My blog posts:
This project is an optmizing Lisp compiler and bytecode VM.
It can compile to JavaScript, or compile to bytecode and execute in a VM.
The bytecode VM is a little faster than Node.js (TODO: benchmarks).
Bytecode VM:
./program --vm --debug < fib10.lisp
0: PushClosure(["n"])
-> 0: LoadVar("n")
-> 1: PushConst(2.0)
-> 2: LessThan
-> 3: Jump(6) // go to 6
-> 4: LoadVar("n")
-> 5: Jump(17) // exit
-> 6: LoadVar("n")
-> 7: PushConst(1.0)
-> 8: Sub(2)
-> 9: LoadVar("fib")
-> 10: CallLambda(1)
-> 11: LoadVar("n")
-> 12: PushConst(2.0)
-> 13: Sub(2)
-> 14: LoadVar("fib")
-> 15: CallLambda(1)
-> 16: Add(2)
1: StoreVar("fib")
2: PushConst(10.0)
3: LoadVar("fib")
4: CallLambda(1)
5: LoadVar("print")
6: CallLambda(1)
55
Compile to JavaScript:
./program --js < fib10.lisp
/*
(let ((fib (lambda (n)
(if (< n 2)
n
(+ (fib (- n 1)) (fib (- n 2)))))))
(print (fib 10)))
*/
let print = console.log;
let fib = (n) => n < 2 ? n : (fib(n - 1) + fib(n - 2));
print(fib(10));
The implemented optimizations are constant folding and propagation, and dead code elimination:
; before optimization
(let ((b 2) (c 3))
(print
(+
(+ b 4 c)
(- b c 7)
)))
; after optimization
(let () (print 1))
The Lisp variant is very similar to Little Lisp.
; atoms
1 ; f64 numbers
a ; symbols
; arithmetic expressions
(+ 1 2) ; 3
(- 1 2) ; -1
; control flow expressions
(< 1 2) ; true
(> 1 2) ; false
(if (< 1 2) (+ 10 10) (+ 10 5)) ; 20
; lambda expressions
(lambda (x) (+ x x)) ; function that doubles
; variable definition
(let ((a 1)) (print a)) ; prints 1
(let ((double (lambda (x) (+ x x)))) (double 2)) ; 4
Required (one of):
--js
output JavaScript to stdout--vm
compile to bytecode and execute in VM
Optional:
--optimize
for optimization--debug
show annotated bytecode
cargo test