Struct Spu 

pub struct Spu(/* private fields */);

The SPU’s memory-moving driver: it owns the SPU’s DMA channel, so the one operation that copies RAM into sound RAM, the ADPCM upload, is a method here and the borrow checker sees a second uploader.

The voice, volume and reverb registers are free functions in this crate: writing one cannot touch RAM, so they need no owner yet.

let mut spu = psx_spu::Spu::new(peripherals.spu_dma);
spu.upload_adpcm(SpuAddr::new(0x1010), &TONE_SAMPLE);

Implementations

impl Spu

pub fn new(dma: SpuDma) -> Self

Take over the SPU: reset it to a sane playable state and write the silence block every one-shot’s repeat address points at.

Sets:

• SPUCNT: enable, unmute, default reverb off, CD audio off
• Main volume: max
• Every voice: silenced (volume 0, ADSR release fire, key-off)
• Reverb input and wet-output depth, pitch-mod, and noise: all disabled
• Transfer mode: “normal” (the upload path we expose)

Call once at boot before any voice operations.

pub fn upload_adpcm(&mut self, dest: SpuAddr, bytes: &[u8])

Upload ADPCM sample bytes to SPU RAM.

Goes over DMA channel 4 when bytes is word-aligned and a whole number of words, which is fast and the only reliable path past the 32-halfword transfer FIFO, and otherwise falls back to the manual-transfer (PIO) path, one halfword per write.

bytes must be a multiple of 2 (one halfword per two bytes). dest must be 8-byte-aligned (the SPU voice-start register’s resolution) and above the silence block.

Panics

If bytes is odd in length, or dest overlaps the silence block at 0x1000.

pub fn release(self) -> SpuDma

Give the DMA token back.

Trait Implementations

impl Debug for Spu

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.