Fix RDRAM initialization

[Rosalie241]

This patch does the following:

1. It changes the corrupt handler's toggle by checking all modules its CC value
2. It fixes the idfield_value function which incorrectly captured the 16-24th bits instead of the 8-16th bits
3. It only captures the first 15 bits instead of the full 16 bits for comparison
4. It adds support for the swapfield value of the AddressSelect register
5. It adds a bounds check to read_rdram_dram_corrupted for the dram address like read_rdram_dram has
6. It fixes the mapping.end value in map_corrupt_rdram to match the one in device.c

This patch allows the libdragon IPL to boot without using hacks like setting the RI_SELECT register to a nonzero value or checking which register is targeted by a broadcast.

I want to thank @rasky for helping me debug some issues.

You can test this code using a simple n64-systemtest build with the libdragon IPL which only runs 1 test: n64-systemtest.zip

You can view the documentation here:

• https://n64brew.dev/wiki/RDRAM_Interface#Memory_addressing
• https://n64brew.dev/wiki/RDRAM#RDRAM_addressing

[kapstok]

Maybe this is not the right place to post this, but I tried to compile and run your version and a new version from ui console and sdl input. This version froze on running the n64-systemtest rom and other roms that use libdragon. Are there also other dependencies that have to be built first? I downloaded and installed the 2.6.0 release as base for the other dependencies.

[Rosalie241]

Maybe this is not the right place to post this, but I tried to compile and run your version and a new version from ui console and sdl input. This version froze on running the n64-systemtest rom and other roms that use libdragon. Are there also other dependencies that have to be built first? I downloaded and installed the 2.6.0 release as base for the other dependencies.

Please read the PR carefully:

This patch allows the libdragon IPL to boot without using hacks like setting the RI_SELECT register to a nonzero value or checking which register is targeted by a broadcast.

this patch only allows the libdragon IPL to boot, it doesn't make n64-systemtest fully pass or make libdragon games run properly, the latter requires RSP accuracy improvements which I might try to do sometime by backporting some simple64 patches and upstreaming those.

I'd also like to mention that a LLE RSP and video plugin will be required regardless (for example parallel-rsp and angrylion-rdp-plus or parallel-rdp) alongside using the cached or pure interpreter, the dynamic recompiler has many issues.

[kapstok]

Ah, thank you for clarifying :)

Thought that the boot of the libdragon IPL was the only obstacle that prevented mupen from fully running libdragon roms.