Hooooooly SHIT. I just came home to Lamb2k1 taking us from slideshow to PLAYABLE AND FUCKING GORGEOUS with his upcoming port of Resident Evil 4 to the Sega Dreamcast!!! GAAAAINZ!! Here's a direct hardware capture of me playing through to the starting cabin with the first enemy encounter... with settings enabled for high visual fidelity, such as 2x horizontal FSAA and far draw distance. Evidently Lamb2k1 and his audacious little clankers have been busy AF, starting with him telling them to redo all of the renderer and math in the port to use my SH4ZAM library to the fullest... which is funny, cuz I just made a few more micro gainz that I have yet to push for some of these commonly used matrix routines! BUT additionally, he's done some voodo such as: 1) Leveraging a special Katana-like "chunking" model format for the rooms. 2) Precomputing a lot of things up-front for room AI. 3) Culling WAY more aggressively than the GC or PS2 versions were doing. 4) Leon's jacket doesn't update as frequently anymore... lmao! There's still a LONG ways to go, and it slows down plenty under load, but... WHOAH.

I don't even know wtf happened, man... I swore I wouldn't get sucked into any of these vibed-up, fully AI-based Sega Dreamcast ports that have been popping up everywhere, all of a sudden... ...but here I am, 6+ hours deep into mapping the matrix skinning and TnL math from Lamb2k's Resident Evil 4 port from Gamecube-targeted routines to SH4ZAM, my accelerated math library for the Sega Dreamcast's SH4 CPU. The prospect of one of my all-time favorite games coming to my all-time favorite console has given me the courage to wade through this Clanker-generated hellish nightmare of AI slop code in search of GAINZ for our little white box. THANK GOD I added first-class support for 3x4 transform matrices to SH4ZAM's XMTRX API, including loading, saving, and operating on their transposes... because of course, the Dreamcast is chad 4x4 column-major at the HW-level, while the Lamecube loved 3x4 row-major matrices... ...as you can see from this direct hardware capture, the game already looks absolutely stunning at the Dreamcast's beautiful and crisp 480p with 2x horizontal FSAA enabled, but... the real question is now HOW IS IT GOING TO RUN IN THE END? NO CLUE, because I'm still trying to get a grasp on just how well the AI optimized the port, given that portions of the code look like absolute dog-shit, totally ignoring the vector instructions on the SH4 FPU for critical-path math... but at the same time, from what I can gather from the clusterfuck that is the renderer, the actual vertex submission code is quite efficient... So I really can't say precisely just how much room is left for GAINZ here just yet, tbh. The port is DEFINITELY extremely CPU-bound atm, but we'll just have to see where this one takes us!