native: 汇编器针对ARM调整注释语法

Author: chai2010

docs/nasm.md

@@ -93,7 +93,9 @@ main:
 
 ### 2.1. 注释
 
-`#` 开头的一行为注释
+`#` 和 `//` 开头的一行为注释.
+
+注意, ARM 平台不支持 `#` 风格的注释(被用作了立即数前缀).
 
 ### 2.2. 标识符
 

internal/native/arm64/lookup.go

@@ -0,0 +1,36 @@
+package arm64
+
+import "wa-lang.org/wa/internal/native/abi"
+
+// 寄存器名字列表
+var _Register = []string{}
+
+// 指令的名字
+// 保持和指令定义相同的顺序
+var _Anames = []string{}
+
+// 根据名字查找寄存器(忽略大小写, 忽略下划线和点的区别)
+func LookupRegister(regName string) (r abi.RegType, ok bool) {
+	if regName == "" {
+		return
+	}
+	for i, s := range _Register {
+		if strEqualFold(s, regName) {
+			return abi.RegType(i), true
+		}
+	}
+	return 0, false
+}
+
+// 根据名字查找汇编指令(忽略大小写, 忽略下划线和点的区别)
+func LookupAs(asName string) (as abi.As, ok bool) {
+	if asName == "" {
+		return
+	}
+	for i, s := range _Anames {
+		if strEqualFold(s, asName) {
+			return abi.As(i), true
+		}
+	}
+	return 0, false
+}

internal/native/arm64/utils.go

@@ -0,0 +1,112 @@
+// Copyright (C) 2026 武汉凹语言科技有限公司
+// SPDX-License-Identifier: AGPL-3.0-or-later
+
+package arm64
+
+import (
+	"fmt"
+	"unicode"
+	"unicode/utf8"
+)
+
+func assert(ok bool, message ...interface{}) {
+	if !ok {
+		if len(message) != 0 {
+			panic(fmt.Sprint(append([]interface{}{"assert failed:"}, message...)...))
+		} else {
+			panic("assert failed")
+		}
+	}
+}
+
+// 忽略大小写
+// 下划线和"."视作相同
+func strEqualFold(s, t string) bool {
+	// ASCII fast path
+	i := 0
+	for ; i < len(s) && i < len(t); i++ {
+		sr := s[i]
+		tr := t[i]
+		if sr|tr >= utf8.RuneSelf {
+			goto hasUnicode
+		}
+
+		// Easy case.
+		if tr == sr {
+			continue
+		}
+
+		// Make sr < tr to simplify what follows.
+		if tr < sr {
+			tr, sr = sr, tr
+		}
+		// ASCII only, sr/tr must be upper/lower case
+		if 'A' <= sr && sr <= 'Z' && tr == sr+'a'-'A' {
+			continue
+		}
+		// '_' 和 '.' 视作相等
+		if (sr == '_' && tr == '.') || (sr == '.' && tr == '_') {
+			continue
+		}
+		return false
+	}
+	// Check if we've exhausted both strings.
+	return len(s) == len(t)
+
+hasUnicode:
+	s = s[i:]
+	t = t[i:]
+	for _, sr := range s {
+		// If t is exhausted the strings are not equal.
+		if len(t) == 0 {
+			return false
+		}
+
+		// Extract first rune from second string.
+		var tr rune
+		if t[0] < utf8.RuneSelf {
+			tr, t = rune(t[0]), t[1:]
+		} else {
+			r, size := utf8.DecodeRuneInString(t)
+			tr, t = r, t[size:]
+		}
+
+		// If they match, keep going; if not, return false.
+
+		// Easy case.
+		if tr == sr {
+			continue
+		}
+
+		// Make sr < tr to simplify what follows.
+		if tr < sr {
+			tr, sr = sr, tr
+		}
+		// Fast check for ASCII.
+		if tr < utf8.RuneSelf {
+			// ASCII only, sr/tr must be upper/lower case
+			if 'A' <= sr && sr <= 'Z' && tr == sr+'a'-'A' {
+				continue
+			}
+			// '_' 和 '.' 视作相等
+			if (sr == '_' && tr == '.') || (sr == '.' && tr == '_') {
+				continue
+			}
+			return false
+		}
+
+		// General case. SimpleFold(x) returns the next equivalent rune > x
+		// or wraps around to smaller values.
+		r := unicode.SimpleFold(sr)
+		for r != sr && r < tr {
+			r = unicode.SimpleFold(r)
+		}
+		if r == tr {
+			continue
+		}
+		return false
+	}
+
+	// First string is empty, so check if the second one is also empty.
+	return len(t) == 0
+}

internal/native/parser/parser.go

@@ -8,6 +8,7 @@ import (
 	"strconv"
 
 	"wa-lang.org/wa/internal/native/abi"
+	"wa-lang.org/wa/internal/native/arm64"
 	"wa-lang.org/wa/internal/native/ast"
 	"wa-lang.org/wa/internal/native/loong64"
 	"wa-lang.org/wa/internal/native/riscv"
@@ -63,6 +64,11 @@ func newParser(cpu abi.CPUType, fset *token.FileSet, filename string, src []byte
 	p.gasSectionAlign = make(map[string]int)
 	p.gasGlobl = make(map[string]bool)
 
+	useArmStyleComment := false
+	if cpu == abi.ARM64 {
+		useArmStyleComment = true
+	}
+
 	switch cpu {
 	case abi.LOONG64:
 		p.scanner = scanner.NewScanner(
@@ -77,6 +83,7 @@ func newParser(cpu abi.CPUType, fset *token.FileSet, filename string, src []byte
 				}
 				return token.NONE
 			},
+			useArmStyleComment,
 		)
 	case abi.RISCV32, abi.RISCV64:
 		p.scanner = scanner.NewScanner(
@@ -91,6 +98,7 @@ func newParser(cpu abi.CPUType, fset *token.FileSet, filename string, src []byte
 				}
 				return token.NONE
 			},
+			useArmStyleComment,
 		)
 	case abi.X64Unix, abi.X64Windows:
 		p.scanner = scanner.NewScanner(
@@ -105,6 +113,22 @@ func newParser(cpu abi.CPUType, fset *token.FileSet, filename string, src []byte
 				}
 				return token.NONE
 			},
+			useArmStyleComment,
+		)
+	case abi.ARM64:
+		p.scanner = scanner.NewScanner(
+			func(ident string) token.Token {
+				// 将原始的寄存器映射到 token.Token 编码
+				if reg, ok := arm64.LookupRegister(ident); ok {
+					return token.REG_ARM64_BEGIN + token.Token(reg)
+				}
+				// 将原始的指令映射到 token.Token 编码
+				if as, ok := arm64.LookupAs(ident); ok {
+					return token.A_ARM64_BEGIN + token.Token(as)
+				}
+				return token.NONE
+			},
+			useArmStyleComment,
 		)
 	default:
 		panic(fmt.Errorf("unknown cpu: %v", cpu))

internal/native/scanner/scanner.go

@@ -30,6 +30,8 @@ type ErrorHandler func(pos token.Position, msg string)
 type Scanner struct {
 	// 查询特定平台的指令和寄存器
 	lookupRegisterOrAs func(ident string) token.Token
+	// 是否为ARM64风格的注释
+	useArmStyleComment bool
 
 	// immutable state
 	file *token.File  // source file handle
@@ -52,9 +54,10 @@ type Scanner struct {
 
 const bom = 0xFEFF // byte order mark, only permitted as very first character
 
-func NewScanner(lookupRegisterOrAs func(ident string) token.Token) *Scanner {
+func NewScanner(lookupRegisterOrAs func(ident string) token.Token, useArmStyleComment bool) *Scanner {
 	s := &Scanner{}
 	s.lookupRegisterOrAs = lookupRegisterOrAs
+	s.useArmStyleComment = useArmStyleComment
 	return s
 }
 
@@ -747,16 +750,43 @@ scanAgain:
 			insertSemi = true
 			tok = token.RBRACK
 		case '#':
-			// #-style comment
-			if s.insertSemi && s.findLineEnd('#') {
+			if s.useArmStyleComment {
+				// ARM语法中 #123 为立即数
+				insertSemi = true
+				tok, lit = s.scanNumber()
+			} else {
+				// #-style comment
+				if s.insertSemi && s.findLineEnd('#') {
+					// reset position to the beginning of the comment
+					s.ch = '#'
+					s.offset = s.file.Offset(pos)
+					s.rdOffset = s.offset + 1
+					s.insertSemi = false // newline consumed
+					return pos, token.SEMICOLON, "\n"
+				}
+				comment := s.scanComment('#')
+				if s.mode&ScanComments == 0 {
+					// skip comment
+					s.insertSemi = false // newline consumed
+					goto scanAgain
+				}
+				tok = token.COMMENT
+				lit = comment
+			}
+
+		case '/':
+			// comment
+			// 对于 ARM64, # 被用于立即数前缀, 因此需要提供不同的注释语法
+			// 对于格式化时, 也需要做特别处理
+			if s.insertSemi && s.findLineEnd('/') {
 				// reset position to the beginning of the comment
-				s.ch = '#'
+				s.ch = '/'
 				s.offset = s.file.Offset(pos)
 				s.rdOffset = s.offset + 1
 				s.insertSemi = false // newline consumed
 				return pos, token.SEMICOLON, "\n"
 			}
-			comment := s.scanComment('#')
+			comment := s.scanComment('/')
 			if s.mode&ScanComments == 0 {
 				// skip comment
 				s.insertSemi = false // newline consumed

internal/native/token/token.go

@@ -106,12 +106,14 @@ const (
 	REG_LOONG_BEGIN Token = 1000 + 100*iota
 	REG_RISCV_BEGIN
 	REG_X64_BEGIN
+	REG_ARM64_BEGIN
 
 	REG_BEGIN     = REG_LOONG_BEGIN
 	REG_LOONG_END = REG_LOONG_BEGIN + 100
 	REG_RISCV_END = REG_RISCV_BEGIN + 100
 	REG_X64_END   = REG_X64_BEGIN + 100
-	REG_END       = REG_X64_END
+	REG_ARM64_END = REG_ARM64_BEGIN + 100
+	REG_END       = REG_ARM64_END
 )
 
 // 指令到 Token 空间的映射
@@ -121,12 +123,14 @@ const (
 	A_LOONG_BEGIN Token = 2000 + 2000*iota
 	A_RISCV_BEGIN
 	A_X64_BEGIN
+	A_ARM64_BEGIN
 	A_END
 
 	A_BEGIN     = A_LOONG_BEGIN
 	A_LOONG_END = A_LOONG_BEGIN + 2000
 	A_RISCV_END = A_RISCV_BEGIN + 2000
 	A_X64_END   = A_X64_BEGIN + 2000
+	A_ARM64_END = A_ARM64_BEGIN + 2000
 )
 
 var tokens = [...]string{
@@ -287,6 +291,9 @@ func (tok Token) RawReg() abi.RegType {
 	if REG_X64_BEGIN <= tok && tok < REG_X64_END {
 		return abi.RegType(tok - REG_X64_BEGIN)
 	}
+	if REG_ARM64_BEGIN <= tok && tok < REG_ARM64_END {
+		return abi.RegType(tok - REG_ARM64_BEGIN)
+	}
 	return 0
 }
 
@@ -301,6 +308,9 @@ func (tok Token) RawAs() abi.As {
 	if A_X64_BEGIN <= tok && tok < A_X64_END {
 		return abi.As(tok - A_X64_BEGIN)
 	}
+	if A_ARM64_BEGIN <= tok && tok < A_ARM64_END {
+		return abi.As(tok - A_ARM64_BEGIN)
+	}
 	return 0
 }