zsnes port to C language

[flaggy]

It looks like you guys started porting things to C, but then stoped. Is there a reason for that? Is there something that just can't be ported to C? It would be very good if everything was written in C because then we could compile zsnes to 64 bits intel processor and several other targets. It would also be interesting to have a arm port of zsnes. All that would be much easier if zsnes were 100% C code.

[Gil_Hamilton]

It's ongoing. Important people have bad cases of real life syndrome, which has several severe symptoms including "no free time for hobby coding".

And I still maintain that whatever comes out won't be ZSNES, since the entire emulation core will be new.

And absolutely everyone knows the advantages. It's why porting a shit-ton of cryptic x86 assembly to C was ever started.

[grinvader]

And by important people he means everyone involved.

And I still maintain that whatever comes out won't be ZSNES, since the entire emulation core will be new.

Well, it's about the spirit imo. A program's identity stems from its guts, direct consequence of the skills and quirks of its authors.
So if there's a reason it won't be ZSNES, it's because zsknight and _Demo_ aren't doing it, essentially.

[flaggy]

I was just curious whether rewriting things in C was a worthy goal, or if it was hopeless. I decided I'd have a go at it. So I began with init function. I found this part

Code: Select all

    ; Initialize volume
    xor edx,edx
    movzx eax,byte[MusicRelVol]
    shl eax,7
    mov ebx,0A3D70A3Dh
    mul ebx
    shr edx,6
    cmp dl,127
    jb .noof
    mov dl,127
.noof
    mov [MusicVol],dl

which I translated to

Code: Select all

    int64_t vol = MusicRelVol << 7;                    
    vol *= 0xA3D70A3D;                                 
    vol >>= 38;
    if (vol > 127)                                     
        vol = 127;
    MusicVol = vol;

Still, I'm puzzled as what does all those shift and multiplying by 0xA3D70A3D means. I ploted the function

((x<<7) * 0xA3D70A3D) >> 38

and I figured that it's very close to the function f(x) = 1.28x - 0.5.

Still, I had no further insight as why to use such an awkward function for translating MusicRelVol into MusicVol. It seems to me that MusicRelVol goes from 0 to 100 and MusicVol goes from 0 to 127. Anybody care to join me in this task?

[flaggy]

I have set up a public repository with the development I'm doing. Everyone is welcomed to help me out making the transition from assembly to C:

https://github.com/aflag/zsnes

I hope eventually we'll be able to merge any changes back to the main development branch.

[kode54]

That was from me being really "clever" by borrowing fixed point multiplication optimizations from Intel's C/C++ compiler a long time ago. No better reason, really.

[grinvader]

Code: Select all

    shr edx,6

>

Code: Select all

    vol >>= 38;

Don't forget to check if your port actually does what the original did while you're at it.

(Edit: i.e. killing eax' contents does not work in other spots)


Regarding your attempt: you'd think starting at init is the right choice, but shortly thereafter you hit the main gui jump and execloop... If you're fine with assembly, feel free to actually trace it out. It's a unique experience.
Mind you, all that block is gonna die with the rewrite. Some of the messy stuff in there causes a good half of the issues.

If you want to help, please join the irc dev channel and hit us up, we always welcome new slaves^Whelpers.

[Fras]

Wasn't there another C port attempt fairly recently, by the way? I don't remember his name, but I think he had done a decent bit of porting before he got taken away by real life.

[grinvader]

Yes, Tron did a lot of work.

[kode54]

Oh yeah, now I remember, I did that because Intel compiler thought that code like that was faster than using division opcodes. Seriously, just use *128/100.