RISC OS news and information from Drobe


Beta! \| About us \| Contact \| RSS	Webspace \| Tech docs \| Downloads \| BBC Micro \| Gallery \| Wallpaper

32bit Adjust on ARM9 breakthrough

By Chris Williams. Published: 8th Oct 2004, 15:00:09.

This is not a drill [Updated]

Advantage Six has announced the A9 - an ARM9 powered computer that runs a 32bit version of 'Embedded RISC OS Adjust'. The A9 press release is light on the details, so we're awaiting a fuller specification and expected shipping dates, although we're told that a retail version will be made available at some point.

Earlier this week, RISCOS Ltd. confirmed the development of a 32bit RISC OS Select. Previously, RISC OS 4 could only run natively on ARM processors that provided a 26bit program counter, and as modern CPU cores provide only 32bit program counters, the OS would need to be made 32bit compatible in order to survive.

ARM9 and RISC OS 4

The ARM9 processor family used by the A9 tends to run between 180 and 250MHz, and basically focuses on being high performance embedded cores, rather than desktop brutes. The main point here is that RISC OS together with a hardware partner has found a new processor and solution space to leverage itself onto. Going by their announcement today, we're assuming they want to move into 'industrial control'.

What we do know is that the A9 will be available as a rack mountable version and also, interestingly, in a wall mountable version with a flatscreen and mobile communications capabilities. We also know that Advantage Six, better known as STD to RISC OS users, are using the RISC OS friendly user environment and BASIC language to sell the system to engineers in need of rapid solution development.

As to the difference between Embedded RISC OS and desktop RISC OS, Ad6 explain: "Embedded versions of RISC OS 4 have exactly the same potential features as the full desktop versions but leave the choice of which features are to be included to the manufacturer, dependant on the end application and target cost."

Stop press at 14:51 8/10/2004
Matt Edgar of STD tells us that as for what ARM9 processor they'll use: "We have a range available to us and will, in conjuction with the OEMs decide on the most suitable for their particular requirement. This is why we have a range of processors/boards available. Some of which are very capable. We have identified a number of initial market areas that we are concentrating on but expect to expand rapidly into others as we demonstrate that we can deliver (reference customers are an important part of this). Obviously details beyond what is in the press release are commercially sensitive."

Also, expected OEM prices for "reasonable quantities" are 400 to 1500GBP. Sample units of the machine are available now, and finalised units will ship depending on the OEMs' requirements, we're told.

Update at 19:43 8/10/2004
drobe.co.uk was able to speak to Castle this evening, and we were able to get their opinion of events. "It's fantastic news", said Castle director Jack Lillingston, "we're always keen to see new RISC OS products in the marketplace". Castle reaffirmed their support and enthusiasm for all things RISC OS, noting that they're keen to see as much deployment of RISC OS as possible, as such things ultimately help RISC OS users and Castle themselves. The 32-bit Adjust developments are of course between ROL and STD, and do not directly involve RISC OS 5 technology or Castle.

Speaking on the matter of the relationship between Castle and RISCOS Ltd, Jack spoke highly of the progress being made with Paul Middleton over the negotiations between RISC OS 5 and Select, stating that the agreement was "99% complete", and that we should expect a joint announcement in the next while.

Links
A9 press release
A9 website
RISCOS Ltd

Discussion

Viewing threaded comments | View comments unthreaded, listed by date

Hoping this means we will get some form of portable RISC OS mini-machine coming now?

Rien

Rien on 8/10/04 3:44PM [ Reply | Report ]

Cool two versions of 32bit happy (presumably hardware independent) RISC OS to choose from ;o)

Bit ironic after all the mud that was thrown Castle's way about them doing lots of work for the embedded market during the 'big spat' to find out that ROL are doing exactly the same.

Is this now a sign of ROL's waning commitment to their 'core desktop market' or just the reality-check that anyone developing RISC OS as an OS will have to expand into non-desktop areas to take the OS forwards.

blahsnr on 8/10/04 3:47PM [ Reply | Report ]

Rien I enquired sometime ago about RON, to hear from PM that ROL were considering the netBook Pro after having failed to get a port for the netBook. After a couple of emails PM then said he couldn't comment any further.

Wacky-backy or an NDA? Who knows ;o)

Doubtless the next RISC OS show will be interesting!

blahsnr on 8/10/04 3:58PM [ Reply | Report ]

Maybe the emebdded RISC OS Select IS a coop between castle and RISC OS??

Rien on 8/10/04 4:04PM [ Reply | Report ]

Rien Are you by any chance in a (Dutch) coffee shop at the moment ;o)

blahsnr on 8/10/04 4:08PM [ Reply | Report ]

No, but I am in an optimistic mood... Rien

Rien on 8/10/04 4:12PM [ Reply | Report ]

The NetbookPro has a variety of crazy hardware that unfortunately don't make it a great choice to port RISC OS to.

mrchocky on 8/10/04 4:20PM [ Reply | Report ]

To Rien, I'm not following your logic. I can't see Castle being the main partner with embedded-32 and letting Advantage Six get the first bite at the cherry! Clearly A6 are the partners with ROL on this embedded version and Castle have been - are - will be partners in an Iyonix version. Great news for everybody!

jc on 8/10/04 4:34PM [ Reply | Report ]

So ROL finally did what they set out to do when they started off with ROL - a 32 bit RISC OS. What a waste since a 32 bit RISC OS is there already... or did they build it on top of Castle RISC OS? Or to put it differently: Do we have two different 32 bit RISC OS threads?

Anyhow: Put a battery in the wall mountable A9 with flatscreen and there you might have your Laptop?

hzn on 8/10/04 6:22PM [ Reply | Report ]

What fab news this is. We might even get a more afordable modern computer than either the Omega or Iyonix (wishful thinking).

Can't wait to see the version released for the desktop market.

Hzn, wouldn't that then be classed as a Tablet PC?

sa110 on 8/10/04 6:44PM [ Reply | Report ]

Maybe Riscstation can do that for us..

Rien on 8/10/04 6:44PM [ Reply | Report ]

Affordable? I think I saw UKP 400 - 1500 and I assume for a reasonably fast one you won't be in the lower half of that price range...

Guess why I wrote "laptop" and not notebook (can be put on my lap) :-)

And then there is the issue with the keyboard... since if the wall mountable one does require a keyboard and mouse it's not a tablet PC either.

hzn on 8/10/04 6:47PM [ Reply | Report ]

jc> Now this is purely speculation but (to quote STD as reported on Drobe in an article by Chris Williams [link]). It was stated that (with respect to STD re-opening) that this "may be subject to Castle Technology Ltd fulfilling certain obligations in a timely manner."

Like (perhaps) supplying a variant of the HAL adjusted to work with the STD ARM9 ? Now that's purely speculation but it seems highly unlikely that after years of foot dragging by ROL that suddenly they have 32bit just *after* they come to a settlement with Castle.

But then that's the marvel of speculation it can mean something or nothing at all.......

AMS on 8/10/04 7:02PM [ Reply | Report ]

Hzn, perhaps STD and ROL have trumped everone and gone for a computer that requres neither keyboard or mouse, just the power of the brain.

sa110 on 8/10/04 7:05PM [ Reply | Report ]

So I wonder if this means that Castle will dump their 32-Bit RO and just move the Iyonix-specifics into A HAL for Adjust-32?

Then we can have just one 32-Bit RISC OS, yay! Mind you, even two 32-Bit RO's is probably going to be easier for developers than a 32-Bit and a 26-Bit RO....

simo on 8/10/04 7:08PM [ Reply | Report ]

Based on what the update states it appears that Castle have said none of their RO5 technology was used in the A9, assuming the HAL is counted in in that then an unqualified congrats *are* due then to ROL, well done guys.

Glad to see these events have not dimmed Castle's enthusiasm for RISC OS, after all they've given us (pre the A9) the first new (and currently only) 32bit hardware - and all without selling a single Windows License to a RISC OS user either (nice !).

As to the A9's upper price of 1500 GBP, if that's the OEM price I dread to think what the retail price will be ! Can't wait to see if the same people who complained about the price of Iyonix will do the same when confronted with a slower ARM9 box at a higher price.

Anyway - congrats again ROL nice work guys - keep it up !!!!

Regards

Annraoi

AMS on 8/10/04 7:29PM [ Reply | Report ]

I wonder if Samsung have actually released that 1.2ghz Arm10 they announced a while ago?

The Doctor on 8/10/04 7:58PM [ Reply | Report ]

Sadly not.

AMS on 8/10/04 8:16PM [ Reply | Report ]

would be great to have RISC OS running on it when they finaly get around to releasing it.

sa110 on 8/10/04 8:21PM [ Reply | Report ]

Is the GPS they mention Global Positioning Satelite or something else?

Revin Kevin on 8/10/04 8:40PM [ Reply | Report ]

I may be wrong but I thought that Pace developed a HAL for an Arm9. Could this be resurfacing?

jlavallin on 8/10/04 8:57PM [ Reply | Report ]

So if this is really a RO Ltd 32-bit RO (rather than a rebadged Castle one) does it answer the question of what the Omega XScale will use?

TonyStill on 8/10/04 9:11PM [ Reply | Report ]

Does this mean that ARM9 cards might soon be available with Select-32 ROMs as an upgrade for the RiscPC?

JWCR on 8/10/04 9:42PM [ Reply | Report ]

What would be the point of that? Select and Adjust are already available for RiscPCs as is a 233mhz Kinetic card upgrade. Cheers!

The Doctor on 8/10/04 9:48PM [ Reply | Report ]

Are not ARM9s faster?

AndrewDuffell on 8/10/04 9:56PM [ Reply | Report ]

Bearing in mind that the RPC was designed for the 50odd MHz ARM8, the fact that it's scaled to a 300MHz StrongARM is fairly good going, I'd say. Given that you don't get the best performance out of a StrongARM in a RPC, what's the point in putting a potentially faster ARM9 chip in there instead? Additionally, where's the commercial incentive to do so? RPCs aren't being made any more (due to lack of available parts).

jmb on 8/10/04 10:07PM [ Reply | Report ]

To Rien: you wrote: Maybe RiscStation could do that.

How about finally paying back the deposit????

Ike on 8/10/04 10:45PM [ Reply | Report ]

I was beign sarcastic, given that so many upgrades and modifications have been available for the RPC over the years, that have extended its life way beyond that of contemporary Wintel machines. Should have added a smiley.

JWCR on 9/10/04 1:08AM [ Reply | Report ]

I was being sarcastic, given that so many upgrades and modifications have been available for the RPC over the years, that have extended its life way beyond that of contemporary Wintel machines. Should have added a smiley.

JWCR on 9/10/04 1:08AM [ Reply | Report ]

Andrew> ARM9's come in different flavours. Some as slow as 33MHz, others clocked faster. The one depicted at the top of this page is an Atmel AT92RM200 some of which manage around 200Mips performance (putting them in or around StrongARM performance). ARM's site says that some ARM9's manage 300Mips (around Kinetic performance?). Some ARM9's do have the Vector Floating Point (VFP) fitted and can do 200+MFlops (but I am not aware of any RISC OS variant that supports that).

And then there's the big inponderable which of the many variants of ARM9 have STD chosen - their webpage does not detail the type or performance.

It probably wouldn't make a whole lot of sense fitting an ARM9 to a RPC as (for integer performance) the improvement would - at best - be negligable. If the VFP option were exploited that might be of benefit - but it is not compatible (at the moment) with the FPE (so again unlikely to bring immediate benefit).

It is more the principle that ROL can now provide a RISC OS variant for processors currently unsupported and 32bit *only* that is the significant news and that *does* open the possibility of using xScale and ARM10 (or ARM11's when they become available). Even at that these would be best exploited in new hardware as some of the performance issues of the RISC PC simply *won't* be addressed by just changing the processor there are other bottlenecks too....

AMS on 9/10/04 2:47PM [ Reply | Report ]

Wasn't the A6 based on the Simtec EP7500FE board? Is this then based on the Simtec Samsung 2410 board?

jjvdgeer on 10/10/04 4:10PM [ Reply | Report ]

Ooops... That should be the A75, not the A6....

jjvdgeer on 10/10/04 4:13PM [ Reply | Report ]

Will this Select 32 work on current systems? Since, SA does support 32bit? Or am I just horribly confused.

anon/203.10.121.89 on 11/10/04 02:53AM [ Reply | Report ]

SA does support 32 bit. As do all CPUs from the ARM6 upwards, in fact.

I can't imagine RO Ltd wanting to keep two lines of code (26- and 32-bit) running but I reserve the right to be wrong.

adrianl on 11/10/04 03:49AM [ Reply | Report ]

So will Aemulor run on ARM9? I think it depended on some debug-hooks in [the Iyonix-variant of the] X-Scale processor, so that would mean it probably doesn't?

jjvdgeer on 11/10/04 11:00AM [ Reply | Report ]

I've only looked through the datasheets quickly, but I think the answer is - sadly - no for the StrongARM engine. (Most of the features needed are in fact present, just not the breakpoint registers)

Having said that the JIT - if I ever finish it

- will, and the ARM6 interpreter of course.

adrianl on 11/10/04 7:04PM [ Reply | Report ]

The non-26bit compatibility of the A9 would probably not matter much to OEM's who will probably use it for applications where the ability to run *old* RISC OS software is probably not significant to them. Being able to develope *new* 32bit apps (probably specific to that OEM's intended use for the A9) would probably be sufficient.

If the A9 is eventually released (in some form) as a RISC OS portable or desktop machine then it's going to be even more reliant on purely 32bit software than the Iyonix. Given the amount of 32bit software out there now though this is probably less of a problem than it once was.

AMS on 13/10/04 6:54PM [ Reply | Report ]

Annroi Re 32bit compatibility; don't forget that Castle, developers and the Iyonix buying public have done most of the hard work in getting essential apps and products converted to 32bit clean tested and on the market.

Adjust-32 based products (assuming executable compatibility is maintained) can simply ride on the coat tails of the Iyonix success, without taking the same degree of financial or technical risk.

blahsnr on 14/10/04 06:58AM [ Reply | Report ]

S Williams > I'd not dispute that for one minute !!!!

We're in a position that A9 can *viably* exist on the existing 32 bit software base precisely because of the efforts of Castle, the (often underappreciated) efforts of developers and those who've stuck with the ARM RISC OS platform and supported it through thick and thin.

AMS on 14/10/04 7:46PM [ Reply | Report ]

I've always understood that without a HAL, a complete rewrite was required to enable Risc OS to run on a 32-bit ARM chip. So presumably 32-bit Adjust now has a HAL, and indeed, a different HAL to RO 5, since the ARM9 chip architecture is different. Does this mean that the two branches of the OS are in fact diverging, and indeed are competing with each other for non-desktop market share?

George

bucksboy on 15/10/04 11:22AM [ Reply | Report ]

A HAL (Hardware Abstraction Layer) makes the work needed to run something on different hardware easier, eg loading a file from a floppy disk, scsi disk or ide disk is the same for a programmer because of hardware abstraction.

Nothing to do with what ARM chip is used really, or 26/32bit. The point is to seperate the hardware specific things from the rest of the OS, for ROS5/Iyonix that meant, for example, video hardware, Nvidia instead of VIDC, so first you move all the VIDC specific things into a seperate module, and write a hardware netural interface between the rest of the OS and the graphics module, then you write an Nvidia module and plug that in instead of the VIDC one. More work for 1 change, but when you come to use a different graphics system, much easier. Repeat for all hardware.

mavhc on 15/10/04 3:46PM [ Reply | Report ]

To mavhc: I was just wondering what had been required to move Adjust from its current, Acorn-proprietary chip dependent 26-bit only state to the point where it will run on ARM9 with, presumably, no requirement for VIDC chips or emulation, and whether this work had been done by ROL from scratch or whether the work already done by Castle on RO5 had been the basis for development. No doubt all will be made clear eventually!

George

bucksboy on 15/10/04 6:55PM [ Reply | Report ]

It's very odd, and very annoying, isn't one reason we use RISC OS because we're close to the developers? ROL never say anything any more.

mavhc on 15/10/04 7:55PM [ Reply | Report ]

MavHC: given the climate over the last couple of years, would /you/ publically discuss your development plans? They have received much criticism from the userbase for NOT having a 32-bit version, despite working on this for some time. After all, the entire OS can't exactly be extensively re-written, 32-bitted and changed to work on new hardware in 1 or 2 months! Even the timing between S3 and this announcement is probably too short (remembering A6 are offering release of the A9 to customers now, albeit not the desktop market).

Announcements /were/ made along the lines of 'we're working on something. We'll tell you about it when we can'. It was also announced for some time that 32-bit work was in gradual progress whilst doing Selects 1, 2 and 3.

The problem was that when more specifics are presented publically, companies in RISC OS land (and I'm not just talking ROL here) will be held to account to the nearest day and forever quoted on the subject ad nauseam. This is precisely why Select components are never pre-announced - it's a 'pay for development' scheme. It's also why you never see pre-announcements any more from Castle, STD and other significant parties until the product is commercially available.

In times of a more united userbase, there probably would be more pre-announcements to arouse interest. However, experience has clearly demonstrated this is counterproductive for many companies who now choose to do their development in private.

In answer to George: the article is clearly stating that development was entirely independent of Castle, I would take that at face value.

anon/172.186.183.84 on 16/10/04 00:09AM [ Reply | Report ]

More to do with STD announcing something about Embedded 32 bit RISC OS and noone actually telling us what that really is. Then there's news that it wasn't to do with Castle, so did ROL 32bit seperately from Castle, weren't they working together, etc etc.

mavhc on 17/10/04 11:18PM [ Reply | Report ]

mavhc: The article explains in paragraph five as to what Embedded ROS is. It's a custom build of Select for 3rd parties, where the 3rd party can choose what OS components to include and not to include.

Which explains why ROL have been really modularising the OS since the start of Select.

Chris. Just me.

diomus on 19/10/04 04:52AM [ Reply | Report ]

Please login before posting a comment. Use the form on the right to do so or create a free account.

Create a new account
Forgot your password?

Search this website