A compiler for javascript targeting LLVM IR, x86 assembly , self interpreting etc...
- How a compiler work and how to build one
- How does code run from source
- How to compile code to Assembly or LLVM IR
import {
Interpreter
} from './src'
let code = `
console.log('=== MemberExpression evaluated =========');
function fn(){
console.log('=== FunctionDeclaration evaluated =====')
}
fn()
`
new Interpreter(code).interpret()
/* output:
* === MemberExpression evaluated =========
* === FunctionDeclaration evaluated =====
*/import {Assembler} from './src'
let kernalCode = `
function printChar($$c){
console.log(1,$$c,1)
}
function printHelper(n) {
if(n>9){
printHelper(n / 10)
}
// 48 is the ASCII code for '0'
printChar(48 + n % 10);
}
function print(n){
printHelper(n)
printChar(10); // 换行
}`
let code = `
function main(){
print(231)
print(4)
}`
let result = new Assembler(kernal+code).compile()result.assembly
.global _main
.text
printChar:
PUSH RBP
MOV RBP, RSP
PUSH 1
MOV RAX, RBP
ADD RAX, 16 # $$c
PUSH RAX
PUSH 1
POP RDX
POP RSI
POP RDI
MOV RAX, 33554436
SYSCALL
PUSH RAX
POP RAX
POP RBP
RET
printHelper:
PUSH RBP
MOV RBP, RSP
# If
# >
PUSH [RBP + 16] # n
PUSH 9
POP RAX
CMP [RSP], RAX
MOV RAX, 0
MOV DWORD PTR [RSP], 1
CMOVA RAX, [RSP]
MOV [RSP], RAX
# End >
POP RAX
TEST RAX, RAX
JZ .else_branch1
# If then
# DIV
PUSH [RBP + 16] # n
PUSH 10
POP RAX
XCHG [RSP], RAX
XOR RDX, RDX
DIV QWORD PTR [RSP]
MOV [RSP], RAX
CALL printHelper
MOV [RSP], RAX
JMP .after_else_branch1
# If else
.else_branch1:
PUSH 0 # Null else branch
.after_else_branch1:
# End if
POP RAX # Ignore non-final expression
# ADD
PUSH 48
# DIV
PUSH [RBP + 16] # n
PUSH 10
POP RAX
XCHG [RSP], RAX
XOR RDX, RDX
DIV QWORD PTR [RSP]
MOV [RSP], RDX
POP RAX
ADD [RSP], RAX
# End ADD
CALL printChar
MOV [RSP], RAX
POP RAX
POP RBP
RET
print:
PUSH RBP
MOV RBP, RSP
PUSH [RBP + 16] # n
CALL printHelper
MOV [RSP], RAX
POP RAX # Ignore non-final expression
PUSH 10
CALL printChar
MOV [RSP], RAX
POP RAX
POP RBP
RET
main:
PUSH RBP
MOV RBP, RSP
PUSH 231
CALL print
MOV [RSP], RAX
POP RAX # Ignore non-final expression
PUSH 4
CALL print
MOV [RSP], RAX
POP RAX
POP RBP
RET
_main:
CALL main
MOV RDI, RAX
MOV RAX, 33554433
SYSCALL
import {Assembler} from './src'
let kernalCode = `
function printChar($$c){
console.log(1,$$c,1)
}
function printHelper(n) {
if(n>9){
printHelper(n / 10)
}
// 48 is the ASCII code for '0'
printChar(48 + n % 10);
}
function print(n){
printHelper(n)
printChar(10); // 换行
}`
let code = `
function main(){
print(311)
// print(4)
}`
let result = new Assembler(kernal+code).llvmCompile()result.assembly
define i64 @printChar(i64 %$$c) {
%sym5 = add i64 1, 0
%sym7 = add i64 %$$c, 0
%sym6 = alloca i64, align 4
store i64 %sym7, i64* %sym6, align 4
%sym8 = add i64 1, 0
%sym9 = add i64 33554436, 0
%sym4 = call i64 asm sideeffect "syscall", "=r,{rax},{rdi},{rsi},{rdx},~{dirflag},~{fpsr},~{flags}" (i64 %sym9, i64 %sym5, i64* %sym6, i64 %sym8)
}
define i64 @printHelper(i64 %n) {
%ifresult8 = alloca i64, align 4
%sym9 = add i64 %n, 0
%sym10 = add i64 9, 0
%sym7 = icmp sgt i64 %sym9, %sym10
br i1 %sym7, label %iftrue11, label %iffalse12
iftrue11:
%sym15 = add i64 %n, 0
%sym16 = add i64 10, 0
%sym14 = udiv i64 %sym15, %sym16
%sym13 = call i64 @printHelper(i64 %sym14)
store i64 %sym13, i64* %ifresult8, align 4
br label %ifend17
iffalse12:
br label %ifend17
ifend17:
%sym6 = load i64, i64* %ifresult8, align 4
%sym19 = add i64 48, 0
%sym21 = add i64 %n, 0
%sym22 = add i64 10, 0
%sym20 = urem i64 %sym21, %sym22
%sym18 = add i64 %sym19, %sym20
%sym6 = call i64 @printChar(i64 %sym18)
}
define i64 @print(i64 %n) {
%sym9 = add i64 %n, 0
%sym8 = call i64 @printHelper(i64 %sym9)
%sym10 = add i64 10, 0
%sym8 = call i64 @printChar(i64 %sym10)
}
define i64 @main() {
%sym10 = add i64 311, 0
%sym9 = call i64 @print(i64 %sym10)
}npm install
Prepare "gcc" and "llc" enviroment or else only interpret.spec.ts could pass the test
npm run test
Further info could reference test cases
- Manual
- Tutorial
- Manual
- Tutorial
- Article