Compile FASM/MASM to JS

Question

Is it possible to compile assembly or x86/x64 code to JavaScript?

Or asm -> llvm-ir -> js?

UPD: So my final question is: I have asm listing (excerpt)

push %rbp
mov %rsp,%rbp
sub $0x10,%rsp
mov 0x5e41(%rip),%eax
lea 0x601a(%rip),%rdx
cltq
movzbl (%rdx,%rax,1),%eax
test %al,%al
jne loc_0040121b
movl $0x0,-0x4(%rbp)
jmp  loc_00401206

Can I run it in browser without modifications with help of emscriptten or any other tool?

The fact is that asm is not compiled, it is assembled. If you want to emulate a machine in JS maybe it's easier to just use the binary output from a conventional assembler. — Margaret Bloom
– Margaret Bloom, Commented Oct 4, 2016 at 11:08
@Ped7g I want to run executable(small program) in browser without access to original source code — OwnageIsMagic
– OwnageIsMagic, Commented Oct 4, 2016 at 18:40
A quick search didn't lead to any x86_64 emulator, only 32b x86 found. For example this looks impressive bellard.org/jslinux (check technical notes, it's basically full 80486 emulation). Actually no wonder, as JS can't natively handle 64b numbers (only about 53b), so it would have to emulate register values as BigInteger -> probably additional 10x performance burden to the already slow emulation. You can maybe try some disassembler which is trying to reconstruct C code, but I think those died ~15-20 years back, when C compiler optimizers started produce too specialized machine code. — Ped7g
– Ped7g, Commented Oct 5, 2016 at 19:05
@Ped7g this one looks more interesting in my opinion copy.sh/v86 BTW: IDA Pro is still feeling good :) — OwnageIsMagic
– OwnageIsMagic, Commented Oct 5, 2016 at 20:03

Ped7g · Accepted Answer · 2016-10-04 21:57:17Z

1

I want to run executable(small program) in browser without access to original source code

So, you actually don't want to compile assembly to js, as you don't have the source. Which is good, as it doesn't make any sense anyway, asm is not like high-level languages having some variables and affecting platform trough some API, in asm basically any instruction can change state of platform a lot, and "variables" is the whole accessible memory.

A way how it makes sense:

You want to execute some machine code already prepared for some target platform (CPU + OS).

"In browser" (directly) it's not possible, browser doesn't process machine code of that CPU, neither does it provide the OS API/ABI/syscall. That executable needs the target OS, so it's instructions have the desired effect.

You can run emulator of that OS written in JS, and run the binary inside the emulator. Search for example for web-online dosbox (running old DOS games in browser, with reasonable speed and not-worst-accuracy, most of them actually being playable and looking as they should).

How to create emulator of your target platform is a bit too broad question and takes somewhat more effort, like decades in some cases. But you may be lucky, that there already exists emulator of your target OS, and has JS port and license allowing you to use it.

There's also second option: if it's really small program, it would be very likely simpler to rewrite it from scratch in JS and in OS/browser agnostic way.

edit: about machine code -> llvm ir:

Let's say simply "no". (although probably like 50% of machine code would be reasonably easy to translate, but the other half would require the translator to fully understand what the code does, which is fully NP problem (or not possible at all)).

edited Oct 4, 2016 at 21:57

answered Oct 4, 2016 at 21:48

Ped7g

16.7k4 gold badges33 silver badges71 bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

OwnageIsMagic Over a year ago

I think that running OS is overkill for such case, but I somehow forgot that calls to libc are platform dependent operations :)

Ped7g Over a year ago

@OwnageIsMagic: well.. yeah. And now consider the machine code even doesn't have to bother to call libc at all, and can access HW/OS directly.

OwnageIsMagic Over a year ago

I can grantee that program doesnt perform any interupts/syscals (explicit)

OwnageIsMagic Over a year ago

asm and machine code are 1:1 (for a given platform). Why you reference them like something very different?

Ped7g Over a year ago

@OwnageIsMagic: because when you have binary file, you have machine code. To get Assembly you have to first disassemble it. But if you assemble that back into machine code, you may get sometimes slightly different executable, in special case where some instruction has two+ possible opcodes (official/unofficial) and disassembler hides that. Also binary header fields from executable are usually not part of Assembly source, so it's another thing generated from scratch during compilation => another room for different bytes output. So relying on machine code only is safer option in emulation.

Collectives™ on Stack Overflow

Compile FASM/MASM to JS

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related