backend: add infinite_register class method to RegisterType

[superlopuh]

It's nice to have a check to make sure that our fake infinite registers don't clash with the real ones.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 91.30%. Comparing base (e9b1efb) to head (675852e).
Report is 12 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3929      +/-   ##
==========================================
- Coverage   91.30%   91.30%   -0.01%     
==========================================
  Files         467      468       +1     
  Lines       58068    58121      +53     
  Branches     5581     5582       +1     
==========================================
+ Hits        53021    53067      +46     
- Misses       3616     3623       +7     
  Partials     1431     1431              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[JosseVanDelm]

Looks good to me, I have a few suggestions though

[JosseVanDelm]

Suggested change

	return f"sse_{index}"
	return f"josse_{index}"

[JosseVanDelm]

Suggested change

	assert isinstance(res.index, NoneAttr), (
	f"Invalid 'infinite' register name {spelling}"
	)
	assert isinstance(res.index, NoneAttr), (
	f"Invalid 'infinite' register name: {spelling} clashes with finite register set"
	)

[superlopuh]

I changed the logic to just provide the prefix, not the whole thing. One of the questions is do we want _ suffixes everywhere? How important do you think it is to be consistent here, should we allow j0 in riscv and avx2_0 on x86, or should we change to j_0 in riscv land?

[compor]

I changed the logic to just provide the prefix, not the whole thing. One of the questions is do we want _ suffixes everywhere? How important do you think it is to be consistent here, should we allow j0 in riscv and avx2_0 on x86, or should we change to j_0 in riscv land?

Not a biggie for me; leaning slightly towards using _ everywhere, since for the RISC-V side is purely a riscemu thing.
Maybe even slightly easier to search/grep?

[superlopuh]

OK then I'll do a quick PR first doing j_ and fj_ for riscv then rebase this on top of that

[superlopuh]

#3934