[Anh302kay]

3DS Port using citro2d for video, and SDL2 for audio. idk if I was supposed put my name in the credits so I didn't do it.

• Uses a software renderer
• Compiles into a CIA
• Banner logos were taken from Google, but it would nice for it to use the logo in the graphics folder, but I didn't know how to convert it into a PNG
• Sonic Advance 2 tested and works on real hardware
• Sonic Advance 1 runs, but crashes when it shows stage results during a memcpy on line 608 of stage_results.c

[freshollie]

Hey. Sorry I've not got round to looking at this (and TBH ports are not a priority for this project)

Some questions:

• Did you use AI for this?
• What is this array? Why is it needed? You are mapping gba pixels to 3ds pixels, but why do you need a precomputed array for this? (smells of AI)
• Why does this need to use SDL for sound output? You'd think you could just output audio directly.

[Anh302kay]

Hello, apologies for the poor quality of the port.

• I did not use any AI for this. I was very tempted to for the banner logos to make imagemagick scripts to convert the ones in the graphic folders but it is not allowed.
• The 3DS stores its textures in a way that requires all textures to be swizzled, and they are not in a simple 1d array. I got the original algorithm for this from here. The array maps pixels so they can be viewed properly on the 3DS. It was originally a function, but I precomputed it into a table with a small separate program to try and improve performance, as the performance was 5-25 fps. I think it had a minimal effect, but when I saw the precomputed table, I realised that the swizzled texture kept pixels together in pairs, so I could map pixels and copy them to the texture in pairs also, which was what ultimately improved performance the most. I can see if there is any noticeable change in performance, and rework the port to reuse the function instead of the precomputed table.
• I cannot lie, it was just because I was extremely lazy and did not want to touch 3DS audio, because I haven't learnt it(I normally use SDL for audio in my 3DS projects). I can try to learn it, and implement it directly.

I cannot upload the program I made to precompute it so I will just paste it here.

#include <iostream>

static size_t posToTex(size_t index)
{
    const int width = 512;
    const int x = index % 400;
    const int y = index / 400;

    // printf("%i , %i \n", x, y);

    return ((((y >> 3) * (width >> 3) + (x >> 3)) << 6)
            + ((x & 1) | ((y & 1) << 1) | ((x & 2) << 1) | ((y & 2) << 2) | ((x & 4) << 2) | ((y & 4) << 3)));
}

int main(){

for(int i = 0; i < 400*240; i++) {
if(i % 2 == 0)
std::cout << posToTex(i)/2 << ", ";
if(i % 16 == 0) std::cout << "\n";
}


}

[freshollie]

Thanks for the info. You could just pre-compute this table at start-up if it does have a performance improvement. Having a static array compiled like this is just way to long to keep in source.

[Anh302kay]

I just updated it to be computed at start-up. I didn't see any noticeable change in performance, but I measured the speed of the table and the speed of the function using the 3DS' built-in tick counter, and the table was faster by roughly 350 microseconds.