@ttsc/win32-arm64 0.33.0 → 0.34.0

package/bin/go/src/simd/archsimd/_gen/simdgen/main.go

@@ -4,11 +4,17 @@
 
 // simdgen is an experiment in generating Go <-> asm SIMD mappings.
 //
-// Usage: simdgen [-xedPath=path] [-q=query] input.yaml...
+// Usage: simdgen [-xedPath=path | -arm64Path=path] [-q=query] input.yaml...
+//
+// Only one of -xedPath or -arm64Path may be specified.
 //
 // If -xedPath is provided, one of the inputs is a sum of op-code definitions
 // generated from the Intel XED data at path.
 //
+// If -arm64Path is provided, one of the inputs is a set of instruction
+// definitions parsed from ARM64 ISA XML files at path (obtained from
+// https://developer.arm.com/-/cdn-downloads/permalink/Exploration-Tools-A64-ISA/ISA_A64/ISA_A64_xml_A_profile-2025-12.tar.gz).
+//
 // If input YAML files are provided, each file is read as an input value. See
 // [unify.Closure.UnmarshalYAML] or "go doc unify.Closure.UnmarshalYAML" for the
 // format of these files.
@@ -37,14 +43,22 @@
 //
 //	go run . -xedPath $XEDPATH -q '{asm: VPADDQ}'
 //
+// For VADD.S4 on ARM64:
+//
+//	go run . -arm64Path $ARM64_ISA_PATH -o yaml -q '{asm: VADD, arrangement: "4S"}'
+//
 // simdgen can also generate Go definitions of SIMD mappings:
 // To generate go files to the go root, run:
 //
-//	go run . -xedPath $XEDPATH -o godefs -goroot $PATH/TO/go go.yaml categories.yaml types.yaml
+//	go run . -xedPath $XEDPATH -o godefs -goroot $PATH/TO/go go_amd64.yaml categories.yaml types.yaml
+//
+// For ARM64:
+//
+//	go run . -arm64Path $ARM64_ISA_PATH -o godefs -goroot $PATH/TO/go go_arm64.yaml categories.yaml types.yaml
 //
 // types.yaml is already written, it specifies the shapes of vectors.
-// categories.yaml and go.yaml contains definitions that unifies with types.yaml and XED
-// data, you can find an example in ops/AddSub/.
+// categories.yaml and go_<arch>.yaml contain definitions that unify with types.yaml and
+// XED/ARM64 ISA data, you can find an example in ops/AddSub/.
 //
 // When generating Go definitions, simdgen do 3 "magic"s:
 // - It splits masked operations(with op's [Masked] field set) to const and non const:
@@ -67,7 +81,7 @@
 // These 3 magics could be disabled by enabling -nosplitmask, -nodedup or
 // -noconstimmporting flags.
 //
-// simdgen right now only supports amd64, -arch=$OTHERARCH will trigger a fatal error.
+// simdgen supports amd64 and arm64 architectures.
 package main
 
 // Big TODOs:
@@ -92,6 +106,7 @@ import (
 	"slices"
 	"strings"
 
+	"simd/archsimd/_gen/simdgen/arm64"
 	"simd/archsimd/_gen/unify"
 
 	"gopkg.in/yaml.v3"
@@ -99,12 +114,15 @@ import (
 
 var (
 	xedPath               = flag.String("xedPath", "", "load XED datafiles from `path`")
+	arm64Path             = flag.String("arm64Path", "", "load ARM64 instruction xml definitions from `path`")
 	flagQ                 = flag.String("q", "", "query: read `def` as another input (skips final validation)")
 	flagO                 = flag.String("o", "yaml", "output type: yaml, godefs (generate definitions into a Go source tree")
 	flagGoDefRoot         = flag.String("goroot", ".", "the path to the Go dev directory that will receive the generated files")
 	FlagNoDedup           = flag.Bool("nodedup", false, "disable deduplicating godefs of 2 qualifying operations from different extensions")
 	FlagNoConstImmPorting = flag.Bool("noconstimmporting", false, "disable const immediate porting from op to imm operand")
-	FlagArch              = flag.String("arch", "amd64", "the target architecture")
+
+	// FlagArch must be pre-initialized to a bogus value because there have been initializations that depended on it
+	FlagArch = flag.String("arch", "must be specified, amd64 or arm64", "the target architecture")
 
 	Verbose = flag.Bool("v", false, "verbose")
 
@@ -142,18 +160,36 @@ func main() {
 		}()
 	}
 
-	var inputs []unify.Closure
-
-	if *FlagArch != "amd64" {
-		log.Fatalf("simdgen only supports amd64")
+	// Default -arch to arm64 when -arm64Path is specified.
+	if *arm64Path != "" && *FlagArch != "arm64" {
+		if *xedPath != "" {
+			log.Fatalf("both -xedPath and -arm64Path specified")
+		}
+		// *FlagArch = "arm64"
 	}
 
-	// Load XED into a defs set.
-	if *xedPath != "" {
-		xedDefs := loadXED(*xedPath)
-		inputs = append(inputs, unify.NewSum(xedDefs...))
+	// Load instructions into the architecture-specific defs set.
+	var defs []*unify.Value
+	switch *FlagArch {
+	case "amd64":
+		if *xedPath != "" {
+			defs = loadXED(*xedPath)
+		}
+	case "arm64":
+		if *arm64Path != "" {
+			var err error
+			defs, err = arm64.Load(*arm64Path)
+			if err != nil {
+				log.Fatalf("loading ARM64 instructions: %s", err)
+			}
+		}
+	default:
+		log.Fatalf("simdgen only supports amd64 and arm64")
 	}
 
+	var inputs []unify.Closure
+	inputs = append(inputs, unify.NewSum(defs...))
+
 	// Load query.
 	if *flagQ != "" {
 		r := strings.NewReader(*flagQ)
@@ -173,7 +209,8 @@ func main() {
 		}
 		inputs = append(inputs, defs)
 
-		if filepath.Base(path) == "go.yaml" {
+		base := filepath.Base(path)
+		if base == "go_amd64.yaml" || base == "go_arm64.yaml" {
 			// These must all be used in the final result
 			for def := range defs.Summands() {
 				must[def] = struct{}{}
@@ -200,13 +237,15 @@ func main() {
 		log.Fatal(err)
 	}
 
+	ok := true
+
 	// Validate results.
 	//
 	// Don't validate if this is a command-line query because that tends to
 	// eliminate lots of required defs and is used in cases where maybe defs
 	// aren't enumerable anyway.
 	if *flagQ == "" && len(must) > 0 {
-		validate(unified, must)
+		ok = validate(unified, must)
 	}
 
 	// Print results.
@@ -238,9 +277,13 @@ func main() {
 			fmt.Fprintf(os.Stderr, "XED decoding generated %d \"errors\" which is not cause for alarm, use -v for details.\n", operandRemarks)
 		}
 	}
+	if !ok {
+		os.Exit(1)
+	}
 }
 
-func validate(cl unify.Closure, required map[*unify.Value]struct{}) {
+func validate(cl unify.Closure, required map[*unify.Value]struct{}) bool {
+	ok := true
 	// Validate that:
 	// 1. All final defs are exact
 	// 2. All required defs are used
@@ -277,5 +320,7 @@ func validate(cl unify.Closure, required map[*unify.Value]struct{}) {
 		// be worth it.
 		fmt.Fprintf(os.Stderr, "%s: def required, but did not unify (%v)\n",
 			def.PosString(), def)
+		ok = false
 	}
+	return ok
 }

package/bin/go/src/simd/archsimd/_gen/simdgen/ops/AddSub/categories.yaml

@@ -15,45 +15,45 @@
   commutative: false
   documentation: !string |-
     // NAME subtracts corresponding elements of two vectors with saturation.
-- go: AddPairs
+- go: ConcatAddPairs
   commutative: false
   out:
   - elemBits: 16|32
   documentation: !string |-
     // NAME horizontally adds adjacent pairs of elements.
     // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [x0+x1, x2+x3, ..., y0+y1, y2+y3, ...].
-- go: AddPairs
+- go: ConcatAddPairs
   commutative: false
   out:
   - elemBits: 64
   documentation: !string |-
     // NAME horizontally adds adjacent pairs of elements.
     // For x = [x0, x1] and y = [y0, y1], the result is [x0+x1, y0+y1].
-- go: SubPairs
+- go: ConcatSubPairs
   commutative: false
   out:
   - elemBits: 16|32
   documentation: !string |-
     // NAME horizontally subtracts adjacent pairs of elements.
     // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [x0-x1, x2-x3, ..., y0-y1, y2-y3, ...].
-- go: SubPairs
+- go: ConcatSubPairs
   commutative: false
   out:
   - elemBits: 64
   documentation: !string |-
     // NAME horizontally subtracts adjacent pairs of elements.
     // For x = [x0, x1] and y = [y0, y1], the result is [x0-x1, y0-y1].
-- go: AddPairsSaturated
+- go: ConcatAddPairsSaturated
   commutative: false
   documentation: !string |-
     // NAME horizontally adds adjacent pairs of elements with saturation.
     // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [x0+x1, x2+x3, ..., y0+y1, y2+y3, ...].
-- go: SubPairsSaturated
+- go: ConcatSubPairsSaturated
   commutative: false
   documentation: !string |-
     // NAME horizontally subtracts adjacent pairs of elements with saturation.
     // For x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [x0-x1, x2-x3, ..., y0-y1, y2-y3, ...].
-- go: AddPairsGrouped
+- go: ConcatAddPairsGrouped
   commutative: false
   out:
   - elemBits: 16|32
@@ -61,7 +61,7 @@
     // NAME horizontally adds adjacent pairs of elements.
     // With each 128-bit as a group:
     // for x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [x0+x1, x2+x3, ..., y0+y1, y2+y3, ...].
-- go: AddPairsGrouped
+- go: ConcatAddPairsGrouped
   commutative: false
   out:
   - elemBits: 64
@@ -69,7 +69,7 @@
     // NAME horizontally adds adjacent pairs of elements.
     // With each 128-bit as a group:
     // for x = [x0, x1] and y = [y0, y1], the result is [x0+x1, y0+y1].
-- go: SubPairsGrouped
+- go: ConcatSubPairsGrouped
   commutative: false
   out:
   - elemBits: 16|32
@@ -77,7 +77,7 @@
     // NAME horizontally subtracts adjacent pairs of elements.
     // With each 128-bit as a group:
     // for x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [x0-x1, x2-x3, ..., y0-y1, y2-y3, ...].
-- go: SubPairsGrouped
+- go: ConcatSubPairsGrouped
   commutative: false
   out:
   - elemBits: 64
@@ -85,13 +85,13 @@
     // NAME horizontally subtracts adjacent pairs of elements.
     // With each 128-bit as a group:
     // for x = [x0, x1] and y = [y0, y1], the result is [x0-x1, y0-y1].
-- go: AddPairsSaturatedGrouped
+- go: ConcatAddPairsSaturatedGrouped
   commutative: false
   documentation: !string |-
     // NAME horizontally adds adjacent pairs of elements with saturation.
     // With each 128-bit as a group:
     // for x = [x0, x1, x2, x3, ...] and y = [y0, y1, y2, y3, ...], the result is [x0+x1, x2+x3, ..., y0+y1, y2+y3, ...].
-- go: SubPairsSaturatedGrouped
+- go: ConcatSubPairsSaturatedGrouped
   commutative: false
   documentation: !string |-
     // NAME horizontally subtracts adjacent pairs of elements with saturation.

package/bin/go/src/simd/archsimd/_gen/simdgen/ops/AddSub/{go.yaml → go_amd64.yaml}

@@ -51,7 +51,7 @@
   - *uint
   out:
   - *uint
-- go: AddPairs
+- go: ConcatAddPairs
   asm: "VPHADD[DW]"
   in: &2any128
   - &any128
@@ -60,19 +60,19 @@
   - *any128
   out: &1any128
   - *any128
-- go: SubPairs
+- go: ConcatSubPairs
   asm: "VPHSUB[DW]"
   in: *2any128
   out: *1any128
-- go: AddPairs
+- go: ConcatAddPairs
   asm: "VHADDP[SD]" # floats
   in: *2any128
   out: *1any128
-- go: SubPairs
+- go: ConcatSubPairs
   asm: "VHSUBP[SD]"  # floats
   in: *2any128
   out: *1any128
-- go: AddPairsSaturated
+- go: ConcatAddPairsSaturated
   asm: "VPHADDS[DW]"
   in: &2int128
   - &int128
@@ -82,11 +82,11 @@
   - *int128
   out: &1int128
   - *int128
-- go: SubPairsSaturated
+- go: ConcatSubPairsSaturated
   asm: "VPHSUBS[DW]"
   in: *2int128
   out: *1int128
-- go: AddPairsGrouped
+- go: ConcatAddPairsGrouped
   asm: "VPHADD[DW]"
   in: &2any256
   - &any256
@@ -95,19 +95,19 @@
   - *any256
   out: &1any256
   - *any256
-- go: SubPairsGrouped
+- go: ConcatSubPairsGrouped
   asm: "VPHSUB[DW]"
   in: *2any256
   out: *1any256
-- go: AddPairsGrouped
+- go: ConcatAddPairsGrouped
   asm: "VHADDP[SD]" # floats
   in: *2any256
   out: *1any256
-- go: SubPairsGrouped
+- go: ConcatSubPairsGrouped
   asm: "VHSUBP[SD]"  # floats
   in: *2any256
   out: *1any256
-- go: AddPairsSaturatedGrouped
+- go: ConcatAddPairsSaturatedGrouped
   asm: "VPHADDS[DW]"
   in: &2int256
   - &int256
@@ -117,7 +117,7 @@
   - *int256
   out: &1int256
   - *int256
-- go: SubPairsSaturatedGrouped
+- go: ConcatSubPairsSaturatedGrouped
   asm: "VPHSUBS[DW]"
   in: *2int256
   out: *1int256

package/bin/go/src/simd/archsimd/_gen/simdgen/ops/AddSub/go_arm64.yaml

@@ -0,0 +1,91 @@
+!sum
+- go: Add
+  asm: "VADD|VFADD"
+  in:
+  - &any
+    go: $t
+  - *any
+  out:
+  - *any
+- go: Sub
+  asm: "VSUB|VFSUB"
+  in:
+  - *any
+  - *any
+  out:
+  - *any
+
+# Signed saturating add — VSQADD
+- go: AddSaturated
+  signed: true
+  asm: "VSQADD"
+  in:
+  - &int
+    go: $t
+    base: int
+  - *int
+  out:
+  - *int
+
+# Unsigned saturating add — VUQADD
+- go: AddSaturated
+  signed: false
+  asm: "VUQADD"
+  in:
+  - &uint
+    go: $t
+    base: uint
+  - *uint
+  out:
+  - *uint
+
+# Signed saturating subtract — VSQSUB
+- go: SubSaturated
+  signed: true
+  asm: "VSQSUB"
+  in:
+  - *int
+  - *int
+  out:
+  - *int
+
+# Unsigned saturating subtract — VUQSUB
+- go: SubSaturated
+  signed: false
+  asm: "VUQSUB"
+  in:
+  - *uint
+  - *uint
+  out:
+  - *uint
+
+# Horizontal add of pairs (concatenated) — VADDP (int)
+- go: ConcatAddPairs
+  signed: true
+  asm: "VADDP"
+  in:
+  - *int
+  - *int
+  out:
+  - *int
+
+# Horizontal add of pairs (concatenated) — VADDP (uint)
+- go: ConcatAddPairs
+  signed: false
+  asm: "VADDP"
+  in:
+  - *uint
+  - *uint
+  out:
+  - *uint
+
+# Horizontal add of pairs (concatenated) — VFADDP (float)
+- go: ConcatAddPairs
+  asm: "VFADDP"
+  in:
+  - &fp
+    go: $t
+    base: float
+  - *fp
+  out:
+  - *fp

package/bin/go/src/simd/archsimd/_gen/simdgen/ops/BitwiseLogic/categories.yaml

@@ -2,11 +2,11 @@
 - go: And
   commutative: true
   documentation: !string |-
-    // NAME performs a bitwise AND operation between two vectors.
+    // NAME performs a bitwise x & y.
 - go: Or
   commutative: true
   documentation: !string |-
-    // NAME performs a bitwise OR operation between two vectors.
+    // NAME performs a bitwise x | y.
 - go: AndNot
   commutative: false
   documentation: !string |-
@@ -14,7 +14,15 @@
 - go: Xor
   commutative: true
   documentation: !string |-
-    // NAME performs a bitwise XOR operation between two vectors.
+    // NAME performs a bitwise x ^ y.
+- go: OrNot
+  commutative: false
+  documentation: !string |-
+    // NAME performs a bitwise x | ^y.
+- go: Not
+  commutative: false
+  documentation: !string |-
+    // NAME returns the bitwise complement of x.
 - go: tern
   commutative: false
   documentation: !string |-

package/bin/go/src/simd/archsimd/_gen/simdgen/ops/BitwiseLogic/go_arm64.yaml

@@ -0,0 +1,43 @@
+!sum
+- go: And
+  asm: "VAND"
+  in:
+  - &any
+    go: $t
+  - *any
+  out:
+  - *any
+- go: Or
+  asm: "VORR"
+  in:
+  - *any
+  - *any
+  out:
+  - *any
+- go: Xor
+  asm: "VEOR"
+  in:
+  - *any
+  - *any
+  out:
+  - *any
+- go: AndNot
+  asm: "VBIC"
+  in:
+  - *any
+  - *any
+  out:
+  - *any
+- go: OrNot
+  asm: "VORN"
+  in:
+  - *any
+  - *any
+  out:
+  - *any
+- go: Not
+  asm: "VNOT"
+  in:
+  - *any
+  out:
+  - *any

package/bin/go/src/simd/archsimd/_gen/simdgen/ops/Compares/go_arm64.yaml

@@ -0,0 +1,81 @@
+!sum
+- go: Equal
+  asm: "VCMEQ"
+  in:
+  - &int
+    go: $t
+    base: int
+  - *int
+  out:
+  - &anyToMask
+    go: $t
+    overwriteBase: int
+    overwriteClass: mask
+
+- go: Equal
+  asm: "VCMEQ"
+  in:
+  - &uint
+    go: $t
+    base: uint
+  - *uint
+  out:
+  - *anyToMask
+
+- go: Greater
+  asm: "VCMGT"
+  in:
+  - *int
+  - *int
+  out:
+  - *anyToMask
+
+- go: Greater
+  asm: "VCMHI"
+  in:
+  - *uint
+  - *uint
+  out:
+  - *anyToMask
+
+- go: GreaterEqual
+  asm: "VCMGE"
+  in:
+  - *int
+  - *int
+  out:
+  - *anyToMask
+
+- go: GreaterEqual
+  asm: "VCMHS"
+  in:
+  - *uint
+  - *uint
+  out:
+  - *anyToMask
+
+- go: Equal
+  asm: "VFCMEQ"
+  in:
+  - &float
+    go: $t
+    base: float
+  - *float
+  out:
+  - *anyToMask
+
+- go: Greater
+  asm: "VFCMGT"
+  in:
+  - *float
+  - *float
+  out:
+  - *anyToMask
+
+- go: GreaterEqual
+  asm: "VFCMGE"
+  in:
+  - *float
+  - *float
+  out:
+  - *anyToMask

package/bin/go/src/simd/archsimd/_gen/simdgen/ops/Converts/categories.yaml

@@ -42,9 +42,14 @@
   regexpTag: "convert"
   documentation: !string |-
     // NAME converts element values to float64.
+- go: "ConvertLo2ToFloat64"
+  commutative: false
+  regexpTag: "convert"
+  documentation: !string |-
+    // NAME converts low-indexed float32 element values to float64.
 
 # Int <-> Int conversions
-- go: "TruncateToInt8"
+- go: "TruncToInt8"
   commutative: false
   regexpTag: "convert"
   documentation: !string |-
@@ -59,7 +64,7 @@
   regexpTag: "convert"
   documentation: !string |-
     // NAME sign-extends element values to int16.
-- go: "TruncateToInt16(Concat)?"
+- go: "TruncToInt16(Concat)?"
   commutative: false
   regexpTag: "convert"
   documentation: !string |-
@@ -74,7 +79,7 @@
   regexpTag: "convert"
   documentation: !string |-
     // NAME sign-extends element values to int32.
-- go: "TruncateToInt32"
+- go: "TruncToInt32"
   commutative: false
   regexpTag: "convert"
   documentation: !string |-
@@ -89,7 +94,7 @@
   regexpTag: "convert"
   documentation: !string |-
     // NAME sign-extends element values to int64.
-- go: "TruncateToUint8"
+- go: "TruncToUint8"
   commutative: false
   regexpTag: "convert"
   documentation: !string |-
@@ -104,7 +109,7 @@
   regexpTag: "convert"
   documentation: !string |-
     // NAME zero-extends element values to uint16.
-- go: "TruncateToUint16(Concat)?"
+- go: "TruncToUint16(Concat)?"
   commutative: false
   regexpTag: "convert"
   documentation: !string |-
@@ -119,7 +124,7 @@
   commutative: false
   documentation: !string |-
     // NAME zero-extends element values to uint32.
-- go: "TruncateToUint32"
+- go: "TruncToUint32"
   regexpTag: "convert"
   commutative: false
   documentation: !string |-