Abbuc Hardware Contest 2008...

[+CharlieChaplin]

Well,

simply click here: http://www.abbuc.de/modules.php?name=DocTr...Id=260#englisch

 

-Andreas Koch.

Edited January 31, 2008 by CharlieChaplin

[+Larry]

Well,

simply click here: http://www.abbuc.de/modules.php?name=DocTr...Id=260#englisch

 

-Andreas Koch.

 

Perhaps this will be the year of a working, salable Atari accelerator! (I can always hope.)

-Larry

[+bob1200xl]

Geee... 500 bucks is almost enough to make it worth our while, isn't it? How fast would it have to go, Larry? What would it have to do?

 

Bob

 

 

 

 

 

Well,

simply click here: http://www.abbuc.de/modules.php?name=DocTr...Id=260#englisch

 

-Andreas Koch.

 

Perhaps this will be the year of a working, salable Atari accelerator! (I can always hope.)

-Larry

[ndary]

Geee... 500 bucks is almost enough to make it worth our while, isn't it? How fast would it have to go, Larry? What would it have to do?

 

Bob

 

 

 

 

How fast we can go? do we need to replace the OS?

 

A plug-in CPU accelerator upgrade that will also add extea memory (512k or more) will be a good candidate

 

Ndary

[+Larry]

Geee... 500 bucks is almost enough to make it worth our while, isn't it? How fast would it have to go, Larry? What would it have to do?

 

Bob

 

How fast we can go? do we need to replace the OS?

 

A plug-in CPU accelerator upgrade that will also add extea memory (512k or more) will be a good candidate

 

Ndary

 

Hi Bob, Nir-

 

Here's what I know; what I think; what I think I know...

 

Not to be picky, but ABBUC's prize is 500 Euro's -- almost $750!

 

The accelerator that is most likely to be released is along the lines of the (so-called) Warp-4. From what I know, it is a update of the Turbo-816, but unlike the T816 will run at about 4X in extended memory. As mentioned before in another thread, this scheme has been done before on an Atari work-alike, but the details were never released (AFAIK).

 

There may be as many as 3 different folks working on this concept. I have heard recently that the issues with extended memory may be close to being solved. (?)

 

I don't know quite what this would do as regards Graphics. There are a lot of gamers, so I suspect they would want the accelerator to be able to do most (all?) graphics modes. I'd be thrilled with text modes.

 

4X would be great. Warp-7 (a different project -- unknown.) There is supposedly another concept for using an "Atari on a chip" (can't remember the name of that device).

 

As Nir says, drop in would be great. Individual "tuning" to a 'scope is probably not what is needed. PAL and NTSC, of course.

 

One school of thought is that the accelerator would have been finished last year, but didn't want to compete with the SIO2USB project. (Speculation, but a reasonable supposition.)

 

Perhaps drac030 could comment on the OS and Basic?

 

I personally always liked the concept that was proposed some years ago in Atari Classics.

 

And BTW, I'd be a very willing beta tester...

 

-Larry

[+bob1200xl]

Geee... 500 bucks is almost enough to make it worth our while, isn't it? How fast would it have to go, Larry? What would it have to do?

 

Bob

 

How fast we can go? do we need to replace the OS?

 

A plug-in CPU accelerator upgrade that will also add extea memory (512k or more) will be a good candidate

 

Ndary

 

Hi Bob, Nir-

 

Here's what I know; what I think; what I think I know...

 

Not to be picky, but ABBUC's prize is 500 Euro's -- almost $750!

 

The accelerator that is most likely to be released is along the lines of the (so-called) Warp-4. From what I know, it is a update of the Turbo-816, but unlike the T816 will run at about 4X in extended memory. As mentioned before in another thread, this scheme has been done before on an Atari work-alike, but the details were never released (AFAIK).

 

There may be as many as 3 different folks working on this concept. I have heard recently that the issues with extended memory may be close to being solved. (?)

 

I don't know quite what this would do as regards Graphics. There are a lot of gamers, so I suspect they would want the accelerator to be able to do most (all?) graphics modes. I'd be thrilled with text modes.

 

4X would be great. Warp-7 (a different project -- unknown.) There is supposedly another concept for using an "Atari on a chip" (can't remember the name of that device).

 

As Nir says, drop in would be great. Individual "tuning" to a 'scope is probably not what is needed. PAL and NTSC, of course.

 

One school of thought is that the accelerator would have been finished last year, but didn't want to compete with the SIO2USB project. (Speculation, but a reasonable supposition.)

 

Perhaps drac030 could comment on the OS and Basic?

 

I personally always liked the concept that was proposed some years ago in Atari Classics.

 

And BTW, I'd be a very willing beta tester...

 

-Larry

 

What was in Atari Classics?

 

I don't think you should have to change the OS or software. Programs that rely on clock timings will be trashed, but WSYNC and that kind of thing should be OK. Running out of ROM (cartridge or OS) will be slow unless we can move it into RAM. We should build that into the device. 14mhz will be tough unless we strip out a lot of features. Probably best to use a 14mhz auxillary processor if we want that kind of speed.

 

How about a 2x for starters? We could probably run that out of flash memory.

 

All the graphics modes would work, just the CPU is faster. I like the idea of an aux processor because it doesn't mess up existing programs. But, you have to code in order to pump up the performance.

 

(the extra $250 is for pencils...)

 

 

Bob

[Defender II]

Hmmmmm......

 

ATARI upgrades .....

 

Hmmmmm.......

[+Larry]

What was in Atari Classics?

 

I don't think you should have to change the OS or software. Programs that rely on clock timings will be trashed, but WSYNC and that kind of thing should be OK. Running out of ROM (cartridge or OS) will be slow unless we can move it into RAM. We should build that into the device. 14mhz will be tough unless we strip out a lot of features. Probably best to use a 14mhz auxillary processor if we want that kind of speed.

 

How about a 2x for starters? We could probably run that out of flash memory.

 

All the graphics modes would work, just the CPU is faster. I like the idea of an aux processor because it doesn't mess up existing programs. But, you have to code in order to pump up the performance.

 

(the extra $250 is for pencils...)

 

Bob

 

 

The very interesting "65816" articles -- Sep/Oct, 1994, p.34-36 & Jan/Feb, 1995, p.27-29.

I could attach a scan if I got permission from the publisher...

 

2x and mostly compatible would be a *terrific* start. Could we "flip a switch" or such for 1X/compatibility with a dual clock?

 

-Larry

Edited February 6, 2008 by Larry

[Rybags]

A no-frills slave processor in a cartridge, needing no modification to the host machine would be a winner in my books.

[drac030]

On cart it would not work, I guess. The entire goal is to have the new and the old memory directly adressable for programs, and this means that the CPU has to be replaced.

 

The OS has to be replaced too, the old XL OS does not allow to switch the CPU to the mode in which it is able to run code in the extended memory (past the first 64k).

[Rybags]

I was thinking more along the lines of a PIC with it's own (or shared) memory.

 

It wouldn't be anything like existing or proposed turbo/16 bit 6502 type projects.

[+bob1200xl]

I'm not aware of any special requirement for accessing data in the extended blocks. It all seems like it would work in 8-bit mode.

 

Regardless, this is a speed upgrade, not necessarily a memory extension. The memory needs to be replaced for higher speed access and it is just as easy to put 512K in there as only 64K.

 

Bob

 

 

 

 

 

 

On cart it would not work, I guess. The entire goal is to have the new and the old memory directly adressable for programs, and this means that the CPU has to be replaced.

 

The OS has to be replaced too, the old XL OS does not allow to switch the CPU to the mode in which it is able to run code in the extended memory (past the first 64k).

[+bob1200xl]

Cool... scan away!

 

 

Yes, of course. You can always drop into 'good old Atari mode'.

 

4x looks just as doable as 2x, I guess. Works out to 4.5x the possible clock cycles. It's mostly a matter of the PCB. Handwired will be 2x. Commercial production should go 4.5x on the same design.

 

Bob

 

 

 

What was in Atari Classics?

 

I don't think you should have to change the OS or software. Programs that rely on clock timings will be trashed, but WSYNC and that kind of thing should be OK. Running out of ROM (cartridge or OS) will be slow unless we can move it into RAM. We should build that into the device. 14mhz will be tough unless we strip out a lot of features. Probably best to use a 14mhz auxillary processor if we want that kind of speed.

 

How about a 2x for starters? We could probably run that out of flash memory.

 

All the graphics modes would work, just the CPU is faster. I like the idea of an aux processor because it doesn't mess up existing programs. But, you have to code in order to pump up the performance.

 

(the extra $250 is for pencils...)

 

Bob

 

 

The very interesting "65816" articles -- Sep/Oct, 1994, p.34-36 & Jan/Feb, 1995, p.27-29.

I could attach a scan if I got permission from the publisher...

 

2x and mostly compatible would be a *terrific* start. Could we "flip a switch" or such for 1X/compatibility with a dual clock?

 

-Larry

[+bob1200xl]

That's actually a good idea - a pass-thru cart like the R-Time8 with a really fast aux processor. I prefer a 65816 because it can run the code that we are using in the 8-bit. One application would be the FP routines in the OS. Only a few parameters to pass and a few results to hand back, right? The cart could present it's memory to the 8-bit in 16K chunks so you could load code and even pass really big arrays to and from it. Biggest drawback is that the cart has no access to the 8-bit's memory itself, only the other way around. (so, you run everything in the cart - the 8-bit just does screen and I/O) (you could point ANTIC at the cartridge space and do page-flipping...) (really fast) (a little 512MB CF card sticking out of the top of the cart...)

 

A 400 with a 14mhz, 512K, 512MB aux cartridge!

 

Heck, you might as well drop the cartridge code in there as well. Why stack carts? The cart takes over the machine on boot and you're off to the races.

 

I'd better quit. Good idea.

 

Bob

 

 

I was thinking more along the lines of a PIC with it's own (or shared) memory.

 

It wouldn't be anything like existing or proposed turbo/16 bit 6502 type projects.

[drac030]

I'm not aware of any special requirement for accessing data in the extended blocks.

 

I am not speaking about accessing data. I am speaking of running code.

 

Regardless, this is a speed upgrade, not necessarily a memory extension.

 

The only speedup you gain in Warp 4 comes from running code in the high memory.

[+Larry]

Cool... scan away!

(snip...)

 

Here are the two 65816 articles as .tif scans.

 

 

[ndary]

A 400 with a 14mhz, 512K, 512MB aux cartridge! sound great..

 

do we need different carts for 800XL/XE and the old A400/800 systems?

with flash memory the cart can also act as a flash drive... am i right?

 

Ndary

[+bob1200xl]

This type of upgrade would run code in a 64K block of its own memory, although you could run it in 16-bit mode if you wanted - it would just be difficult to develop the routines without a system platform that ran 16-bit code. (maybe an Apple IIGS?). There are no interrupts to deal with, no HALTs, no REF. Everything would run at high speed, but, in 8-bit mode, the only use for the extra memeory is data storage.

 

What does the Warp 4 configuration look like?

 

Bob

 

 

 

 

I'm not aware of any special requirement for accessing data in the extended blocks.

 

I am not speaking about accessing data. I am speaking of running code.

 

Regardless, this is a speed upgrade, not necessarily a memory extension.

 

The only speedup you gain in Warp 4 comes from running code in the high memory.

[+bob1200xl]

No, no reason to make different carts for the different systems. All of the OSs allow the cart to take over very early in the boot. That's all we need.

 

The CF card would be a flash drive for the cartridge processor. Only indirectly would it be available to the Atari. (I guess it would work OK on the Atari, too)

 

The problem with this project is that it needs a whole bunch of code to make it work. It would run legacy stuff perfectly because you could always just switch it off. But, it's not going to do anything without somebody programming it. An internal h/w hack would run faster with NO code changes.

 

It's a trade-off. Which is better?

 

Bob

 

 

 

A 400 with a 14mhz, 512K, 512MB aux cartridge! sound great..

 

do we need different carts for 800XL/XE and the old A400/800 systems?

with flash memory the cart can also act as a flash drive... am i right?

 

Ndary

[+bob1200xl]

Oh, yeah... that's what I'm talking about. Thanks.

 

So - you've been following this discussion? How would you suggest we proceed? A simple 2x hack? Do you think that would be enough to win?

 

Bob

 

 

 

Cool... scan away!

(snip...)

 

Here are the two 65816 articles as .tif scans.

 

 

[+Larry]

Oh, yeah... that's what I'm talking about. Thanks.

 

So - you've been following this discussion? How would you suggest we proceed? A simple 2x hack? Do you think that would be enough to win?

 

Bob

 

I guess you are asking me?

 

If 2X is pretty straight-froward, I'd go for that as opposed to something that might get bogged down. Probably there are things that will be learned from the 2X that can be built upon later. Maybe the "MarkII" is 4.5X the next year. You can see from the emulator that 200% makes a very significant performance improvement.

 

"Drop In." Fewer connecting wires is definitely better. Works in XL's and XE's. If it is flexible, compatible, doable, installable, I think it might be a winner.

 

Now if it happens that the 4X gets done this year, and its 2X vs 4X (ceteris paribus), then the 4X will probably win. But the work has been going on for some time on the 4X and it's not there yet, AFAIK.

 

I'd go with simple -- I like "sure things." A "bird in the hand," etc. After you have mulled this over a bit, why don't you jot down a list of reasonable capabilities -- sort of a spec list.

 

FWIW,

Larry

Edited February 7, 2008 by Larry

[drac030]

This type of upgrade would run code in a 64K block of its own memory, although you could run it in 16-bit mode if you wanted - it would just be difficult to develop the routines without a system platform that ran 16-bit code. (maybe an Apple IIGS?). There are no interrupts to deal with, no HALTs, no REF. Everything would run at high speed, but, in 8-bit mode, the only use for the extra memeory is data storage.

 

I don't understand what you are talking about. 65C816 by itself is not faster than 6502. It only can run at higher clock speed, and can do 16-bit calculations and moves with single instructions. To use the 16-bit instructions normally (with interrupts enabled) you have to replace the OS.

 

What does the Warp 4 configuration look like?

 

It is an ordinary Atari with the CPU replaced and some (1 MB for example) of additional, directly addressable fast RAM located past the first 64k.

[+bob1200xl]

Rybags was suggesting a cartridge based auxillary processor that would take over some of the Atari's workload. Such a device would run at a much higher clock speed than the original system, perhaps 8 times faster. With no interrupts or refresh or HALTs from ANTIC, the throughput would approach 12 times the Atari in GR.0. The advantage to using a 65816 is that code running in the Atari could be exported to the cartridge with few or no changes required. The FP calculations, for example. If you were running a word processor, the pages could be kept in cartridge memory and manipulated by the aux processor, based on commands from the main program running in the Atari.

 

No, the 65816 isn't any faster than a 6502, unless you increase the clock speed. That's the plan - higher clock speeds.

 

The nice thing about the cartridge is that it can take over the OS very early in the boot. It can then configure the system as it wishes. And, you don't have to open the covers.

 

Does the Warp 4 run at higher clock speeds?

 

Bob

 

 

 

 

 

This type of upgrade would run code in a 64K block of its own memory, although you could run it in 16-bit mode if you wanted - it would just be difficult to develop the routines without a system platform that ran 16-bit code. (maybe an Apple IIGS?). There are no interrupts to deal with, no HALTs, no REF. Everything would run at high speed, but, in 8-bit mode, the only use for the extra memeory is data storage.

 

I don't understand what you are talking about. 65C816 by itself is not faster than 6502. It only can run at higher clock speed, and can do 16-bit calculations and moves with single instructions. To use the 16-bit instructions normally (with interrupts enabled) you have to replace the OS.

 

What does the Warp 4 configuration look like?

 

It is an ordinary Atari with the CPU replaced and some (1 MB for example) of additional, directly addressable fast RAM located past the first 64k.

[drac030]

Rybags was suggesting a cartridge based auxillary processor that would take over some of the Atari's workload. Such a device would run at a much higher clock speed than the original system, perhaps 8 times faster. With no interrupts or refresh or HALTs from ANTIC, the throughput would approach 12 times the Atari in GR.0. The advantage to using a 65816 is that code running in the Atari could be exported to the cartridge with few or no changes required. The FP calculations, for example. If you were running a word processor, the pages could be kept in cartridge memory and manipulated by the aux processor, based on commands from the main program running in the Atari.

 

Ok, I understand now what are you referring to. Surely, it would speedup FP calculations and such, but in fact FP calculations are rarely used. I guess that all the gain would be wasted on communications with the cartridge and its coprocessor. You know, the calculations even the BASIC interpreter does are mostly ordinary integer stuff, most of the speed gain you can do with an accelerator comes from the fact that every CPU instruction is executed faster - not from the fact that just FP calculations are faster.

 

The nice thing about the cartridge is that it can take over the OS very early in the boot. It can then configure the system as it wishes.

 

I can't imagine how it could configure the system f.e. to use the native mode interrupts, in the other way than replacing the XL OS with something loaded into the RAM below the OS. This is "replacing the OS" I mentioned above and it is way better done by replacing the ROM. There are already some 32-in-one OS-es, so what is the problem with adding one more, this time really necessary?

 

Does the Warp 4 run at higher clock speeds?

 

Yes. The high RAM (past the $FFFF) is clocked at 7 MHz currently, maybe 14 MHz in the near future.

[Shawn Jefferson]

Would you be able to use this "cartridge" processor to do math operations, or other data transformations ? Sort of like the Suzy chip works in the Lynx. You can load the multiplication or division registers, tell it to go, and then check back a bit later for your answer. I can imagine that some programs and games could benefit from even this type of speedup. A 3D ray tracer for instance may be able to utilize this to make a 25/30 fps 1st person shooter in a graphics mode that didn't make you think you were playing on a Vic20.

 

I think if you build it, and then, especially, write support into the various emulators, programmers may start using it to actually build something cool. It's not even that far out of the range of things that were done "in the day". Wasn't there a cartridge that came with a game that off-loaded some of the calculations?