[dev.simd] simd, cmd/compile: add widening unsigned converts 8->16->32

Change-Id: If0bde7154bd622573375eba5539fd642b8ef9d2f
Reviewed-on: https://go-review.googlesource.com/c/go/+/696555
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

3 files changed, 128 insertions, 5 deletions

diff --git a/src/simd/_gen/simdgen/ops/Converts/categories.yaml b/src/simd/_gen/simdgen/ops/Converts/categories.yaml
index cc6c419dcc..b4c7d468e9 100644
--- a/src/simd/_gen/simdgen/ops/Converts/categories.yaml
+++ b/src/simd/_gen/simdgen/ops/Converts/categories.yaml
@@ -2,9 +2,24 @@
 - go: ConvertToInt32
   commutative: false
   documentation: !string |-
-    // ConvertToInt32 converts element values to int32.
+    // NAME converts element values to int32.
 
 - go: ConvertToUint32
   commutative: false
   documentation: !string |-
-    // ConvertToUint32Masked converts element values to uint32.
+    // NAME converts element values to uint32.
+
+- go: ConvertToUint16
+  commutative: false
+  documentation: !string |-
+    // NAME converts element values to uint16.
+
+- go: ConvertToUint16x8
+  commutative: false
+  documentation: !string |-
+    // NAME converts 8 lowest vector element values to uint16.
+
+- go: ConvertToUint32x4
+  commutative: false
+  documentation: !string |-
+    // NAME converts 4 lowest vector element values to uint32.
diff --git a/src/simd/_gen/simdgen/ops/Converts/go.yaml b/src/simd/_gen/simdgen/ops/Converts/go.yaml
index 4e251728bf..be0f157b40 100644
--- a/src/simd/_gen/simdgen/ops/Converts/go.yaml
+++ b/src/simd/_gen/simdgen/ops/Converts/go.yaml
@@ -19,3 +19,75 @@
     go: $u
     base: uint
     elemBits: 32
+
+- go: ConvertToUint16x8
+  asm: "VPMOVZXBW"
+  in:
+  - &u8x16
+    base: uint
+    elemBits: 8
+    bits: 128
+  out:
+  -
+    base: uint
+    elemBits: 16
+    bits: 128
+
+- go: ConvertToUint16
+  asm: "VPMOVZXBW"
+  in:
+  - *u8x16
+  out:
+  -
+    base: uint
+    elemBits: 16
+    bits: 256
+
+- go: ConvertToUint16
+  asm: "VPMOVZXBW"
+  in:
+  -
+    base: uint
+    elemBits: 8
+    bits: 256
+  out:
+  -
+    base: uint
+    elemBits: 16
+    bits: 512
+
+- go: ConvertToUint32x4
+  asm: "VPMOVZXWD"
+  in:
+  - &u16x8
+    base: uint
+    elemBits: 16
+    bits: 128
+  out:
+  -
+    base: uint
+    elemBits: 32
+    bits: 128
+
+- go: ConvertToUint32
+  asm: "VPMOVZXWD"
+  in:
+  - *u16x8
+  out:
+  -
+    base: uint
+    elemBits: 32
+    bits: 256
+
+- go: ConvertToUint32
+  asm: "VPMOVZXWD"
+  in:
+  -
+    base: uint
+    elemBits: 16
+    bits: 256
+  out:
+  -
+    base: uint
+    elemBits: 32
+    bits: 512
diff --git a/src/simd/ops_amd64.go b/src/simd/ops_amd64.go
index 76bbf738cb..79f5dc8523 100644
--- a/src/simd/ops_amd64.go
+++ b/src/simd/ops_amd64.go
@@ -1212,23 +1212,59 @@ func (x Float32x8) ConvertToInt32() Int32x8
 // Asm: VCVTTPS2DQ, CPU Feature: AVX512
 func (x Float32x16) ConvertToInt32() Int32x16
 
+/* ConvertToUint16 */
+
+// ConvertToUint16 converts element values to uint16.
+//
+// Asm: VPMOVZXBW, CPU Feature: AVX2
+func (x Uint8x16) ConvertToUint16() Uint16x16
+
+// ConvertToUint16 converts element values to uint16.
+//
+// Asm: VPMOVZXBW, CPU Feature: AVX512
+func (x Uint8x32) ConvertToUint16() Uint16x32
+
+/* ConvertToUint16x8 */
+
+// ConvertToUint16x8 converts 8 lowest vector element values to uint16.
+//
+// Asm: VPMOVZXBW, CPU Feature: AVX
+func (x Uint8x16) ConvertToUint16x8() Uint16x8
+
 /* ConvertToUint32 */
 
-// ConvertToUint32Masked converts element values to uint32.
+// ConvertToUint32 converts element values to uint32.
 //
 // Asm: VCVTPS2UDQ, CPU Feature: AVX512
 func (x Float32x4) ConvertToUint32() Uint32x4
 
-// ConvertToUint32Masked converts element values to uint32.
+// ConvertToUint32 converts element values to uint32.
 //
 // Asm: VCVTPS2UDQ, CPU Feature: AVX512
 func (x Float32x8) ConvertToUint32() Uint32x8
 
-// ConvertToUint32Masked converts element values to uint32.
+// ConvertToUint32 converts element values to uint32.
 //
 // Asm: VCVTPS2UDQ, CPU Feature: AVX512
 func (x Float32x16) ConvertToUint32() Uint32x16
 
+// ConvertToUint32 converts element values to uint32.
+//
+// Asm: VPMOVZXWD, CPU Feature: AVX2
+func (x Uint16x8) ConvertToUint32() Uint32x8
+
+// ConvertToUint32 converts element values to uint32.
+//
+// Asm: VPMOVZXWD, CPU Feature: AVX512
+func (x Uint16x16) ConvertToUint32() Uint32x16
+
+/* ConvertToUint32x4 */
+
+// ConvertToUint32x4 converts 4 lowest vector element values to uint32.
+//
+// Asm: VPMOVZXWD, CPU Feature: AVX
+func (x Uint16x8) ConvertToUint32x4() Uint32x4
+
 /* CopySign */
 
 // CopySign returns the product of the first operand with -1, 0, or 1,