| Subject: RE: Ideas on new Atari |
Author: Evan Langlois (67.187.51.187)
Date: 03-12-2005 07:20
Where did you hear that the ACP team wanted to do emulation? What I saw was that they wanted harwdare compatibility. The're design is based on an older Coldfire that doesn't have an MMU, so any compatibility with apps other than the most cleanly written will mean hardware level compatibility and they site this as a goal. I think moving to a newer coldfire with an MMU to remap and emulate the hardware for less-than-perfectly-clean apps is a better way to go. There design has a number of disadvantages in hardware design and complexity. The coldfire already has tons of built-in support for the features that chips like the MFP would add, so adding another chip to perform those capabilities just for compatibility is redundant and adds to the parts count.
They also have a few theories that I feel are completely backwards. For example they site that they won't use ATI video cards because only the 2D API is open and not the 3D API. However, the 3D API for some of the older chips is open .. up to the 9200 I believe. I'm using one at the moment, and I don't use binary ATI drivers, only free Xorg drivers, and my 3D works fine. The error in their logic is that they would use this supervidel chip which doesn't even HAVE a 3D processor to program. Why site the lack of documentation on the 3D hardware if your alternative is to use a design that doesn't have the capability at all. I don't even think the Supervidel has hardware assisted 2D. Is there any support for video scaling or accelerated drawing primitives? Did they even include a bit-blit processor? The 30 bit color doesn't impress me either when the human eye doesn't see all the color you get from 24 bit, yet alone another 2 bits for each color gun. The developement boards for the Coldfire use an S3 chip or something (doesn't say exactly, but I'm guessing S3). Probably not the fastest at 3D, but I don't need 3D in this situation, and moderately accelerated 2D is better than a non-accelerated supervidel. For the price, soldering a readily available, well-tested, graphics controller to the board is way more cost effective than any custom chip solution. Chips sold by the millions for multiple vendors are cheaper than custom ones almost always. Soldering it in would suck for a desktop, as would the lack of 3D, but for en embedded device, who cares?
The Desse looks interesting, but I don't know if its fast enough to be of much use if your main processor is already so much faster. Is there enough of an existing code base of DSP software to warrant compatibility with the 56001? There are also DSP translators to port the 56001 code to newer DSPs. Compared to costs of traditional hardware, its quite pricey. Look at the Echo Mia (the card I use), or the SoundBlaster Audigy line, both 24 bit, 96Khz (or 192Khz) and professional ADCs and DACs (at least for the MIA, with balanced in and out). Again, use something they make already in quantity, solder it in, no custom chips, and no replacements. If you need more inputs, there are USB solutions.
The final and most major difference is they seem to be targetting a desktop, and I don't see enough of a market, and they want to design all this without commercial support. I definately think a commercial product is necessary to support the design costs and prototype expenses and marketing expenses. The design needs of a desktop is very different than a portable device.
Now, if I ever go further than just tossing around ideas for possible feedback, I'd certainly work with the ACP for basic porting of the OS to the coldfire, although I'd want to make quite a few more changes than they would, particularly to the audio APIs. I'd also need someone to help with the $930 they want for the v4e development board for testing :P Long way before I need one of those though. Emulators can go a long way.
|
