Resident evil for Atari

[shoggoth]

darksun9210 wrote:
and i think the 3D accelerator project just ended up being vapour.

If you're referring to the SuperVIDEL project, it's definitely alive. I saw the hardware myself a few weeks ago. The CTPCI project is alive as well; I tried it myself on a ColdFire evaluation board.

hmmm, as the binareys are 68020, maybe someone could emulate? :lol:

Provided that the binaries only use basic OS functions, it should be fairly easy to implement some kind of OS wrapper functionality on the Amiga. Once you go GUI, or rely on unix functionality, it's a lot more difficult. This in turn generally means applications which tend to bang the hardware directly, which makes it downright impossible to emulate at sensible speeds.

[shoggoth]

countzero wrote:
well I got some bad news for you then. we had a fabled coldfire board in the works too, but in the end it turned out coldfire is not fully 68k compatible and the software layer which translates 68k to coldfire code dropped the performance to sinilar level to our 50 MHz 060 boards. Elbox canceled the project completely. I'm sure your 100MHz CT63 boards would kick the hell out of any coldfire board.

I've done some research on this topic on "that other 68k" platform. Basically, one has to deal with three cases.

A: applications which are inherently incompatible with the CF and need to be patched

B: applications which are compatible enough (emulation by means of trapped emulation of missing instructions)

C: applications which works out of the box.

Case C is rare, case B is not completely uncommon, and A has to be handled through user land emulation. All these cases has to be catered for by the OS, which means you either have to patch your OS completely or get access to the source code.

Supervisor mode can be dealt with (at least on the platform I'm referring to). I've worked on case A (user land emulation). Performance suffer greatly for those apps, but since you don't emulate hardware nor OS functions, it isn't that obvious to the user for everyday apps. An interesting approach would be JIT emulation - HP did an experiment once, where they used some form av JIT emulation where the host architecture was identical to the guest architecture. The interesting part is that in some cases, the JIT emulated environment ran faster than the "real" conterpart, due to the fact that the JIT compiler could eliminate unneeded checks in the code (branches etc) since it had enough runtime information to completely rule out certain execution paths. If this was possible on the CF (I don't know if it is), case A could be pretty fast too.

All this means that CF incompatibilites can't be solved by nifty hardware solutions; it can only be solved on the OS level.

I don't know how well all this applies to the Amiga though. Why ColdFire? Personally I like 68k coding, but I want new hardware.

[shoggoth]

darksun9210 wrote:
Shoggoth did pen the following...

If you're referring to the SuperVIDEL project, it's definitely alive. I saw the hardware myself a few weeks ago. The CTPCI project is alive as well; I tried it myself on a ColdFire evaluation board.

Sweet!  :-D

The SuperVIDEL concept is really neat. The CT60 is designed to allow expansion boards to replace legacy hardware registers. In the case of the SuperVIDEL, this means it's compatible "out of the box", even without drivers. It's also capable of snooping accesses to ST-ram (i.e. chipram) to provide support for apps that aren't SuperVIDEL-aware.

A really nice solution imo, can't wait to try it out. Just like the Amiga, these machines suffer from poor video ram bandwidth (and planar graphics in <=8bpp), which means a lot of CPU cycles are wasted just for C2P conversion and copying to video ram. I predict approx 2x speed increase for some of my projects, perhaps even more since I won't have to buffer & convert stuff.