Rebb and myself started discovering Dreamcast just recently, so several weeks ago, Rebb came up with an idea that we should try to do a fast prod for Assembly - a suggestion that I couldn't resist, even that it seems impossible from the beginning.
Anyway, even that it ended up as a simple production, a LOT of effort was spent to do it. Unbelievable as it sounds, we fought the unpredictable problems on every single step of a way. And not having a proper dev hardware until a weekend before the party haven't helped (broadband adapter is definitely a must for DC dev, if you ask me).
Regarding the performance... on every single thing we did, we couldn't believe the (bad) performance we're getting. We've dropped several great ideas and effects because they were just unbearable to watch.
That may be caused by bad coding, but
when you have unrolled loop that is a only a call to the polygon draw function, and you're not able to draw more than 100 textured polygons (with alpha transparency) per one frame, then you start questioning other things. E.g. maybe it was caused by the fact that we haven't use a single line of Sega copyrighted code (original SDK), rather the latest KallistiOS (manually compiled from git). Or, maybe we did something wrong during the setup the crossdev, as we used the same how-to, but it is what it is.
(let's me just mention here that I had 100x better polygon performance with the same parameters on 4 years older PSX)
Also, trying to get a knowledge from 10+ years old forum threads isn't much fun. Especially when half of those methods and functions are obsolete in today's version of KOS.
Anyway, we hope that we'll discover how to get around the performance problems (or solve them completely), and that this won't be our last release for Dreamcast. Even that it included too many sleepless nights (most painful was the ones before the trip), I've enjoyed working on the platform, and especially collaborating with all the guys included. And with all the problems we fought, at least it wasn't boring.
) It can always be better, we could work on it for years, and never release it, but here it is, like it or not. It's far from being our best production, but we gave our best in a given time with a given tools.
And if you don't like it, please go make a demo about it.