[rsp] Introduce microcode build process, disassemble and build both rspboot versions

[Thar0]

This PR introduces a build system for rsp microcodes, using armips as the assembler:

• Microcode sources are in a new directory rsp/
• Microcode sources are preprocessed with the C preprocessor
• armips assembles the preprocessed source into separate text and data binary files
• rsp2elf.s is a template file for assembling those binaries into a unified elf file for the microcode
• The elf file is included in the spec for linking into the project

To exercise this build system, both variants of rspboot are included. The anti-piracy variant (rspboot_ap) is used in N64 versions while the non-anti-piracy variant (rspboot) is used in all others.

Other microcodes to come in future PRs.

[Pepe20129]

Wouldn't it be better to combine rspboot.s & rspboot_ap.s and use #ifdefs?

[Thar0]

The two files haven't got enough common text to warrant that, even the regular functions unrelated to anti-piracy measures were rewritten. I think once you get to just a single ifdef for all the code in a file it makes more sense to just have separate files at that point.

[Dragorn421]

Build system looks good, I want to take a closer look at rspboot.s before merging

Can you document how you came up with the single file armips and from which revision of it

[Dragorn421]

I find it mildly interesting that rspboot is assembled from origin IMEM_START_VA instead of IMEM_START like rspboot_ap

[Dragorn421]

for grepability I would personally prefer if this macro took the full OS_TASK_OFF_ macro name directly

Suggested change

	#define OSTASK_FIELD(field) (-0x40 + OS_TASK_OFF_##field)($zero)
	#define OSTASK_FIELD(fieldOffset) (-0x40 + fieldOffset)($zero)

[Thar0]

I don't think it's worth prioritizing grep-friendliness over readability here, the latter should be the common case here

[Dragorn421]

where / by what was the semaphore acquired?

[Thar0]

The semaphore can be released whether or not it was acquired, in which case it's a no-op. This is probably done for deadlock prevention, in case some previous task or the cpu had acquired the semaphore and failed to release it

[Dragorn421]

Suggested change

	// Return the semaphore and jump to the ucode that was loaded
	jr $7
	mtc0 $zero, SP_SEMAPHORE
	// Release the semaphore as a just-in-case measure and jump to the ucode that was loaded
	jr $7
	mtc0 $zero, SP_SEMAPHORE

?

[Dragorn421]

Suggested change

	// Checks that $4, $5, $6 and $11 all have expected register values set by IPL3
	// Checks that $4, $5, $6 and $11 all have expected register values set by the cic6105 microcode

afaik IPL3 never runs anything on the rsp (merely at some point the cpu executes from imem/dmem)

[Thar0]

This isn't the case, IPL3 runs RSP code

[Dragorn421]

Oh 🤔 all flavors of it?

[Thar0]

Not all, but the case we care about (X105) does

[Dragorn421]

Should we add a mention of X105 to this comment?

[Pepe20129]

Apparently it's possible to mark a file as auto-generated so that GitHub hides the changes by default by adding this to the .gitattributes file (docs):

tools/armips.cpp linguist-generated=true

[Dragorn421]

Apparently it's possible to mark a file as auto-generated so that GitHub hides the changes by default by adding this to the .gitattributes file (docs):

tools/armips.cpp linguist-generated=true

Thanks for the suggestion but I don't think we really need that
(I would also rather see changes to this file for security reasons)

[Pepe20129]

(I would also rather see changes to this file for security reasons)

IIRC, it only collapses it by default, you can still expand it, it's mostly to avoid lagging out the ui when you don't want to see the changes in that file.