Amiga One X 1000 and Xena/Xorro

[Dandy]

@Dandy

This is "classic" Amiga forum so most users here see it from "classic" view of point.

Hmmm - here at my end it tells me:
» Amiga.org » Amiga computer related discussion » Amiga Hardware Issues and discussion » Amiga One X 1000 and Xena/Xorro

No mention of "classic" - sorry to disappoint you...

[OlafS3]

Hmmm - here at my end it tells me:
» Amiga.org » Amiga computer related discussion » Amiga Hardware Issues and discussion » Amiga One X 1000 and Xena/Xorro

No mention of "classic" - sorry to disappoint you...

Please stop!

[KimmoK]

So you can load code in it? Can you use the official tools (f.e. on windows) to create it? ...

From the thread so far: you can load code to xena from AOS4 command line.
But "compiler" runs only on windows.
There's a "study" of putting up a server to handle that compilation for all x1000 etc. users.
(another option might be emulator to run windows tools on AOS4, but so far no luck)


(And to my understanding there's a study if it is possible to implement a system that enables swapping the running program/code per thread, without needing to restart the whole chip or something. That would be pretty cool. Perhaps unique even.)
(& IIRC, the chip is XS1-L2 https://www.xmos.com/download/public/XS1-L02A-QF124-Datasheet%28X1466E%29.pdf http://www.a-eon.com/18-10-2013-5.pdf)

And because I stumbled on that x1000 PDF... why it lists only 8GB RAM supported? I thought PA6T can address 32GB of RAM?

[OlafS3]

From the thread so far: you can load code to xena from AOS4 command line.
But "compiler" runs only on windows.
There's a "study" of putting up a server to handle that compilation for all x1000 etc. users.
(another option might be emulator to run windows tools on AOS4, but so far no luck)

For the second you would need either something like VMWare or WINE (and I am not sure if it would work with WINE at all). More realistic (personal guess) is to either create a kind of chain, develop and compile on Windows and then automatically transfer it to amiga or develop it on Windows and just finally test it on Amiga.

[KimmoK]

@xena warrior

Is xena able to generate interrupt to PA6T?
Is PA6T able to generate interrupt to xena?
What is the max bandwidth between PA6T and xena?
Is it possible to do DMA transfer to PA6T RAM from xena?


I think I lost the link to PA6T-xena-xorro schematic, is it somewhere as free?


Then, just playing with ideas...
Work for all those cores/threads:
1: Control.
2: Data in. (CPU interface)
3: Data out. (CPU interface)
4: D9-1-io
5: D9-2-io
6: D9-3-io
7: D9-4-io
8: Floppy-control (kryoflux on xcore)
9: Floppy-data
10: i2s-out  / SPDIF
11: i2s-in
12: Fan-Control
13: funky-led-control
14: watchdog (PA6T might have wd built in, but I'm not sure if one can use it on AOS4)
15: parallel-port-out or DA/DIO out or RPi I/O emu or arduino I/O emu. ...
16: parallel-port-in or AD/DAQ in ...

I wonder ... once such SW is loaded to xcore, would it be simple to just configure I/O on the fly (by control process) and then set needed threads/cores to work...

guestions galore:
-is it possible to have xena running when PA6T is off/sleep/halted?

out of curiosity, old xena thread @ xmoslinkers: http://archive.xmoslinkers.org/forum/viewtopic.php?t=700

[smerf]

@TrevorDick,

Wouldn't it be better to bring out a couple of boards for an affordable price to bring people back to the platform and to start providing support. The Amiga would of never made it at such a huge price, plus Commodore started support by a mag showing how to program the hardware and all the new commands needed to give programmers support on their product. Next thing you need to do is get a software company to start making software for it (if you can find one that microslick hasn't already over powered by their software usage EULA, giving them the right to program microsick programs) anyhow just sitting on a computer and bringing it out is just part of the program, so now we have a new computer with an OS, and some programs. Not very attractive or interesting at this point of time, especially for the price.

I believe the reason they gave up on the Amiga was the designers and engineers gave little room for expansion especially in the graphics and sound area, and what was made for expansion was later on to expensive or obsoleted the main boards that came out. In other words everything was specialized which later became the death bell and road block for future development.

[smerf]

No, this is an Amiga forum that is open and welcome to all flavors of the Amiga experience.  It is one of, if not THE oldest Amiga forum site, so it began before many of the newer Amiga flavors existed, but that does not mean that it is ONLY for Classic Amiga users.



The majority of users in the Amiga community are "Classic" Amiga owners/users, so it is no surprise that a "Classic" Amiga point of view is expressed more often here.  The new owner(s) want to make this site a welcoming place for all users of any Amiga or Amiga inspired system.

Please keep this in mind when posting to these forums.

@AmigaDave,

Very True, It isn't just for Classic Amiga users, they have let Mac users slowly take over this board, it has become more Mac orientated with their Amiga Emulation programs (which I don't really blame Mac users, since the Amiga even with how old it is, is a way better platform to use and way more fun)

Just goes to show they will let anyone on this board, (even me, LOL) but then they banned one of the most entertaining, and knowledgeable users named FRANKO, who is a pure Amiga user, who even I admire, since I am mostly a classic Amiga user, who slowly moved over the the PC platform for more modern day games and graphic and sound improvements. I know Bill Gates took the Amiga OS, examined it and slowly moved its ideas and principles to what the PC is today, if you ever saw his office, he had an Amiga 1000 sitting on his back shelf so he wouldn't lose direction of which way to take his OS and the PC, which was new hardware, new graphics, new expansion slots, and the undoing of old exhausted time ridden hardware. While Amiga stayed on its course with old PPC technology, and expansion devices, which threw them out of the main stream high end computers. We saw it coming but didn't believe it could happen, no Commodore was to ahead of all the other platforms, yeah right!!!

smerf

[spirantho]

@TrevorDick,

Wouldn't it be better to bring out a couple of boards for an affordable price to bring people back to the platform and to start providing support. The Amiga would of never made it at such a huge price, plus Commodore started support by a mag showing how to program the hardware and all the new commands needed to give programmers support on their product.

The Amiga in real terms was actually more expensive when it started than the X1000. It was when the A500 came out that Commodore started to become popular, but that's why we have the Sam range (although it does seem like there's not so many for sale at the moment).
The problem is that "affordable price" varies from person to person. Many people want an Amiga-compatible board that's custom designed with drivers made for just a few hundred dollars each, and that's just impossible. The Amiga market has nothing like the economies of scale that the PC market does, nor does it even have those that the Amiga did originally. The Amiga 500 was in a massive market; that market has shrunk tremendously since those days, so while Commodore could invest and build many thousands of units to recoup costs, that's impossible these days because demand is so much lower.

Next thing you need to do is get a software company to start making software for it (if you can find one that microslick hasn't already over powered by their software usage EULA, giving them the right to program microsick programs) anyhow just sitting on a computer and bringing it out is just part of the program, so now we have a new computer with an OS, and some programs. Not very attractive or interesting at this point of time, especially for the price.

AmigaKit and A-Eon are trying to do just this, but the main problem here is that people's expectations have moved. In the 1980's, a good program could be very simple (e.g. Kindwords, Deluxe Paint) Now people expect applications of the level of complexity of Microsoft Word and Photoshop before they're taken seriously. We're never going to have the massive coding teams that PC and console software makers have, there's no point in trying. What company would invest millions of dollars in development costs if they're only going to sell a thousand copies max.?

I believe the reason they gave up on the Amiga was the designers and engineers gave little room for expansion especially in the graphics and sound area, and what was made for expansion was later on to expensive or obsoleted the main boards that came out. In other words everything was specialized which later became the death bell and road block for future development.

I think the real reason they gave up on the Amiga was simply money. It became less commercially viable in about 1993, when the PC clones took over; those who continued to develop for the Amiga were pushed more and more to the sidelines until they too went to the "Dark Side".

This is a topic for another thread, though. Let's not derail this one about Xena etc.

[LyleHaze]

Interesting stuff and many thanks for your work on it

So you can load code in it? Can you use the official tools (f.e. on windows) to create it? Then you would still need a kind of chain to transfer it from PC to X1000/X5000 to test it there. Or propably more convenient develop and test the code on PC and then transfer it to Amiga and test it there. What XMOS chip is used? Perhaps better to use the same chip and configuration on PC for development and testing and then transfer it and test it on Amiga?

Yes, Yes,Yes,Yes, XS1-L16A.
The X1000 _had_ an XS1-L02, but since then the salesmen have taken liberties with how things are named. The exact same chip is now the XS1-L16A.

The older includes are now deprecated, and they have since renamed threads as cores, cores are now tiles, and a bunch of other changes that seem to be designed just to make conversation awkward. i.e. since "core" had a completely different meaning than it does now, any time it is used the speaker must stop to explain all this renaming mess.

I have _no_ XMOS hardware of any type on my windows machine, I just develop and sim there, and bring the executable over to the X1000 when it's ready for testing on metal.

But every thing you asked is "yes" and has been since October 2012 when I did the public demonstration at AmiWest. I often transfer the executable by FTP, but a flash drive works fine too.

[LyleHaze]

@xena warrior

Is xena able to generate interrupt to PA6T?
Is PA6T able to generate interrupt to xena?
What is the max bandwidth between PA6T and xena?
Is it possible to do DMA transfer to PA6T RAM from xena?

1>Yes, but I have not yet implemented it.

2>No, there's no such thing as a Xena interrupt pin.  However, one thread(core) of core(tile) One is dedicated to localbus communications, and is addressable as ..16k?.. of flat space. not sure about the size, I'd have to look it up. So there's no need to interrupt it, just perform a read or write to any address that is part of Xenas address range. Note: The range all maps to Xena Space, NOT to RAM.. that space is all comms, not storage.

3> I've forgotten what max bandwidth is on the X1000, it moved around a lot while developing the interface. The interface has been enhanced on Cyrus, but I have not yet had a chance to clock it.

4> No, all Xena memory is internal to the chip, and the only way in is through I/O of some sort.. JTAG is for loading code or debugging, localbus is normally for communication between AmigaOS and Xena tasks..but the I/O is so damned fast that the best way to get high bandwidth in and out is by using chip I/O. The Xena chip is practically all I/O, most can be organized into single bits or groups of 4/8/16 bit groups.
For instance, a simple (no handshaking) serial port needs only two single bits. ANY two single bits of I/O, and if you want only TX or RX then onlly a single IO pin is needed. So you can basically have whatever/as many as you'd like.. the code to make a serial port is just a few lines long, and the speed can go into the megabit range easily.
The SDCard interface uses a couple (3?) individual control bits, plus a 4 bit bus if you choose the wide/fast SDCard interface.

I think I lost the link to PA6T-xena-xorro schematic, is it somewhere as free?

It is part of the X1000 technical reference manual. That question is best answered by A-EON.

Then, just playing with ideas...
Work for all those cores/threads:
1: Control.
2: Data in. (CPU interface)
3: Data out. (CPU interface)
4: D9-1-io
5: D9-2-io
6: D9-3-io
7: D9-4-io
8: Floppy-control (kryoflux on xcore)
9: Floppy-data
10: i2s-out  / SPDIF
11: i2s-in
12: Fan-Control
13: funky-led-control
14: watchdog (PA6T might have wd built in, but I'm not sure if one can use it on AOS4)
15: parallel-port-out or DA/DIO out or RPi I/O emu or arduino I/O emu. ...
16: parallel-port-in or AD/DAQ in ...

I wonder ... once such SW is loaded to xcore, would it be simple to just configure I/O on the fly (by control process) and then set needed threads/cores to work...

The CPU interface is already implemented. much the rest depends on implementation. while all of it at once might be too much for a single Xena chip, I doubt you'd need all of it at once.. This is where "software defined silicon" comes in.. you just tell it what you want it to be at any given time, and it becomes whatever the code says.
Or you could add more Xena chips, and then extend a link from core1(tile1) so they can all access the localbus interface if needed. But see my previous comments on isolation.

guestions galore:
-is it possible to have xena running when PA6T is off/sleep/halted?

A hard reset of the Amiga also resets Xena. A soft reset of the Amiga has no effect on Xena at all. Of course localbus communication requires both sides to be up.

out of curiosity, old xena thread @ xmoslinkers: http://archive.xmoslinkers.org/forum/viewtopic.php?t=700

Interesting read, I had not seen that one earlier.
Nice to read a bit from the other side of the fence.

[KimmoK]

Thank you for that info! (again)

...
3> I've forgotten what max bandwidth is on the X1000, it moved around a lot while developing the interface. The interface has been enhanced on Cyrus, but I have not yet had a chance to clock it.
....

At some point PA6T-xena-xorro schematic was shown in a-eon gallery, from that & xcore specs I figured out that PA6T-xena bus bandwidth is 30Mbytes/s or less. That is plenty for control and some data. (for larger needs one can use the PCIe inline with xena I/O connector)


btw. nasty thing how they stirred and mixed the chip&thread&core namings, sometimes as an ordinary engineer that annoys greatly (It's not enough having to all the time learn new things, because of those PR dudes one needs to re-learn also some of the old)

...a thought about xena->CPU interrupt ... very nice that interrupt can be generated, it would be nice to do interrupt driven things rather than the M$ USB polling methology, there's better use for CPUs than polling interfaces.

[KimmoK]

about dev board...

Surely the person affording to buy x1000 affords to buy the xorro devboard as well.

But IF xena&xorro wold exist on cheaper motherboard, I wonder if normal empty PCIex4 (/x8)= prototyping board could be used as well to get I/O out of xena...

(after quick check... did not spot cheaper than amikit board ... surely somewhere...)

[WolfToTheMoon]

Question to A-eon...

While adding (co)proccessors to the Aone systems that are not used by almost anyone, why not develop something useful like a remake of the old x86 bridgeboards Amigas had. Something with an Atom or AMD Beema on a PCIe card, that is plently fast and passively cooled...

[KimmoK]

...remake of the old x86 bridgeboards ...

I think there exist suitable x86-on-PCIe -cards. "Only" the SW needs to be made. The SW work needed might be too much vs getting cheap x86 node and connect to it via RDesktop...

+ If a-eon x86-on-PCIe custom HW would be made, getting M$ OS support for it costs millions I believe ... let's not enter there!



UPDATE: silly stuff, MAME & xcore
http://www.xcore.com/forum/viewtopic.php?f=21&t=2449&p=15308&hilit=joystick&sid=68204a68fee6dc45d080fc053e3d8fed#p15308


UPDATE: if someone does not know what xena/xcore is...
https://www.xmos.com/download/public/xCORE-Architecture%28X9650D%29.pdf

[LiveForIt]

I think there exist suitable x86-on-PCIe -cards. "Only" the SW needs to be made. The SW work needed might be too much vs getting cheap x86 node and connect to it via RDesktop...

+ If a-eon x86-on-PCIe custom HW would be made, getting M$ OS support for it costs millions I believe ... let's not enter there!

I agree Rdesktop and VNC solve that easily, besides there exists towers, that makes it easy to fit two full size motherboards into one tower.

[LyleHaze]

Thank you for that info! (again)

...a thought about xena->CPU interrupt ... very nice that interrupt can be generated, it would be nice to do interrupt driven things rather than the M$ USB polling methology, there's better use for CPUs than polling interfaces.

It is there, and surely will get used. So far I have really had no need.
The "unique" properties of links on an XMOS chip create certain problems.. Serialized data can flow extremely fast, but will stall for any read without write or write without read.  Ther is no "read until there is no more", as the last read would hang the reader indefinitely.

This _should_ be easy enough to work around, but if your solution is not 100%, then your Xena will hang indefinitely.

As a result, my first projects use a very simple structure where all reads and writes are 16 bits, no matter what, and every transaction between the localbus and the Xena chip is broken down into individual "write then read" loops. This helps keep things simple, but not as fast as they could be at all.

Also a result of this is that I have had no need for interrupts at all yet.

I do have one-directional streaming for higher bandwidth data paths.. but I'm probably gertting into more detail than casual conversation requires.

The line is there, and connected to a GPIO pin on the PA6T. One day when I or someone else needs an interrupt, then progress will be made on this front.
One other user has already inquired about using this line for something else, and I have discoraged that, obviously because once this becomes an interrupt, spurious events for random reasons will not be good for the system.

There is a big list of "stuff to do".. my current project is not a perfect example of properly structured code.. It's more proof of concept than anything else.

There are other users at XCore who are working on a system that will dynamically allocate threads(cores) and also control the switching fabric to allocate their necessary resources, all on the fly. This is obviously better than our current system, which limits us to pre-compiled complete projects.

One possible use that has not yet been mentioned, there is code available for both SHA2 and AES encryption. I'm no crypto guy but having that code available might be useful. I don't know if the X1000 has a guaranteed unique serial number anywhere, but combining that with the crypto stuff could be useful for secure confirmation of machine I.D... and it could be done with no Xorro card or hardware at all.. just what the machine ships with already.

Hell the Xena chip may have a unique ID in itself. that would certainly streamline things. I'll have to look into that.

Nobody loves copy protection, but if it can be done in a way that does not interfere with performance for the registered user, then I don't mind it either.

I need to go.. Nice chatting with you!

LyleHaze