Need help with homebrew developing

Are you sure the audio file is not too big to fit into the 3DS's memory? It probably crashes when you allocate the memory. Try streaming it instead, like in the audio/streaming 3ds example.

 

That's probably still too big when using linearAlloc, I really don't understand the differences between that and the standard malloc in devkitpro but memory allocated with linearAlloc seems to be some special kind of memory that works with the DSP. With streaming you'll only have to allocate a few kilobytes with linearAlloc and the actual audio data can be copied to the buffer either from a separately allocated file buffer or streamed directly from the file.

 

You don't need an extra folder, you can open the file from the same location in the same way. The basic streaming example from 3ds-examples is here: https://github.com/devkitPro/3ds-examples/blob/master/audio/streaming/source/main.c
You could also look at my attempts of streaming audio if the official example does not explain everything, but there's a lot of other unimportant code in this file: https://github.com/ic-scm/smashcustommusic-client-3ds/blob/master/3ds/source/audio.h
If you're still having trouble figuring it out, I could try writing a simple raw file streamer later.

 

You should remove the code for notes and sine waves, the important thing here is setting up the NDSP buffers and then filling them. In the fill_buffer function, the "s16 sample" variable is the audio sample that is about to be added to the buffer. In the next line it is then duplicated into 2 stereo samples, if you are trying to play stereo audio then you could replace the two mentions of "sample" in that line with two different sample variables. If you had two input buffers (2 channels for stereo) of the same size as the output buffer that is passed to the fill_buffer function, the fill_buffer function could look something like this:

Code:

void fill_buffer(void *audioBuffer,size_t offset, size_t size, s16 *buffer_left, s16 *buffer_right ) {
    u32 *dest = (u32*)audioBuffer;
    
    for (int i=0; i<size; i++) {
        dest[i] = (buffer_left[i] << 16) | (buffer_right[i] & 0xffff);
    }
    
    DSP_FlushDataCache(audioBuffer,size);
}

I'll try writing the streamer in a few hours, currently busy with some other things. Would you be able to attach one of your raw audio files so I can understand the format of it?

 

The difference is regular heap makes no guarantee whatsoever that the virtual memory you map is contiguous physical memory. Linear heap does and it's required for hardware like the DSP because it doesn't know/care about virtual memory.

 

@EinTim I wrote the streaming example: https://github.com/ic-scm/3ds-simple-audio-stream
This still loads the entire file into memory instead of streaming directly from file, but it should work for files that aren't too big to fit into the 3DS memory. The audio file is loaded from romfs:/audio.bin, but you can change that in the code.