rudolpe@jacobs.CS.ORST.EDU (Eric Hans Rudolph) (05/15/91)
I am one of the biggest amiga fans, you can bet on it. While trying to think of ways to speed up some games I am developing, and thinking of Dave Haynie's protests about idiots messing with hardware registers and thinking of Dave saying they HAVE to keep the amiga line compatible, I've decided to design my own computer, from scratch, no joke. I'm a CmpE here at OSU and I bet I have the qualifications. I was looking at the inside of the A500 board here on my desk and decided it's a weenie. I mean, it's the best on the block, but it's not good ENOUGH. Agnus is a killer chip. Absolutely amazing, but I know I and some ohter knowledgable sources put together could do better. The rest of the chip set, the custom ones are nice, and innovative, but slow and out of date. So, I have some plans in mind to put together a screamer computer. That's the name. Screamer. And this computer will be dedicated to the Hardware/Software hack. I've most the plans put together already, including the custom Co-Processor (laid out almost to the bone) and the hardware stipulations. Anyone want to become famous and band together with me? Let's make something amazing. Not for the money, but because the market needs it badly!!!! mail me at rudolpe@jacobs.cs.orst.edu :-)
rang@cz2.ics.uci.edu (Roger Penaranda Jr. Ang) (05/17/91)
In article <1991May15.100140.19240@lynx.CS.ORST.EDU> rudolpe@jacobs.CS.ORST.EDU (Eric Hans Rudolph) writes: >So, I have some plans in mind to put together a >screamer computer. That's the name. Screamer. And this computer will be >dedicated to the Hardware/Software hack. I've most the plans put together >already, including the custom Co-Processor (laid out almost to the bone) and >the hardware stipulations. Anyone want to become famous and band together >with me? Let's make something amazing. Not for the money, but because >the market needs it badly!!!! > >mail me at rudolpe@jacobs.cs.orst.edu > >:-) Whoa, you're talking developing a custom chip, are you? Good luck!! If I'm not mistaken it'll cost you *lots* of $$$, as in tens of thousands (correct me if I'm wrong). You do have your chip already designed to the layout level,right???? Also, remember there are two things that really give you performance (as in what the user perceive): the architecture and the OS. So, if I understand you correctly, you're going to build a computer from scratch (including OS) with some custom chips? Well if you get a handful of people, you should be ready to market in, oohh, about 5 years (at least!!! :-). But what's that??? You're machine may not look so spiff compared to what might be available by then? Oh well, that's the computer biz. Roger P. Ang (rang@ICS.UCI.EDU) "Hey, Dyna-Pink. Nice tattoo." Grad student at the Mr. Wonderful - DynaMan Dept. of Info. & Comp. Sci. Univ. of California, Irvine.
ckp@grebyn.com (Checkpoint Technologies) (05/17/91)
In article <1991May15.100140.19240@lynx.CS.ORST.EDU> rudolpe@jacobs.CS.ORST.EDU (Eric Hans Rudolph) writes: >I am one of the biggest amiga fans, you can bet on it. While trying to think >of ways to speed up some games I am developing, and thinking of Dave Haynie's >protests about idiots messing with hardware registers and thinking of Dave >saying they HAVE to keep the amiga line compatible, I've decided to design >my own computer, from scratch, no joke. I'm a CmpE here at OSU and I bet I >have the qualifications. I was looking at the inside of the A500 board here >on my desk and decided it's a weenie. I mean, it's the best on the block, but >it's not good ENOUGH. Agnus is a killer chip. Absolutely amazing, but I know >I and some ohter knowledgable sources put together could do better. >The rest of the chip set, the custom ones are nice, and innovative, but >slow and out of date. So, I have some plans in mind to put together a >screamer computer. That's the name. Screamer. And this computer will be >dedicated to the Hardware/Software hack. I've most the plans put together >already, including the custom Co-Processor (laid out almost to the bone) and >the hardware stipulations. Anyone want to become famous and band together >with me? Let's make something amazing. Not for the money, but because >the market needs it badly!!!! Here are some things to consider: -- You don't want to be stuck with a graphics architecture in custom chips on the motherboard, like the Amiga currently is. Your graphics should be designed to be upgradeable. Put it on a daughtercard, or in a bus slot. -- You don't want your applications to become dependent upon the specifics of the graphics architecture, or else you'll never be able to rev it. Therefore you must provide your applications with an abstract view of the graphics device which is not too limiting, which is extremely fast for games and animations, and which allows for upgrades to future graphics capabilities. If you choose a poor abstraction, then developers will put their efforts into circumventing it. -- You don't want this graphics interface abstraction to be too closely tied to a particular hardware architecture. For example, the Amiga must be used with random-access RAM, because the alignment restrictions of VRAM are too limiting. However, VRAM offers a far greater bandwidth to raster memory. -- To prevent, I mean really prohibit, programmers from touching the hardware in ways that make future revs impossible, you must also provide a hardware abstraction of the graphics device. Something like TIGA, where the interface is thru a skinny pair of host-interface registers to a smart graphcis co-processor. (Developers love to do illegal things. You have to make it truly impossible.) -- If the OS uses MMU tables and supervisor/user protection to prevent applications from handling the interface to the graphics device directly, then a hardware abstraction to the graphics device might be unneccesary. But it had better be really developer-proof. -- TIGA can be downloaded with custom support code for applications; you probably want to support this because it offers a tremendous performance boost. However, TIGA is tied to TI's own graphics chips, but you'll want to be able to incorporate the latest and fastest RISC chips in your graphics engine if the need arises. So you should adopt an abstraction for graphics co-processor code which will be independent of the actual machine interpretation. Perhaps a pseudo-code (like QuickDraw) which would be interpreted, or a source language (PostScript) which would be compiled when downloaded into the graphics device. -- Surely your operating system will provide suport for everything everyone here has asked for: low memory requirements, real-time response (a requirement for multimedia), virtual memory, resource tracking. Try OS-9, I think it'll do. There's my first two cents. Don't get the idea I'm kidding or trying to discourage you; I'd buy a machine like this, especially if it cost close to what an Amiga does. However, I'm aware these are expensive things to do. -- Richard Krehbiel, private citizen ckp@grebyn.com (Who needs a fancy .signature?)