Games such as Pocket Quake and Leo's Flight Simulator do not rely on the OS or hardware for drawing graphic primitives. They have their own routines that perform the drawing. When the frame is built it is blitted directly to the display hardware. To my knowledge it is possible to directly access the video hardware buffer of any Windows CE device made. Before Windows CE 3.0 there was a "Game API" created by a third party (I don't know his name offhand) that allowed for direct screen access. Even Jimmy Software used this library for their early games on CE 2.11 Palm-Sized PCs. So yes, it is possible for PQ to run on the Jornada 720, etc. I doubt it will run faster than the iPaq, because the only thing that counts in this case is processor speed, memory speed, and emulated floating point math speed. The iPaq has better benchmark results in those areas than any other device (* see note). Some devices (the Casios) greatly excel at OS drawn graphic primitives. Again, the problem is that "real" games do not rely on the those routines, which still may be slower than when the application peforms those functions with its own custom code.<br>While I'm talking about performance I'll bring up another interesting item. I was reading through the docs included with the Quake source code. Many routines, specifically the 3D rendering / drawing code, were written in both C and in optimized x86 assembly. John Carmack says that if the optimized x86 assembly code is used then Quake will run almost twice as fast then when the C routines are used. Of course PQ uses the C code, as assembly is processor specific. So what that means is that if someone were to recode those assembly routines for a Pocket PC processor, then a huge increase in performance would be seen. Of course there is still the Fixed Point optimizations, which result in even greater improvement. So at this point, I believe that it is possible for PQ to run at 60 fps on the iPaq, if key routines were written in assembly, and a global fixed point conversion were done. At this time I have a lot of other "real" responsibilities going on that have kept me from working on PQ. I still follow the discussion boards daily. Hopefully within the next month I'll be able to dedicate a slice of time to PQ, during which I should be able to finish a big fixed point conversion.<br><br>Dan East<br><br>* The @migo churns out slightly faster benchmark results than the iPaq in almost all areas, even though it uses the exact same processor. My little conspiracy theory is that Palmax just slightly overclocked the main board (at the hardware level, probably via the timing crystal, etc), for the
sole purpose of slightly besting the iPaq for publicity / sales reasons. It would only take a few mhz to achieve the 1-2% improvement we see. Further evidence that the system as a whole is running over-speed is that the Media Player played songs too fast (sorta like the Chipmunks). Palmax let that one slip by. They probably had to tweak the OS to make up for timing problems, and missed that one. If there are timing problems with 3rd party games and other multimedia apps then it would further reinforce my theory.
