A Moogle  🏳️‍🌈 💻 🎮 🎶 🇸🇪

.. is ever populated. Even the HotGen ones aren't perfectly stable. There's just enough different with the SoC to be annoying, but in many cases I feel like I'm working with incorrect data for the video.

Lazertag (the one using the Div opcodes) is an interesting one because it's clearly yet another developer (Tiger) with a different codebase again, so it's likely doing some other things differently too.

Really they need a good sit down with the code and some extensive studying, but people such as Tahg are far better at doing that than I am. I point to Gormiti because it's one of the simpler cases being just ROM based, rather than loading into RAM, thus ruling out problem sources

Gormiti is interesting. It seems to be setting up the tmap0 tile graphics base to be 004188b7, and it also uses "RegSet" mode. Just wondering, could it be that it should be hitting the CS2 space (which contains that range) instead of going through space->read_word?

it's possible, but IIRC the main problem is that the tables in main RAM used for the tilemaps are blank, same with Wireless Hunting, Triple Header Sports, and one of the Walking Dead games at least. There's some convoluted code to set up a bunch of stuff....

... that includes a pointer to where the actual tile definitions are, but after creating that table it does nothing with it, doesn't seem to set any pointers to it, and never copies data.

I left some notes about that function (in the case of Wireless Hunting) at the top of generalplus_gpl16250soc_video.cpp. Gormiti has basically the same type of thing going on tho. I did wonder if it was some 'static tilemap' mode where it doesn't need copying to RAM, but I'd ...

... still need some pointers to these structures in the video regs for that, and there aren't any.