New RISC OS target: D2Plug

Previous PlugComputers had one problem: no graphics output, so the only way to get a sensible RISC OS computer would have been graphics output via USB – with all its assorted performance problems.

Now, GlobalScale and Marvell have announced a new PlugComputer model named D2Plug: http://www.plugcomputer.org/development-kits/d2plug.html

Nice things about it:

• graphics up to 1080p@60 Hz
• eSATA connector
• complete with case
• costs less than 200 EUR
• 1 GB DDR3 RAM
• 1000bT Ethernet (not via USB)
• integrated flash
• S/PDIF optical out
• SD card interface

Not so nice things about it:

• single core, 800 MHz

According to various benchmarks, the newer Marvell stuff (ARMv6/v7) is a lot faster than the older Sheeva/Kirkwood offerings (ARMv5) – superscalar core as well as larger L1 and L2 caches.

It also looks like Marvell is now taking a more open approach wrt docs and source code for their chips.

If someone would like to have a (serious) go at porting RISC OS to this platform, I am prepared to sponsor hardware as well as the C/C++ Dev Kit as needed.

Is the process of porting RISC OS well enough documented?

That’s a good specification, despite the comparatively low CPU speed; you can have all the CPU speed in the world, but if the display output is no good (I’m looking at you, BeagleBoard!) it’s not actually a very attractive piece of hardware as a stand-alone computer.

It’s still a pity that the now very old Iyonix hardware can drive a 30" monitor over VGA at 2560×1600×16Mx30Hz, while several years later, newer hardware – cheaper, but still, much newer – can’t do anything like that. If it could, I’d seriously consider replacing my Iyonix to save space (it’s in one of those huge Panther cases).

Is the process of porting RISC OS well enough documented?

There’s always room for improvement, but as long as you (i.e. whoever starts a port) is the kind of programmer that is able to work things out without relying on excessive amounts of hand-holding then I don’t think you’ll run into any show stopping issues.

My how to port RISC OS guide is a good place to start. Steps 4 & 5 (getting your test code into the build system) is something that I/we would be more than willing to do for you, since intimate knowledge of how the build system works is perhaps a bit too much to ask if this would be your first full exposure to it. I’ll probably flesh out that doc a bit more today, since I can think of a few bits to add to it.

Andrew, we have ARMini users running at 1920×1200 successfully. Whilst it isn’t 2560×1600 (you need dual link DVI for that), anything higher means monitors costing circa 1000ukp…

Dave,

the biggest problem with porting RISC OS is that you must be an expert in many areas. The two most important ones are probably an intimate knowledge of how RISC OS works on the lowest level, and of course you have to be an ARM assembler expert. Having a lot of patience helps, too. Having previously worked with lousily-specced hardware at the lowest possible level is also helpful. I guess fiddling with strange USB devices qualifies for this however ;-)

I guess fiddling with strange USB devices qualifies for this however ;-)

;-) indeed!

I wasn’t asking on behalf of myself: I’ve got quite enough other work for RISC OS queued up. It was a naive question, but on that would need answering for anyone considering a RISC OS port.

The ARMADA 510 here is also used in the Wyse T50 and CompuLab SBC-A510.

[Edit: Meanwhile, the Armada 310 is on this CoreExpress module.]
[Edit2: Also used in the CuBox.]