platform: clean up Acekard platforms and implement DMA by lifehackerhansol · Pull Request #163 · LNH-team/pico-loader

lifehackerhansol · 2026-04-20T07:24:16Z

Added a "IoRpgPlatformSpecifics" struct which defines some platform-specific magic numbers to use in patch codes
Merged all patch codes into acekard-common directory, leveraging IoRpgPlatformSpecifics to inject magic numbers as needed
Optimized a few routines to reduce patch code space
Implemented DMA support

Gericom · 2026-05-03T08:22:47Z

The dma patch code you added looks very similar. Can we avoid duplication?

lifehackerhansol · 2026-05-03T08:38:11Z

Not really, because of the different parameters each platform needs:

R4iDSN commands start with 0xAB, while others start with 0xD5
AKRPG card state command returns its "state" in the lower nibble, while the rest returns it in the upper nibble

We could programmatically patch those instructions I suppose, but I haven't tried it.

Gericom · 2026-05-03T08:54:39Z

Not really, because of the different parameters each platform needs:
* R4iDSN commands start with 0xAB, while others start with 0xD5

* AKRPG card state command returns its "state" in the lower nibble, while the rest returns it in the upper nibble
We could programmatically patch those instructions I suppose, but I haven't tried it.

I guess another option could be to create an .inc file with the code with some #ifdefs. And then in the actual .s files, set the corresponding ifdef and #include the .inc file.

lifehackerhansol · 2026-05-10T21:46:06Z

I guess another option could be to create an .inc file with the code with some #ifdefs. And then in the actual .s files, set the corresponding ifdef and #include the .inc file.

I went with a slightly different solution inspired by compactflash-common, and unified all patches under acekard-common.

This only gets used in IoRpgCmdSdio which is actually a u8 argument.

…e patch code

Use this struct for any platform-specific data that may need injecting into patch codes.

…e shift The C0 command (read SD state) returns the "state" in different bits of the response: - lower halfbyte on AKRPG - upper halfbyte everywhere else This is defined in the PlatformSpecifics struct, so use that to patch the code on the fly instead of using a register to pass this through.

Most of the code is the same, except for the command sent to the card. Use the new PlatformSpecifics struct to change this on the fly.

Now that SdWaitForState no longer requires passing a platform-specific parameter, the patch code becomes exactly the same.

Most of the code is the same, except for the command sent to the card. Use the new PlatformSpecifics struct to change this on the fly.

- push r0-r1 to stack before the sdsc shift - use r1 for readSd SDSC shift now that it's free, freeing up a register - keep r4 as 0x040001a0 across routine

- use r4=0x040001A0 from previous subroutine - use mov/lsls instead of ldr pc, removing unnecessary pool usage - we can assume that when starting the transfer, the previous command used (SdWaitForState) set our MCCNT1 to 0xA7406004, since we need 0xA1406004 we can just strb the 0xA1 instead of loading entire u32 from pool

- Use r4=0x040001A0 from previous subroutine to reduce pool usage

lifehackerhansol force-pushed the rpg-dma branch from 68e388f to 0b73b0e Compare May 10, 2026 21:44

lifehackerhansol changed the title ~~platform: add DMA support for the RPG family of flashcarts~~ platform: clean up Acekard platforms and implement DMA May 10, 2026

lifehackerhansol force-pushed the rpg-dma branch 2 times, most recently from b9973f9 to ec463ab Compare May 29, 2026 08:05