Also want to ask about guidelines for editor config, e.g.

• max symbols on line (from the current code it seems you use ~100)
• tabs (currently using 2 spaces for tab as that's what I'm used to and like for Javascript)

Thanks, I’m away for a bit and will review this later.

Regarding editor config, it’s probably a mess since there isn’t auto formatting for Pawn and a couple modules are contributions already.

I personally do 4 spaces and max 120 chars per line, because that’s the width of the github preview on desktop and as such frequently used for a modern limit, so I prefer contributions in that format.

@namazso I've verified these changes working with:

AM4 system
Motherboard: MSI B550 Unify-X
CPUs: 1800X (SummitRidge), 1600AF (PinnacleRidge), 3600 (Matisse), 3900X (Matisse), 5300G (Cezanne), 5600X (Vermeer)

AM5 system
Motherboard: Asus Apex X870
CPUs: 9950X3D, 9600X, 9900X (GraniteRidge), Epyc 4124p, 7950X (Raphael)

APU
6800HS (Rembrandt)

Unfortunately my 240GE seems to have lost all PCI-E lanes and I can't boot to windows or use external VGA, so I'm unable to test RavenRidge desktop.

What is left

• Fix the mapping of physical memory by choosing one of these two options:

1. Remove the second base address from CPU class 3, like here: d4f1557
2. Handle iomap based on the cpu class - introduce global g_class variable and mask g_table_base appropriately, thus rendering the high dword for RavenRidge and derivates obsolete.

Based on my code, these codenames use 32bit address, but I can't be sure this is 100% right.
Also when used with 32bit Windows, all CPUs use the low dword and there's a special command SetToolsDramAddress to set that address for use.

APU
RavenRidge, FireFlight, Dali, Picasso, Renoir, Lucienne, Cezanne, Mero, VanGogh

CPU
SummitRidge, Naples, Whitehaven, PinnacleRidge, Colfax, Matisse, CastlePeak, Rome, Vermeer, Chagall, Milan

I guess it would be fine to use whatever the table base address is returned by SMU with the exception of class 3.

Thanks!

Removing [WIP] tag, it seems to be working on the CPUs me and my users managed to test. Verified to work on Phoenix as well.
Should be alright for a "first" version. The only outsdanding issue would be with the base address of class 3.