Supercomputer

Only for programmers and BIOS gurus with technical questions.
Post Reply
azcromntic
New visitors - please read the rules.
Posts: 3
Joined: Sun Nov 16, 2003 9:56 am

I have a couple of old computers lying around that I infrequently use to test 'this' or 'that'. I am very interested in using these computers to make a supercomputer. My idea is to use the floppy drive "ports" on each mother boad to connect the two motherboards together for a "communication" channel.

I had two other older computers (66Mhz and 133Mhz) connected in this manner and surprisingly they would both boot and recognize the floppy controller (from an old 5" floppy drive, guts taken out). The ribbon cable went from one floppy connection on the motherboard to the floppy controller (middle connector on cable) then to the other motherboard.

There were other complexities that kept me from moving forward with the experiment and I gave it up. Until I realized that those problems might be solved from the BIOS.

I would rather do away with the floppy as a "pass through" device and just have the ribbon cable from one motherboard to the other. Then, in the BIOS, to verify that it was indeed working, I could write some kind of communication protocol to identify each motherboard to the other. Basically output to the terminal a message.

For instance: Extend the BIOS of computer A and B to include some communication verification code. Once the normal BIOS code finished and the new extended code is entered, A sends a "message" to B using the floppy "port" (I think it is called an interrupt). B recognizes the message, processes it (prints it to the screen) then sends a reply. A recognizes the reply, processes it (screen again) and the test is done.

This test would do one of two things. If it didn't work (or you know it can't be done) it would help me to put closure on this crazy idea. OR, if it did work (possible with 'xyz' adjustments) it would motivate me to continue and spend endless hours and countless dollars working to improve it.

I have Assembly language knowledge and can write software. I have some training in electronics but not in depth enough to move forward (as you might be able to tell already). I am just looking for anyone willing to offer input, ideas and suggestions.

Thanks!

Andrew
maman
Master Flasher
Posts: 173
Joined: Sun Mar 31, 2002 2:08 pm
Location: Taka Bonerate National Park, Indonesia
Contact:

azcromntic wrote:....

I would rather do away with the floppy as a "pass through" device and just have the ribbon cable from one motherboard to the other. Then, in the BIOS, to verify that it was indeed working, I could write some kind of communication protocol to identify each motherboard to the other. Basically output to the terminal a message.

For instance: Extend the BIOS of computer A and B to include some communication verification code. Once the normal BIOS code finished and the new extended code is entered, A sends a "message" to B using the floppy "port" (I think it is called an interrupt). B recognizes the message, processes it (prints it to the screen) then sends a reply. A recognizes the reply, processes it (screen again) and the test is done. ...

Andrew
I think what are you looking for is an expansion/option ROM, since this ROM will be executed after the main BIOS. I've developed an experimental code in this area, and it already worked. I suggest that you proceed with the ISA Option/Expansion ROM type, since looking at your hardware spec, it's probably only available via this approach. You can see an example here, in my reply to Ger@ld question, i.e. http://www.wimsbios.com/phpBB2/viewtopic.php?t=4617.
And for PCI expansion ROM example you can download here:
http://forums.biosmods.com/showthread.p ... eadid=1449
I hope these can help you :D
azcromntic
New visitors - please read the rules.
Posts: 3
Joined: Sun Nov 16, 2003 9:56 am

I will take a look and do some experiments. Thanks for the info!!
NickS
BIOS Bodhisattva
Posts: 3145
Joined: Fri May 03, 2002 10:34 am
Location: Thames Valley, UK

TBH, using the floppy port doesn't seem such a hot idea to me. The FDD interface uses a serial data stream at 500 Kbits/sec. If you have ECP/EPP parallel ports you can probably do better.
Tested patched BIOSes. Untested patched BIOSes.
Emails *will* be ignored unless the subject line starts "Wim's BIOS forum"
azcromntic
New visitors - please read the rules.
Posts: 3
Joined: Sun Nov 16, 2003 9:56 am

I guess for learning purposes it would be much easier to use the Parallel ports, might be less expensive on hardware too, if I screw something up.
Then again I wouldn't be learning the same things, I think? Thanks for the tip.
NickS
BIOS Bodhisattva
Posts: 3145
Joined: Fri May 03, 2002 10:34 am
Location: Thames Valley, UK

Then again I wouldn't be learning the same things, I think?
True....
Tested patched BIOSes. Untested patched BIOSes.
Emails *will* be ignored unless the subject line starts "Wim's BIOS forum"
Ritchie
BIOS Guru
Posts: 761
Joined: Wed Oct 30, 2002 5:17 am

I am no expert in this area but since high speed interconnectivity options are available such as ethernet, usb and firewire it may be better to connect them using one of these methods and then see what software is available that may possibly be able to drive 2 or more machines toghether as some kind of virtual, single, supercomputer.

I have no idea what software may be around for that kind of purpose but would think that it would be at least theoretically possible. And probably the more you can get into the core code of the system - somewhere at a basic system layer, the better. An application layer would limit the single virtual supercomputer to that application.

Add-on/Edit

I have just realised that using a high speed interface also makes sense because to merge the hardware functions of two or more machines you have to get as close to the bus speeds of the main boards in your interface as possible and parallel ports and floppy drive interfaces probably fall way below main board bus speeds.
maman
Master Flasher
Posts: 173
Joined: Sun Mar 31, 2002 2:08 pm
Location: Taka Bonerate National Park, Indonesia
Contact:

Ritchie wrote:..... since high speed interconnectivity options are available such as ethernet.....

--> I have no idea what software may be around for that kind of purpose but would think that it would be at least theoretically possible <-- .....
I would say that it's not only theoritically possible, my "simple kernel", that I've provided the links to download above actually built according to meet the bootp protocol which is used for virtual terminal, you can see more on the source code. The problem is it's supposed to run on PCI based LAN card, which I assume are not available to all the platform that Mr. azcromntic uses to do his experiments :)
Post Reply