cassiel@well.sf.ca.us (Paul Theodoropoulos) (05/03/91)
At the back of AmigaWorld's April '91 issue, in the section devoted to rumors, speculation, etc., there was an intrigueing idea proposed: Designing an 040 board that would allow the 030/882 to continue functioning, essentially as coprocessors to the 040. They stated that all that would be needed to implement this would be to "tweak" the system software so that it would recognize both chips were aboard. A wonderful idea! though the 030/882 combination would be grossly slower than the 040, it seems to me that on our multitasking system, they could be used to manage the more mundane system tasks, while the real "gut-busters" would simply be sliced to the 040 exclusively. Unfortunately, as i understand it the Fast Slot on the A3000 motherboard is a bus master slot - it disables the 030. Would a crafty hardware firm be able to create a workaround? is it feasible? is it worth the trouble? - it seems a waste to throw out the 7.5 mips the A3000 pumps out...27+ Mips would be screaming.... just speculating out loud, i guess....
psteffn@pogo.gnu.ai.mit.edu (Paul Steffen) (05/03/91)
"Tweaking" the system software is quite an understatment if you were modifying the operating system to run on 2 cpu's. The closest you could probrably get to utilizing two cpu's is to have AmigaDOS loaded on both and use some kind of hardware interface [perhaps dual ported RAM like Bridgeboard uses] and a PIPE: type device. It would be running like two separate Amigas connected via something like ParNet/SerNet.
daveh@cbmvax.commodore.com (Dave Haynie) (05/03/91)
In article <24561@well.sf.ca.us> cassiel@well.sf.ca.us (Paul Theodoropoulos) writes: > At the back of AmigaWorld's April '91 issue, in the section devoted >to rumors, speculation, etc., there was an intrigueing idea proposed: >Designing an 040 board that would allow the 030/882 to continue functioning, >essentially as coprocessors to the 040. They stated that all that would be >needed to implement this would be to "tweak" the system software so that >it would recognize both chips were aboard. That will certainly work, though implying software overhead is no more than a "tweak" grossly oversimplifies the problem. My guess is that, for the first such systems anyone does this way, AmigaOS will run on the 68040, while the 68030 will be used as some kind of I/O processor or something, not actual somewhat-symmetric multiprocessing. >Unfortunately, as i understand it the Fast Slot on the A3000 motherboard >is a bus master slot - it disables the 030. Yes and no. The only way to allow a Coprocessor device to access motherboard resources is for it to master the motherboard bus. You can't have both processors running a cycle at the same time on the bus, of course. This hardly means that the 68030 is disabled, it just gets kicked off the bus when the coprocessor device wants the bus. The 68040, for example, is normally off its processor bus, requesting bus access when it wants to use the bus. This meshes quite well with the 68030's habit of acting as a default bus master. Any time the coprocessor device is running from cache or any private memory it has, the 68030 can run free. This is very typical of tightly coupled multiprocessor systems. -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "That's me in the corner, that's me in the spotlight" -R.E.M.
Fletcher@cup.portal.com (fletcher sullivan segall) (05/04/91)
> > At the back of AmigaWorld's April '91 issue, in the section devoted >to rumors, speculation, etc., there was an intrigueing idea proposed: >Designing an 040 board that would allow the 030/882 to continue functioning, >essentially as coprocessors to the 040. They stated that all that would be >needed to implement this would be to "tweak" the system software so that >it would recognize both chips were aboard. Sounds like an interesting idea. Unfortunately most structure locking is done with Disable/Enable, or Forbid/Permit rather than with system semaphores, so the whole system would frequently revert to mono-processing. But I doubt that the Rom/Kernel will be modified to run multi-processing. No doubt OS/9 will step forward quickly to fill the gap to some degree, and I imagine a variety of hacks will come out to allow certain system functions to run on the '030 as a coprocessor. This might include blitter emulation, or various device support routines. > >A wonderful idea! though the 030/882 combination would be grossly slower >than the 040, it seems to me that on our multitasking system, they could be >used to manage the more mundane system tasks, while the real "gut-busters" >would simply be sliced to the 040 exclusively. > >Unfortunately, as i understand it the Fast Slot on the A3000 motherboard >is a bus master slot - it disables the 030. Your information is incorrect. The coprocessor slot on the A3000 fully supports using both processors simultaneously (this is one of the changes from the A2000). > >Would a crafty hardware firm be able to create a workaround? is it feasible? >is it worth the trouble? - it seems a waste to throw out the 7.5 mips the >A3000 pumps out...27+ Mips would be screaming.... > I think the 68030 at 25MHz is only about 4.25 (V2,1) MIPS. -ss Fletcher@cup.portal.com
cy0q+@andrew.cmu.edu (Chad O. Yoshikawa) (05/04/91)
How about(in 2.xx of AmiDos) adding a switch feature that would allow the user to specify, or automically make the active screen, a screen that would be run on the '040 w/ its application. Then the background applications would use the '030/'882 as their processors? Is this possible? Chad
cs171wbu@sdcc5.ucsd.edu (Dave) (05/04/91)
It is probably not practical to expect both an 030 and an 040 to use the same ram (on the motherboard) which is how the 040 acceleartor slot on the 3000 is designed to wor. If you wanted to have both an 040 and an 030, you would want the 040 put in one of the 32 bit slots on the 3000, with its own private ram, acting as a coprocessor for the 030 (which could ofload processes to the 040 to run in the background). An 040 accelerator for the 3000, which uses the 040 slot and has 64bit ram will hopefully be available soon. An 040 is 3-5 times faster than an 030 at the same clock speed (for integer operataions) and even faster for floating point. Commodore needs to push this accelerator development - You can currently get a next computer with an 040 and a 200 meg hard drive for $4000 (educational prices), monochrome monitor only, but its high res. Commodore should also develop a standard 8 and 32 bit color graphics card (with many megs of graphics memory) and hopefully a graphics coprocessor like i860. If they don't they will start loosing multimedia marketshare. You can get a nextcube with a 32bit graphics card, i860, stand video in/out, video compression chips, etc. and a very high res color monitor (not to mention a 68040, hard drive, (and optical drive too I think)) for about $12000. The video toaster can hardly compete. Image running color x-windows and/or advanced amiga dos windows with a 68040 accelerator and a 32 bit color card and high res color monitor (at least 1200 X 800) !!!! The next can compress video in real time and store up to 1 hour of video on a hard drive. The amiga 3000 is a good platform, being a full 32 bit machine ( except for the old graphics/sound chips). They need to develop the 040 card and 32 bit color card and push them hard to keep up with next and other companies. (And hopefully these will be cheaper when they become available for the amiga than they are on competitors computer systems). Write Commodore/Amiga and tell them that we want 32 bit color!! And we want it as a standard card which amiga will release with future versions of the 3000 or 4000(maybe 3500Graphics) or something, so that software producers can write software specifically for it (the other 24 bit color cards aren't standard for the Amiga line and we won't see wide software support). Amiga has to take command and make the transition from the old 12 bit HAM to true full 32 bit color! This is necessary for a multimedia machine. Hey, that Amiga 3000 deal sounds like a real good way to upgrade your performance. I have a 3000 (which I got through the educational purchase plan) and I like it alot. Some good peripherials are citizen gsx140 24 pin printer (around $300), and practical modem 9600sa (9600 v.32 for around $480) - SHOP IN COMPUTER EDGE MAGAZINE! Some software crashes on the 3000, namely alot of games, but they are continually releasing upgrades to software which will run on the 3000. I looked at a demo for Amiga unix (full blown system 5 release IV), at warners, and it looked nice. Need a large hard drive though and you need to have rom chips installed which boot the unix system. Hopefully, we will eventually all be running unix with amigados running on top of unix, mac dos running on top of unix, xwindows running on top of unix, etc. - the true multienvironment machine. Another note, Apple stock fell $7.00. Ha Ha Ha... They sell their machines for such inflated prices that they deserve to loose market share. The IIfx is a 40 mhz 68030 at some inflated price like $6000 (+ costs for monitor, keyboard, etc.) For that price you can almost get a nextcube with an 040 and i860 risc processor which will run 10 times faster and has 32 bit graphics and lots of slots, and excellent sound, and runs unix/mach, has eithernet, and will probably be able to emulate the whole mac line in the near future. Who are they trying to fool? And those "low cost" color macs are pretty pitiful as well - they are stripped down in computing power (some aren't even sloted for a math coprocessor), expandability, and the color isn't even that good.
panther@st1.vuw.ac.nz (05/06/91)
In article <15543@life.ai.mit.edu>, psteffn@pogo.gnu.ai.mit.edu (Paul Steffen) writes: > "Tweaking" the system software is quite an understatment if you > were modifying the operating system to run on 2 cpu's. The closest > you could probrably get to utilizing two cpu's is to have AmigaDOS > loaded on both and use some kind of hardware interface [perhaps dual ported > RAM like Bridgeboard uses] and a PIPE: type device. It would be running > like two separate Amigas connected via something like ParNet/SerNet. Well, how about this as a varient. Since we now have our famous 'workbench in a window', it _may_ be possible to patch the DOS 2.0 to kick up two workbench windows, one attached to each processor and a segment of memory. This would be an interesting option. The other would be to create a 'virtual network' within the machine similar to the one suggested above, and use a technique that is used on DECnetted VAXes, and has been used on Acorn BBCs to meld the processing power. Either way, you have to do a LOT of work-either write a (probably fairly extensive) OS patch, or a fairly complex hunk of network simulation software... Hmmmm... Much Luck Rodger Donaldson panther@st3.vuw.ac.nz Run Disclaimer [CLI SHell2] Ahh...forget it....
daveh@cbmvax.commodore.com (Dave Haynie) (05/07/91)
In article <Uc8Tm5y00WBL44zmlp@andrew.cmu.edu> cy0q+@andrew.cmu.edu (Chad O. Yoshikawa) writes: >How about(in 2.xx of AmiDos) adding a switch feature that would allow the >user to specify, or automically make the active screen, a screen that >would be run on the '040 w/ its application. Then the background >applications would use the '030/'882 as their processors? Is this >possible? If you solve the "big" problem, processor allocation is easy. And this is a big problem. The operating system, at present, doesn't know about multiple processors. The solutions it offers to deal with synchronization problems when dealing with multiple tasks are insufficient for handling multiple processors. If you really want AmigaOS running on multiple CPUs, lots of changes will have to go in. If those changes were there, scheduling tasks on different CPUs could be done in a variety of ways, though I doubt manual scheduling would be all that efficient in most cases. >Chad -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "That's me in the corner, that's me in the spotlight" -R.E.M.
daveh@cbmvax.commodore.com (Dave Haynie) (05/07/91)
In article <18991@sdcc6.ucsd.edu> cs171wbu@sdcc5.ucsd.edu (Dave) writes: >You can get a nextcube with a 32bit graphics card, i860, stand video in/out, >video compression chips, etc. and a very high res color monitor (not to >mention a 68040, hard drive, (and optical drive too I think)) for about >$12000. The video toaster can hardly compete. You sound a little confused here. First of all, this "NextDimension" board, while it can (according to the spec sheet) display normal video in a small window, is not a device for real video work. The Video Toaster is a totally different creature, designed specifically for video work. Saying it can't compete is like saying Coke can't compete with Michelob; sure, they're both drinks (or video display devices), but they are for different things. Also, the C-Cubed compression chip originally slated to be on this NeXT card has reportedly been deleted from the specifications. This card is supposed to sell for around $4000. Incidently, the Toaster's display kicks out full NTSC overscan at something like 1500x960 pixels. >And we want it as a standard card which amiga will release with >future versions of the 3000 or 4000(maybe 3500Graphics) or something, >so that software producers can write software specifically for >it (the other 24 bit color cards aren't standard for the Amiga >line and we won't see wide software support). That's also the wrong approach. You can't get away with making a $4000 extra a standard thing on a $3000 computer. There are already a number of decent 24-bit options for the Amiga, some even reasonably priced. What would make them all a little easier to deal with would be retargetable graphics, where each application program can interrogate the graphics system, find out what graphics capabilities are available, and act accordingly, without any need to know exactly what hardware is in place. In time, that will happen. -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "That's me in the corner, that's me in the spotlight" -R.E.M.
daveh@cbmvax.commodore.com (Dave Haynie) (05/07/91)
In article <41966@cup.portal.com> Fletcher@cup.portal.com (fletcher sullivan segall) writes: >>Unfortunately, as i understand it the Fast Slot on the A3000 motherboard >>is a bus master slot - it disables the 030. >Your information is incorrect. The coprocessor slot on the A3000 fully >supports using both processors simultaneously (this is one of the changes >from the A2000). Neither system permanently disables the corresponding motherboard processor. Both the A2000 and the A3000 permit a coprocessor device to get on and off the motherboard bus, allowing the motherboard processor to stay active, thus the name "coprocessor" slot. Now, neither motherboard bus is dual ported or anything weird like that, they have to share it just as other bus masters share the bus, but such sharing is very possible. The bus arbitration time on the A3000 coprocessor slot is much less than that of the A2000 coprocessor slot. And the 68030 style of bus locking is much easier to deal with than the 68000 kind, which would have been required in any A2000 card that attempted to keep the 68000 active. >>Would a crafty hardware firm be able to create a workaround? is it feasible? >>is it worth the trouble? - it seems a waste to throw out the 7.5 mips the >>A3000 pumps out...27+ Mips would be screaming.... >I think the 68030 at 25MHz is only about 4.25 (V2,1) MIPS. MIPS can generally mean whatever you want it to mean, I think the term was originally coined by Humpty Dumpty. If you're thinking of DHRYSTONE 2.1, the 25MHz A3000 usually generates in the neighborhood of 7,000-8,000, depending on compiler and various optimizations selected. >Fletcher@cup.portal.com -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "That's me in the corner, that's me in the spotlight" -R.E.M.
es1@cunixb.cc.columbia.edu (Ethan Solomita) (05/07/91)
In article <21302@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Haynie) writes: >What would make >them all a little easier to deal with would be retargetable graphics, where >each application program can interrogate the graphics system, find out what >graphics capabilities are available, and act accordingly, without any need to >know exactly what hardware is in place. In time, that will happen. > ------------------------- Aahhhh! Now we've GOT him! >-- >Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" > {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy > "That's me in the corner, that's me in the spotlight" -R.E.M. -- Ethan "Brain! Brain! What is Brain?"
cassiel@well.sf.ca.us (Paul Theodoropoulos) (05/07/91)
Fletcher@cup.portal.com (fletcher sullivan segall) writes: >I think the 68030 at 25MHz is only about 4.25 (V2,1) MIPS. My 25Mhz A3000 runs at 7.537697 MIPS, at least according to the public domain program called "mips". of course, MIPS is really YAMB - "yet another meaningless benchmark". but it has some value, principally entertainment! ;^) paul theodoropoulos cassiel@well.sf.ca.us
daveh@cbmvax.commodore.com (Dave Haynie) (05/08/91)
In article <1991May7.011720.26702@cunixf.cc.columbia.edu> es1@cunixb.cc.columbia.edu (Ethan Solomita) writes: >In article <21302@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Haynie) writes: >>What would make >>them all a little easier to deal with would be retargetable graphics, where >>each application program can interrogate the graphics system, find out what >>graphics capabilities are available, and act accordingly, without any need to >>know exactly what hardware is in place. In time, that will happen. >> ------------------------- > Aahhhh! Now we've GOT him! Well, without going into too much detail, I can safely say this will happen because it's a requirement of the Amiga market, and becoming more so as more frame buffer type devices appear. Since most software companies working on Amiga stuff are not interested in pursuing the IBM-PC route and writing specific drivers for each board out there, some kind of retargetable graphics system is a virtual certainty. Ideally, Commodore sets the standard, since that way, it comes with every Amiga and can be more integrated into the OS. But if Commodore doesn't do it, I expect the third parties will ultimately agree upon something of their own design. There are a number of groups in the Amiga industry working on "collective" projects. While this is certainly larger than the ones I have heard of, it's not at all beyond the scope of this kind of project. -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "That's me in the corner, that's me in the spotlight" -R.E.M.
mark@calvin..westford.ccur.com (Mark Thompson) (05/11/91)
In article <21302@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Haynie) writes: >Incidently, the Toaster's display kicks out full NTSC overscan at something >like 1500x960 pixels. Actually, it outputs 768 x 482 which is the maximum resolution NTSC can handle (anything greater would have to be thrown away). The 1500 x 960 numbers that Dave is mentioning refers to the resolution that Lightwave 1.0 renders to in antialias mode. This resolution is then filtered down to 768 x 482. While the perceived resolution is much higher, the actual output is still within NTSC limitations. >First of all, this "NextDimension" board, >while it can (according to the spec sheet) display normal video in a small >window, is not a device for real video work. The NextDimension does have some appeal to it but as Dave said, it is not a broadcast video device like the Toaster and would have some real big problems competing in that market. %~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~% % ` ' Mark Thompson CONCURRENT COMPUTER % % --==* RADIANT *==-- mark@westford.ccur.com Principal Graphics % % ' Image ` ...!uunet!masscomp!mark Hardware Architect % % Productions (508)392-2480 (603)424-1829 & General Nuisance % % % ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
dtiberio@eeserv1.ic.sunysb.edu (David Tiberio) (05/11/91)
In article <21302@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Haynie) writes: >In article <18991@sdcc6.ucsd.edu> cs171wbu@sdcc5.ucsd.edu (Dave) writes: >>And we want it as a standard card which amiga will release with >>future versions of the 3000 or 4000(maybe 3500Graphics) or something, >>so that software producers can write software specifically for >>it (the other 24 bit color cards aren't standard for the Amiga >>line and we won't see wide software support). > >That's also the wrong approach. You can't get away with making a $4000 extra >a standard thing on a $3000 computer. There are already a number of decent >24-bit options for the Amiga, some even reasonably priced. What would make >them all a little easier to deal with would be retargetable graphics, where >each application program can interrogate the graphics system, find out what >graphics capabilities are available, and act accordingly, without any need to >know exactly what hardware is in place. In time, that will happen. Maybe Commodore should get together with some of the 24 bit companies, and they could hack up a motherboard modification for a high-end Amiga, or possibly a UNIX machine. >-- >Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" > {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy > "That's me in the corner, that's me in the spotlight" -R.E.M. -- David Tiberio SUNY Stony Brook 2-3481 AMIGA DDD-MEN "If you think that we're here for the money, we could live without it. But the world isn't too good here, and it wasn't always like that." Un ragazzo di Casalbordino, Italia.
daveh@cbmvax.commodore.com (Dave Haynie) (05/14/91)
In article <62063@masscomp.westford.ccur.com> mark@calvin.westford.ccur.com (Mark Thompson) writes: >In article <21302@cbmvax.commodore.com> daveh@cbmvax.commodore.com (Dave Haynie) writes: >>Incidently, the Toaster's display kicks out full NTSC overscan at something >>like 1500x960 pixels. >Actually, it outputs 768 x 482 which is the maximum resolution NTSC can >handle (anything greater would have to be thrown away). Well, I'm no NTSC wiz, but technically, at least on the horizontal, NTSC is analog. There are no such things as pixels, per se. An NTSC TV can't resolve individual pixels at 1500 or so across, but it can resolve the differences between pixel changes at higher-than-70ns rates (you get 70ns pixels out of a computer generating 768 across in full overscan). For example, fire up your favorite titling program in hires, draw some simple letters, and note the jaggies. Then kick that into "superhires" mode (if you have an ECS machine), draw something that looks the same, and note the relative absense of jaggies. You'll see the same effect on TV. That's the reason that cable TV and local commercials done on the Amiga have jaggies in their titling, and the titles on Wide World of Sports or similar network shows don't have any jaggies. Real high end machines may give you 2000 pixels across. You can't resolve individual pixels at that rate, but you can resolve changes. As for the Toaster, I don't know what it really kicks out. Some of my video buddies mentioned 1500x960, and I figured it was at least spitting out a real 1500 pixels on the horizontal. If not, you'll be able to do better looking titles on an ECS system, at least until the Toaster kicks in some antialiasing (my guess is that 768 dots with full NTSC color palette and good antialiasing would make cleaner looking letters than 1500 with three colors). >% ` ' Mark Thompson CONCURRENT COMPUTER % -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "That's me in the corner, that's me in the spotlight" -R.E.M.