neil@amiga.UUCP (Neil Katin) (10/18/85)
Amiga is now on the net. We have been reading the postings for the last week or so, and should be able to send outgoing news now. I'm the one who posted the note to net.news.newsite, and I've started getting mail about "I want to be a developer..." or "I already am a developer, and had a question..." or "I got an amiga and it has this bug..." (sigh). Unfortunately, I am not the right person to deal with those sort of questions. I have created a mail account on amiga called amiga!support. If you send your postings here it will get to the right people (is current send the msg to a local newsgroup amiga.support...). Personally, I haven't been on USENET for about two years. I remember in times past people were upset with having a "for profit" support line that went through uucp. Would someone please let me know what the current policy on this is? We are willing to do a direct connection to anyone who has more than an occational question. For those of you who are "real" developers: remember that we already have the Amiga on-line support system. You should have gotten info on this with your developers kit. For the rest of you, we really would like to have your input (and want to fix whatever bugs there are...) but remember that no one can deal with a non-specific bug report. It's hard to know where to start on replying to some of the misinformation that is going around. The one that I REALLY wanted to address is the bouncing ball. The ball really DOES only use about 8% of the CPU time. It is done by playing with the bit-plane pointers and color map every vertical frame. As Bob Pariseau (our first software person, and current VP software) says "there's less here than meets the eye". However it's a legitimate hack, and shows some of the easy magic you can do with the machine. In the same posting, someone was talking about how 70% of the bus bandwidth was used for painting the ball. There is a faint kernel of truth in this, but it is misleading. The amiga's memory is run at twice the speed of the cpu. This bandwidth is divided between the display (bit planes), the blitter, the copper (co-processor), the cpu, and other stuff (disk dma, sprite data, etc). The display has the highest priority. The copper comes next. The blitter and the cpu trade off whatever is left. The "other stuff" only uses a low percentage of bus bandwidth, so doesn't affect this discussion. There is enougth bus bandwidth to support eight low resolution bit planes, or four high res planes. (note that for other reasons the amiga does not support more than six bit planes...). Once you have this many planes displaying, there is only bus bandwidth for the cpu during the horizontal and vertical blanking periods. In "normal" operation, when the workbench screen is being displayed (2 high res bit planes), the 68000 can run at full speed if the blitter is not running. If the blitter is running then they share the available bandwidth (but the blitter is doing its job faster than the 68000 could do the equivalent things...). Now, back to boing (the bouncing ball demo). This is a five bit plane, low res image. The display is using ~62% of the available bus bandwidth. The cpu can therefore run at ~75% of maximum speed, NOT the 30% that previous postings would have you believe. Hope this clears up some of these questions... Neil Katin amiga!neil
wjr@utai.UUCP (William Rucklidge) (10/22/85)
In article <132@amiga.amiga.UUCP> neil@rocky.UUCP (Neil Katin) writes: > There is enougth bus bandwidth to support eight low resolution bit > planes, or four high res planes. (note that for other reasons the > amiga does not support more than six bit planes...). Once you have > this many planes displaying, there is only bus bandwidth for the cpu > during the horizontal and vertical blanking periods. > > In "normal" operation, when the workbench screen is being displayed > (2 high res bit planes), the 68000 can run at full speed if the blitter > is not running. If the blitter is running then they share the available > bandwidth (but the blitter is doing its job faster than the 68000 could > do the equivalent things...). > > Neil Katin > amiga!neil Hum... this raises some questions. From the block diagram I saw in BYTE, it seems that the Amiga has a layered bus structure. Does anyone know if this is in fact the case? This would mean that the processor would only be slowed down/locked out when it was trying to access the lower 512K bank, and could run at full speed if it was using other memory, regardless of what display mode was active or what the blitter was doing... if you had an external RAM expansion you would be able to run the processor at full speed almost all the time, since your programs would be in non- contended memory. This would be a lovely thing to have... full processor speed and lots of bit planes and blitter action. -- William Rucklidge University of Toronto UUCP {ihnp4 utzoo decwrl uw-beaver}!utcsri!utai!wjr CSNet wjr@toronto BITNET wjr at utoronto This message brought to you with the aid of the Poslfit Committee.
eric@caip.RUTGERS.EDU (Eric Lavitsky) (10/24/85)
> From: wjr@utai.UUCP (William Rucklidge) > Newsgroups: net.micro.amiga > Subject: Re: Amiga's listening to you. > Message-ID: <819@utai.UUCP> > Date: 22 Oct 85 18:36:34 GMT > Hum... this raises some questions. From the block diagram I saw in BYTE, > it seems that the Amiga has a layered bus structure. Does anyone know if > this is in fact the case? This would mean that the processor would only > be slowed down/locked out when it was trying to access the lower 512K > bank, and could run at full speed if it was using other memory, regardless > of what display mode was active or what the blitter was doing... if you > had an external RAM expansion you would be able to run the processor at > full speed almost all the time, since your programs would be in non- > contended memory. > > This would be a lovely thing to have... full processor speed and lots > of bit planes and blitter action. > > -- You are correct. The lower 512K of the address space is known as 'chip' memory. The 3 custom VLSI's can only see this memory. All memory added above an beyond this is called 'fast' memory, because the 68000 can access this memory at full speed without worrying about contention from other devices. In fact, if you look at the board design, it strongly reflects this feature of the architecture. Any memory added on the side expansion port is linked directly to the 68000, but not the VLSI's like the internal 512K is. Applications that sit in fast memory should really scream, even when a lot of graphics is going on. Hopefully, Commodore will announce their expansion box soon so we can fully take advantage of the Amiga's address space. The T-Card may be nice, but if I expand, I want to be able to go all the way... Eric -- ARPA: LAVITSKY@RUTGERS UUCP: ...{harvard,seismo,ut-sally,sri-iu,ihnp4}!topaz!eric SNAIL: 16 Oak St., Flr 2 New Brunswick, NJ 08903