
+			},
+		},
+	},
+	{
+		name:   "VCVTPD2PSXMasked128",
+		argLen: 2,
+		asm:    x86.AVCVTPD2PSX,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:   "VCVTPD2PSY128",
+		argLen: 1,
+		asm:    x86.AVCVTPD2PSY,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 2147418112}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14
+			},
+			outputs: []outputInfo{
+				{0, 2147418112}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14
+			},
+		},
+	},
+	{
+		name:   "VCVTPD2PSYMasked128",
+		argLen: 2,
+		asm:    x86.AVCVTPD2PSY,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:   "VCVTPS2PD256",
+		argLen: 1,
+		asm:    x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 2147418112}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14
+			},
+			outputs: []outputInfo{
+				{0, 2147418112}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14
+			},
+		},
+	},
+	{
+		name:   "VCVTPS2PD512",
+		argLen: 1,
+		asm:    x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:   "VCVTPS2PDMasked256",
+		argLen: 2,
+		asm:    x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:   "VCVTPS2PDMasked512",
+		argLen: 2,
+		asm:    x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
 		name:   "VCVTQQ2PD128",
 		argLen: 1,
 		asm:    x86.AVCVTQQ2PD,
@@ -42524,6 +42686,116 @@ var opcodeTable = [...]opInfo{
 		},
 	},
 	{
+		name:      "VCVTPD2PS256load",
+		auxType:   auxSymOff,
+		argLen:    2,
+		symEffect: SymRead,
+		asm:       x86.AVCVTPD2PS,
+		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, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:      "VCVTPD2PSMasked256load",
+		auxType:   auxSymOff,
+		argLen:    3,
+		symEffect: SymRead,
+		asm:       x86.AVCVTPD2PS,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 72057594037977087}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15 SB
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:      "VCVTPD2PSXMasked128load",
+		auxType:   auxSymOff,
+		argLen:    3,
+		symEffect: SymRead,
+		asm:       x86.AVCVTPD2PSX,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 72057594037977087}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15 SB
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:      "VCVTPD2PSYMasked128load",
+		auxType:   auxSymOff,
+		argLen:    3,
+		symEffect: SymRead,
+		asm:       x86.AVCVTPD2PSY,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 72057594037977087}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15 SB
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:      "VCVTPS2PD512load",
+		auxType:   auxSymOff,
+		argLen:    2,
+		symEffect: SymRead,
+		asm:       x86.AVCVTPS2PD,
+		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, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:      "VCVTPS2PDMasked256load",
+		auxType:   auxSymOff,
+		argLen:    3,
+		symEffect: SymRead,
+		asm:       x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 72057594037977087}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15 SB
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:      "VCVTPS2PDMasked512load",
+		auxType:   auxSymOff,
+		argLen:    3,
+		symEffect: SymRead,
+		asm:       x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{1, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 72057594037977087}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15 SB
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
 		name:      "VCVTQQ2PD128load",
 		auxType:   auxSymOff,
 		argLen:    2,
@@ -52891,6 +53163,86 @@ var opcodeTable = [...]opInfo{
 		},
 	},
 	{
+		name:         "VCVTPD2PSMasked256Merging",
+		argLen:       3,
+		resultInArg0: true,
+		asm:          x86.AVCVTPD2PS,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{2, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281472829161472},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+				{1, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:         "VCVTPD2PSXMasked128Merging",
+		argLen:       3,
+		resultInArg0: true,
+		asm:          x86.AVCVTPD2PSX,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{2, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281472829161472},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+				{1, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:         "VCVTPD2PSYMasked128Merging",
+		argLen:       3,
+		resultInArg0: true,
+		asm:          x86.AVCVTPD2PSY,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{2, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281472829161472},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+				{1, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:         "VCVTPS2PDMasked256Merging",
+		argLen:       3,
+		resultInArg0: true,
+		asm:          x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{2, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281472829161472},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+				{1, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
+		name:         "VCVTPS2PDMasked512Merging",
+		argLen:       3,
+		resultInArg0: true,
+		asm:          x86.AVCVTPS2PD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{2, 71494644084506624}, // K1 K2 K3 K4 K5 K6 K7
+				{0, 281472829161472},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+				{1, 281474976645120},   // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X15 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+			outputs: []outputInfo{
+				{0, 281472829161472}, // X0 X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X16 X17 X18 X19 X20 X21 X22 X23 X24 X25 X26 X27 X28 X29 X30 X31
+			},
+		},
+	},
+	{
 		name:         "VCVTQQ2PDMasked128Merging",
 		argLen:       3,
 		resultInArg0: true,
@@ -89845,6 +90197,21 @@ var opcodeTable = [...]opInfo{
 		generic: true,
 	},
 	{
+		name:    "ConvertToFloat32Float64x2",
+		argLen:  1,
+		generic: true,
+	},
+	{
+		name:    "ConvertToFloat32Float64x4",
+		argLen:  1,
+		generic: true,
+	},
+	{
+		name:    "ConvertToFloat32Float64x8",
+		argLen:  1,
+		generic: true,
+	},
+	{
 		name:    "ConvertToFloat32Int32x4",
 		argLen:  1,
 		generic: true,
@@ -89905,6 +90272,16 @@ var opcodeTable = [...]opInfo{
 		generic: true,
 	},
 	{
+		name:    "ConvertToFloat64Float32x4",
+		argLen:  1,
+		generic: true,
+	},
+	{
+		name:    "ConvertToFloat64Float32x8",
+		argLen:  1,
+		generic: true,
+	},
+	{
 		name:    "ConvertToFloat64Int32x4",
 		argLen:  1,
 		generic: true,
diff --git a/src/cmd/compile/internal/ssa/rewriteAMD64.go b/src/cmd/compile/internal/ssa/rewriteAMD64.go
index e489b26430..a6fc4bfbaf 100644
--- a/src/cmd/compile/internal/ssa/rewriteAMD64.go
+++ b/src/cmd/compile/internal/ssa/rewriteAMD64.go
@@ -604,6 +604,20 @@ func rewriteValueAMD64(v *Value) bool {
 		return rewriteValueAMD64_OpAMD64VCVTDQ2PSMasked256(v)
 	case OpAMD64VCVTDQ2PSMasked512:
 		return rewriteValueAMD64_OpAMD64VCVTDQ2PSMasked512(v)
+	case OpAMD64VCVTPD2PS256:
+		return rewriteValueAMD64_OpAMD64VCVTPD2PS256(v)
+	case OpAMD64VCVTPD2PSMasked256:
+		return rewriteValueAMD64_OpAMD64VCVTPD2PSMasked256(v)
+	case OpAMD64VCVTPD2PSXMasked128:
+		return rewriteValueAMD64_OpAMD64VCVTPD2PSXMasked128(v)
+	case OpAMD64VCVTPD2PSYMasked128:
+		return rewriteValueAMD64_OpAMD64VCVTPD2PSYMasked128(v)
+	case OpAMD64VCVTPS2PD512:
+		return rewriteValueAMD64_OpAMD64VCVTPS2PD512(v)
+	case OpAMD64VCVTPS2PDMasked256:
+		return rewriteValueAMD64_OpAMD64VCVTPS2PDMasked256(v)
+	case OpAMD64VCVTPS2PDMasked512:
+		return rewriteValueAMD64_OpAMD64VCVTPS2PDMasked512(v)
 	case OpAMD64VCVTQQ2PD128:
 		return rewriteValueAMD64_OpAMD64VCVTQQ2PD128(v)
 	case OpAMD64VCVTQQ2PD256:
@@ -2809,6 +2823,15 @@ func rewriteValueAMD64(v *Value) bool {
 		return rewriteValueAMD64_OpConstBool(v)
 	case OpConstNil:
 		return rewriteValueAMD64_OpConstNil(v)
+	case OpConvertToFloat32Float64x2:
+		v.Op = OpAMD64VCVTPD2PSX128
+		return true
+	case OpConvertToFloat32Float64x4:
+		v.Op = OpAMD64VCVTPD2PSY128
+		return true
+	case OpConvertToFloat32Float64x8:
+		v.Op = OpAMD64VCVTPD2PS256
+		return true
 	case OpConvertToFloat32Int32x16:
 		v.Op = OpAMD64VCVTDQ2PS512
 		return true
@@ -2845,6 +2868,12 @@ func rewriteValueAMD64(v *Value) bool {
 	case OpConvertToFloat32Uint64x8:
 		v.Op = OpAMD64VCVTUQQ2PS256
 		return true
+	case OpConvertToFloat64Float32x4:
+		v.Op = OpAMD64VCVTPS2PD256
+		return true
+	case OpConvertToFloat64Float32x8:
+		v.Op = OpAMD64VCVTPS2PD512
+		return true
 	case OpConvertToFloat64Int32x4:
 		v.Op = OpAMD64VCVTDQ2PD256
 		return true
@@ -29165,6 +29194,191 @@ func rewriteValueAMD64_OpAMD64VCVTDQ2PSMasked512(v *Value) bool {
 	}
 	return false
 }
+func rewriteValueAMD64_OpAMD64VCVTPD2PS256(v *Value) bool {
+	v_0 := v.Args[0]
+	// match: (VCVTPD2PS256 l:(VMOVDQUload512 {sym} [off] ptr mem))
+	// cond: canMergeLoad(v, l) && clobber(l)
+	// result: (VCVTPD2PS256load {sym} [off] ptr mem)
+	for {
+		l := v_0
+		if l.Op != OpAMD64VMOVDQUload512 {
+			break
+		}
+		off := auxIntToInt32(l.AuxInt)
+		sym := auxToSym(l.Aux)
+		mem := l.Args[1]
+		ptr := l.Args[0]
+		if !(canMergeLoad(v, l) && clobber(l)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPD2PS256load)
+		v.AuxInt = int32ToAuxInt(off)
+		v.Aux = symToAux(sym)
+		v.AddArg2(ptr, mem)
+		return true
+	}
+	return false
+}
+func rewriteValueAMD64_OpAMD64VCVTPD2PSMasked256(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	// match: (VCVTPD2PSMasked256 l:(VMOVDQUload512 {sym} [off] ptr mem) mask)
+	// cond: canMergeLoad(v, l) && clobber(l)
+	// result: (VCVTPD2PSMasked256load {sym} [off] ptr mask mem)
+	for {
+		l := v_0
+		if l.Op != OpAMD64VMOVDQUload512 {
+			break
+		}
+		off := auxIntToInt32(l.AuxInt)
+		sym := auxToSym(l.Aux)
+		mem := l.Args[1]
+		ptr := l.Args[0]
+		mask := v_1
+		if !(canMergeLoad(v, l) && clobber(l)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPD2PSMasked256load)
+		v.AuxInt = int32ToAuxInt(off)
+		v.Aux = symToAux(sym)
+		v.AddArg3(ptr, mask, mem)
+		return true
+	}
+	return false
+}
+func rewriteValueAMD64_OpAMD64VCVTPD2PSXMasked128(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	// match: (VCVTPD2PSXMasked128 l:(VMOVDQUload128 {sym} [off] ptr mem) mask)
+	// cond: canMergeLoad(v, l) && clobber(l)
+	// result: (VCVTPD2PSXMasked128load {sym} [off] ptr mask mem)
+	for {
+		l := v_0
+		if l.Op != OpAMD64VMOVDQUload128 {
+			break
+		}
+		off := auxIntToInt32(l.AuxInt)
+		sym := auxToSym(l.Aux)
+		mem := l.Args[1]
+		ptr := l.Args[0]
+		mask := v_1
+		if !(canMergeLoad(v, l) && clobber(l)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPD2PSXMasked128load)
+		v.AuxInt = int32ToAuxInt(off)
+		v.Aux = symToAux(sym)
+		v.AddArg3(ptr, mask, mem)
+		return true
+	}
+	return false
+}
+func rewriteValueAMD64_OpAMD64VCVTPD2PSYMasked128(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	// match: (VCVTPD2PSYMasked128 l:(VMOVDQUload256 {sym} [off] ptr mem) mask)
+	// cond: canMergeLoad(v, l) && clobber(l)
+	// result: (VCVTPD2PSYMasked128load {sym} [off] ptr mask mem)
+	for {
+		l := v_0
+		if l.Op != OpAMD64VMOVDQUload256 {
+			break
+		}
+		off := auxIntToInt32(l.AuxInt)
+		sym := auxToSym(l.Aux)
+		mem := l.Args[1]
+		ptr := l.Args[0]
+		mask := v_1
+		if !(canMergeLoad(v, l) && clobber(l)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPD2PSYMasked128load)
+		v.AuxInt = int32ToAuxInt(off)
+		v.Aux = symToAux(sym)
+		v.AddArg3(ptr, mask, mem)
+		return true
+	}
+	return false
+}
+func rewriteValueAMD64_OpAMD64VCVTPS2PD512(v *Value) bool {
+	v_0 := v.Args[0]
+	// match: (VCVTPS2PD512 l:(VMOVDQUload256 {sym} [off] ptr mem))
+	// cond: canMergeLoad(v, l) && clobber(l)
+	// result: (VCVTPS2PD512load {sym} [off] ptr mem)
+	for {
+		l := v_0
+		if l.Op != OpAMD64VMOVDQUload256 {
+			break
+		}
+		off := auxIntToInt32(l.AuxInt)
+		sym := auxToSym(l.Aux)
+		mem := l.Args[1]
+		ptr := l.Args[0]
+		if !(canMergeLoad(v, l) && clobber(l)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPS2PD512load)
+		v.AuxInt = int32ToAuxInt(off)
+		v.Aux = symToAux(sym)
+		v.AddArg2(ptr, mem)
+		return true
+	}
+	return false
+}
+func rewriteValueAMD64_OpAMD64VCVTPS2PDMasked256(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	// match: (VCVTPS2PDMasked256 l:(VMOVDQUload128 {sym} [off] ptr mem) mask)
+	// cond: canMergeLoad(v, l) && clobber(l)
+	// result: (VCVTPS2PDMasked256load {sym} [off] ptr mask mem)
+	for {
+		l := v_0
+		if l.Op != OpAMD64VMOVDQUload128 {
+			break
+		}
+		off := auxIntToInt32(l.AuxInt)
+		sym := auxToSym(l.Aux)
+		mem := l.Args[1]
+		ptr := l.Args[0]
+		mask := v_1
+		if !(canMergeLoad(v, l) && clobber(l)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPS2PDMasked256load)
+		v.AuxInt = int32ToAuxInt(off)
+		v.Aux = symToAux(sym)
+		v.AddArg3(ptr, mask, mem)
+		return true
+	}
+	return false
+}
+func rewriteValueAMD64_OpAMD64VCVTPS2PDMasked512(v *Value) bool {
+	v_1 := v.Args[1]
+	v_0 := v.Args[0]
+	// match: (VCVTPS2PDMasked512 l:(VMOVDQUload256 {sym} [off] ptr mem) mask)
+	// cond: canMergeLoad(v, l) && clobber(l)
+	// result: (VCVTPS2PDMasked512load {sym} [off] ptr mask mem)
+	for {
+		l := v_0
+		if l.Op != OpAMD64VMOVDQUload256 {
+			break
+		}
+		off := auxIntToInt32(l.AuxInt)
+		sym := auxToSym(l.Aux)
+		mem := l.Args[1]
+		ptr := l.Args[0]
+		mask := v_1
+		if !(canMergeLoad(v, l) && clobber(l)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPS2PDMasked512load)
+		v.AuxInt = int32ToAuxInt(off)
+		v.Aux = symToAux(sym)
+		v.AddArg3(ptr, mask, mem)
+		return true
+	}
+	return false
+}
 func rewriteValueAMD64_OpAMD64VCVTQQ2PD128(v *Value) bool {
 	v_0 := v.Args[0]
 	// match: (VCVTQQ2PD128 l:(VMOVDQUload128 {sym} [off] ptr mem))
@@ -35833,6 +36047,18 @@ func rewriteValueAMD64_OpAMD64VMOVDQU32Masked256(v *Value) bool {
 		v.AddArg2(x, mask)
 		return true
 	}
+	// match: (VMOVDQU32Masked256 (VCVTPS2PD256 x) mask)
+	// result: (VCVTPS2PDMasked256 x mask)
+	for {
+		if v_0.Op != OpAMD64VCVTPS2PD256 {
+			break
+		}
+		x := v_0.Args[0]
+		mask := v_1
+		v.reset(OpAMD64VCVTPS2PDMasked256)
+		v.AddArg2(x, mask)
+		return true
+	}
 	// match: (VMOVDQU32Masked256 (VCVTDQ2PD256 x) mask)
 	// result: (VCVTDQ2PDMasked256 x mask)
 	for {
@@ -36744,6 +36970,18 @@ func rewriteValueAMD64_OpAMD64VMOVDQU32Masked512(v *Value) bool {
 		v.AddArg2(x, mask)
 		return true
 	}
+	// match: (VMOVDQU32Masked512 (VCVTPS2PD512 x) mask)
+	// result: (VCVTPS2PDMasked512 x mask)
+	for {
+		if v_0.Op != OpAMD64VCVTPS2PD512 {
+			break
+		}
+		x := v_0.Args[0]
+		mask := v_1
+		v.reset(OpAMD64VCVTPS2PDMasked512)
+		v.AddArg2(x, mask)
+		return true
+	}
 	// match: (VMOVDQU32Masked512 (VCVTDQ2PD512 x) mask)
 	// result: (VCVTDQ2PDMasked512 x mask)
 	for {
@@ -37571,6 +37809,30 @@ func rewriteValueAMD64_OpAMD64VMOVDQU64Masked128(v *Value) bool {
 		v.AddArg4(x, y, z, mask)
 		return true
 	}
+	// match: (VMOVDQU64Masked128 (VCVTPD2PSX128 x) mask)
+	// result: (VCVTPD2PSXMasked128 x mask)
+	for {
+		if v_0.Op != OpAMD64VCVTPD2PSX128 {
+			break
+		}
+		x := v_0.Args[0]
+		mask := v_1
+		v.reset(OpAMD64VCVTPD2PSXMasked128)
+		v.AddArg2(x, mask)
+		return true
+	}
+	// match: (VMOVDQU64Masked128 (VCVTPD2PSY128 x) mask)
+	// result: (VCVTPD2PSYMasked128 x mask)
+	for {
+		if v_0.Op != OpAMD64VCVTPD2PSY128 {
+			break
+		}
+		x := v_0.Args[0]
+		mask := v_1
+		v.reset(OpAMD64VCVTPD2PSYMasked128)
+		v.AddArg2(x, mask)
+		return true
+	}
 	// match: (VMOVDQU64Masked128 (VCVTQQ2PSX128 x) mask)
 	// result: (VCVTQQ2PSXMasked128 x mask)
 	for {
@@ -38410,6 +38672,18 @@ func rewriteValueAMD64_OpAMD64VMOVDQU64Masked256(v *Value) bool {
 		v.AddArg4(x, y, z, mask)
 		return true
 	}
+	// match: (VMOVDQU64Masked256 (VCVTPD2PS256 x) mask)
+	// result: (VCVTPD2PSMasked256 x mask)
+	for {
+		if v_0.Op != OpAMD64VCVTPD2PS256 {
+			break
+		}
+		x := v_0.Args[0]
+		mask := v_1
+		v.reset(OpAMD64VCVTPD2PSMasked256)
+		v.AddArg2(x, mask)
+		return true
+	}
 	// match: (VMOVDQU64Masked256 (VCVTQQ2PS256 x) mask)
 	// result: (VCVTQQ2PSMasked256 x mask)
 	for {
@@ -44249,6 +44523,19 @@ func rewriteValueAMD64_OpAMD64VPBLENDMQMasked512(v *Value) bool {
 		v.AddArg4(dst, x, y, mask)
 		return true
 	}
+	// match: (VPBLENDMQMasked512 dst (VCVTPD2PS256 x) mask)
+	// result: (VCVTPD2PSMasked256Merging dst x mask)
+	for {
+		dst := v_0
+		if v_1.Op != OpAMD64VCVTPD2PS256 {
+			break
+		}
+		x := v_1.Args[0]
+		mask := v_2
+		v.reset(OpAMD64VCVTPD2PSMasked256Merging)
+		v.AddArg3(dst, x, mask)
+		return true
+	}
 	// match: (VPBLENDMQMasked512 dst (VCVTQQ2PD512 x) mask)
 	// result: (VCVTQQ2PDMasked512Merging dst x mask)
 	for {
@@ -45583,6 +45870,44 @@ func rewriteValueAMD64_OpAMD64VPBLENDVB128(v *Value) bool {
 		v.AddArg3(dst, x, v0)
 		return true
 	}
+	// match: (VPBLENDVB128 dst (VCVTPD2PSX128 x) mask)
+	// cond: v.Block.CPUfeatures.hasFeature(CPUavx512)
+	// result: (VCVTPD2PSXMasked128Merging dst x (VPMOVVec64x2ToM <types.TypeMask> mask))
+	for {
+		dst := v_0
+		if v_1.Op != OpAMD64VCVTPD2PSX128 {
+			break
+		}
+		x := v_1.Args[0]
+		mask := v_2
+		if !(v.Block.CPUfeatures.hasFeature(CPUavx512)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPD2PSXMasked128Merging)
+		v0 := b.NewValue0(v.Pos, OpAMD64VPMOVVec64x2ToM, types.TypeMask)
+		v0.AddArg(mask)
+		v.AddArg3(dst, x, v0)
+		return true
+	}
+	// match: (VPBLENDVB128 dst (VCVTPS2PD256 x) mask)
+	// cond: v.Block.CPUfeatures.hasFeature(CPUavx512)
+	// result: (VCVTPS2PDMasked256Merging dst x (VPMOVVec32x4ToM <types.TypeMask> mask))
+	for {
+		dst := v_0
+		if v_1.Op != OpAMD64VCVTPS2PD256 {
+			break
+		}
+		x := v_1.Args[0]
+		mask := v_2
+		if !(v.Block.CPUfeatures.hasFeature(CPUavx512)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPS2PDMasked256Merging)
+		v0 := b.NewValue0(v.Pos, OpAMD64VPMOVVec32x4ToM, types.TypeMask)
+		v0.AddArg(mask)
+		v.AddArg3(dst, x, v0)
+		return true
+	}
 	// match: (VPBLENDVB128 dst (VCVTQQ2PD128 x) mask)
 	// cond: v.Block.CPUfeatures.hasFeature(CPUavx512)
 	// result: (VCVTQQ2PDMasked128Merging dst x (VPMOVVec64x2ToM <types.TypeMask> mask))
@@ -49194,6 +49519,44 @@ func rewriteValueAMD64_OpAMD64VPBLENDVB256(v *Value) bool {
 		v.AddArg3(dst, x, v0)
 		return true
 	}
+	// match: (VPBLENDVB256 dst (VCVTPD2PSY128 x) mask)
+	// cond: v.Block.CPUfeatures.hasFeature(CPUavx512)
+	// result: (VCVTPD2PSYMasked128Merging dst x (VPMOVVec64x4ToM <types.TypeMask> mask))
+	for {
+		dst := v_0
+		if v_1.Op != OpAMD64VCVTPD2PSY128 {
+			break
+		}
+		x := v_1.Args[0]
+		mask := v_2
+		if !(v.Block.CPUfeatures.hasFeature(CPUavx512)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPD2PSYMasked128Merging)
+		v0 := b.NewValue0(v.Pos, OpAMD64VPMOVVec64x4ToM, types.TypeMask)
+		v0.AddArg(mask)
+		v.AddArg3(dst, x, v0)
+		return true
+	}
+	// match: (VPBLENDVB256 dst (VCVTPS2PD512 x) mask)
+	// cond: v.Block.CPUfeatures.hasFeature(CPUavx512)
+	// result: (VCVTPS2PDMasked512Merging dst x (VPMOVVec32x8ToM <types.TypeMask> mask))
+	for {
+		dst := v_0
+		if v_1.Op != OpAMD64VCVTPS2PD512 {
+			break
+		}
+		x := v_1.Args[0]
+		mask := v_2
+		if !(v.Block.CPUfeatures.hasFeature(CPUavx512)) {
+			break
+		}
+		v.reset(OpAMD64VCVTPS2PDMasked512Merging)
+		v0 := b.NewValue0(v.Pos, OpAMD64VPMOVVec32x8ToM, types.TypeMask)
+		v0.AddArg(mask)
+		v.AddArg3(dst, x, v0)
+		return true
+	}
 	// match: (VPBLENDVB256 dst (VCVTQQ2PD256 x) mask)
 	// cond: v.Block.CPUfeatures.hasFeature(CPUavx512)
 	// result: (VCVTQQ2PDMasked256Merging dst x (VPMOVVec64x4ToM <types.TypeMask> mask))
diff --git a/src/cmd/compile/internal/ssagen/simdintrinsics.go b/src/cmd/compile/internal/ssagen/simdintrinsics.go
index 4eae25a18f..607311d3b2 100644
--- a/src/cmd/compile/internal/ssagen/simdintrinsics.go
+++ b/src/cmd/compile/internal/ssagen/simdintrinsics.go
@@ -261,6 +261,9 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Uint8x16.ConcatShiftBytesRight", opLen2Imm8(ssa.OpConcatShiftBytesRightUint8x16, types.TypeVec128, 0), sys.AMD64)
 	addF(simdPackage, "Uint8x32.ConcatShiftBytesRightGrouped", opLen2Imm8(ssa.OpConcatShiftBytesRightGroupedUint8x32, types.TypeVec256, 0), sys.AMD64)
 	addF(simdPackage, "Uint8x64.ConcatShiftBytesRightGrouped", opLen2Imm8(ssa.OpConcatShiftBytesRightGroupedUint8x64, types.TypeVec512, 0), sys.AMD64)
+	addF(simdPackage, "Float64x2.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Float64x2, types.TypeVec128), sys.AMD64)
+	addF(simdPackage, "Float64x4.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Float64x4, types.TypeVec128), sys.AMD64)
+	addF(simdPackage, "Float64x8.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Float64x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Int32x4.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Int32x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Int32x8.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Int32x8, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Int32x16.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Int32x16, types.TypeVec512), sys.AMD64)
@@ -273,6 +276,8 @@ func simdIntrinsics(addF func(pkg, fn string, b intrinsicBuilder, archFamilies .
 	addF(simdPackage, "Uint64x2.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Uint64x2, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Uint64x4.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Uint64x4, types.TypeVec128), sys.AMD64)
 	addF(simdPackage, "Uint64x8.ConvertToFloat32", opLen1(ssa.OpConvertToFloat32Uint64x8, types.TypeVec256), sys.AMD64)
+	addF(simdPackage, "Float32x4.ConvertToFloat64", opLen1(ssa.OpConvertToFloat64Float32x4, types.TypeVec256), sys.AMD64)
+	addF(simdPackage, "Float32x8.ConvertToFloat64", opLen1(ssa.OpConvertToFloat64Float32x8, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Int32x4.ConvertToFloat64", opLen1(ssa.OpConvertToFloat64Int32x4, types.TypeVec256), sys.AMD64)
 	addF(simdPackage, "Int32x8.ConvertToFloat64", opLen1(ssa.OpConvertToFloat64Int32x8, types.TypeVec512), sys.AMD64)
 	addF(simdPackage, "Int64x2.ConvertToFloat64", opLen1(ssa.OpConvertToFloat64Int64x2, types.TypeVec128), sys.AMD64)