| |
Aemulor BeagleBoard |
| |
This is a long thread. Click here to view the threaded list. |
| |
|
Adrian Lees |
Message #112448, posted by adrianl at 00:36, 20/12/2009 |
Member
Posts: 1637
|
Hi, for anyone who still likes 'firsts' screenshots, here's a very early prototype of Aemulor running Publisher on my BeagleBoard... |
| |
 [ Log in to reply ] |
| |
|
Jeffrey Lee |
Message #112449, posted by Phlamethrower at 00:52, 20/12/2009, in reply to message #112448 |
 Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
First!
Is that a WIP copy of Geminus I spot there as well?  |
| |
[ Log in to reply ] |
| |
|
Alan Robertson |
Message #112450, posted by nytrex at 01:04, 20/12/2009, in reply to message #112449 |
Member
Posts: 118
|
Excellent work mate.
And thanks for keeping us informed of your developments. |
| |
[ Log in to reply ] |
| |
|
Adrian Lees |
Message #112451, posted by adrianl at 01:11, 20/12/2009, in reply to message #112449 |
Member
Posts: 1637
|
Aye, but it doesn't get very far at the moment. Only the JPEG rendering is likely to be very useful on the BB currently. I'm sure many of the other accelerations would be useful too, but the code will need quite a bit more work because - for example - in the case of cacheing window contents/sprites, it currently expects the cached data to be in video RAM which is contiguous and outside the OS's control. |
| |
[ Log in to reply ] |
| |
|
Terje Slettebø |
Message #114651, posted by tslettebo at 20:48, 18/6/2010, in reply to message #112448 |
Member
Posts: 3
|
Hi Adrian.
Any progress on this one? I'd liked to be able to run software like John Kortink's Translator, which is supposed to mostly work on the Iyonix with Aemulor.
Would it be possible for you to check if it works with Aemulor on the BeagleBoard? |
| |
[ Log in to reply ] |
| |
|
Trevor Johnson |
Message #116322, posted by trevj at 20:09, 25/1/2011, in reply to message #114651 |
Member
Posts: 660
|
Adrian's website has a link to http://www.aemulor.com/, which is recently registered. I wonder how things are coming on with this. I'd be a paying customer, especially if it'd allow playing old Acorn games on the Pandora. |
| |
[ Log in to reply ] |
| |
|
Jeffrey Lee |
Message #116323, posted by Phlamethrower at 20:26, 25/1/2011, in reply to message #116322 |
Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot Hot stuff
Posts: 15100
|
Adrian's website has a link to http://www.aemulor.com/, which is recently registered. I wouldn't get your hopes up; to me it looks like the domain has simply expired and been snapped up by a squatter.
I wonder how things are coming on with this. Last I heard (which would be just over a year ago) Adrian was disappointed that it didn't look like it was going to be possible to reuse the technique that he'd used to get high-speed emulation working on the Iyonix (due to some of the CPU's debugging registers not being accessible from the CPU itself).
I'd be a paying customer, especially if it'd allow playing old Acorn games on the Pandora. I'd say that a polished/optimised version of ArcEm is a much better thing to pin your hopes on. |
| |
[ Log in to reply ] |
| |
|
Trevor Johnson |
Message #116324, posted by trevj at 21:57, 25/1/2011, in reply to message #116323 |
Member
Posts: 660
|
I wouldn't get your hopes up; to me it looks like the domain has simply expired and been snapped up by a squatter. You're right.
I'd say that a polished/optimised version of ArcEm is a much better thing to pin your hopes on. Well, this very short list isn't alphabetical, so it was probably intended to be chronological... and now turns out to also be in order of decreasing viability.
[Edited by trevj at 22:05, 25/1/2011] |
| |
[ Log in to reply ] |
| |
|
Trevor Johnson |
Message #118517, posted by trevj at 07:54, 2/9/2011, in reply to message #116323 |
Member
Posts: 660
|
"[...] Aemulor in due course." |
| |
[ Log in to reply ] |
| |
|
Herbert zur Nedden |
Message #118518, posted by hzn at 15:32, 3/9/2011, in reply to message #112448 |
Member
Posts: 11
|
Fantastic news. This is the missing thing that keeps me and some others to move to the BeagleBoard or some other new hardware.
Adrian, please make it available! I am happy to put money on the line for this one and others are too for sure.
If you want get in touch: hzn.2011@hq.gag.de
Herbert |
| |
[ Log in to reply ] |
| |
|
Chris Evans |
Message #118543, posted by CJE at 11:10, 5/9/2011, in reply to message #118518 |
CJE Micros chap
Posts: 228
|
Fantastic news. This is the missing thing that keeps me and some others to move to the BeagleBoard or some other new hardware.
Herbert Unfortunately probably not good good news for Beaglebaord/ARMini owners:
AIUI Adrian has said that Aemulor for the Beagleboard/ARMini can't be done due the CPU lacking certain important features (Access to some control registers, I think). The ARM core in the Rasberry Pi I think uses a different ARM core.
May be those features are useable on that core!
n.b. The lack of Aemulor is the main reason we (CJE Micro's) decided no to do a BeagleBoard Bundle. As we want to offer something to replace a RiscPC etc and as R-Comp said in their announcement of the ARMini "However, it is important to be realistic. This machine [THE ARMini] is not a direct replacement for a RiscPC or even Iyonix, and we don't want people ordering/buying with that in mind."
Chris |
| |
[ Log in to reply ] |
| |
|
Andrew Rawnsley |
Message #118545, posted by arawnsley at 11:40, 5/9/2011, in reply to message #118543 |
R-Comp chap
Posts: 600
|
I think it is best not to speculate too much either way on this at this stage. Any comments *must* come from Adrian, and he (understandbly) keeps a low profile. |
| |
[ Log in to reply ] |
| |
|
Trevor Johnson |
Message #118546, posted by trevj at 11:59, 5/9/2011, in reply to message #118543 |
Member
Posts: 660
|
The ARM core in the Rasberry Pi I think uses a different ARM core. Yes, it's an ARM1176JZF-S. |
| |
[ Log in to reply ] |
| |
|
Chris Evans |
Message #118552, posted by CJE at 13:22, 5/9/2011, in reply to message #118545 |
CJE Micros chap
Posts: 228
|
I think it is best not to speculate too much either way on this at this stage. Any comments *must* come from Adrian, and he (understandbly) keeps a low profile. I wouldn't call my main point speculation but a warning. Several people I have spoken to assumed Aemulor was available for the BeagleBoard/ARMini.
You are right in hopeing for a comment from Adrian. I've tried contacting him myself, but the email address I have I suspect isn't valid anymore! |
| |
[ Log in to reply ] |
| |
|
Jeff Doggett |
Message #118563, posted by jeff-doggett at 06:56, 6/9/2011, in reply to message #118552 |
Member
Posts: 21
|
I hate to sound negative but I feel that Aemulor has done more harm than good for the 32 bit market. The fact that it exists means that developers haven't bothered to produce 32bit compatible apps. |
| |
[ Log in to reply ] |
| |
|
Andrew Poole |
Message #118564, posted by andypoole at 07:36, 6/9/2011, in reply to message #118563 |
Posts: 5558
|
I hate to sound negative but I feel that Aemulor has done more harm than good for the 32 bit market. The fact that it exists means that developers haven't bothered to produce 32bit compatible apps. I doubt it. A lot of developers left the RISC OS scene a while back, or aren't interested in updating their apps, so either way you'd still not have as many apps to run on 32bit machines if Aemulor wasn't around. |
| |
[ Log in to reply ] |
| |
|
Trevor Johnson |
Message #118567, posted by trevj at 09:05, 6/9/2011, in reply to message #118564 |
Member
Posts: 660
|
Also worth noting Bounty proposal: Emulator. |
| |
[ Log in to reply ] |
| |
|
Adrian Lees |
Message #118595, posted by adrianl at 09:21, 8/9/2011, in reply to message #118543 |
Member
Posts: 1637
|
I don't want to make any further comments/commitments just yet, but I should correct one point....Aemulor can be made to work on both processors (the screenshot at the top of this thread shows that, and it in fact started life on a StrongARM RiscPC), but it uses features of the XScale hardware to gain a substantial speed increase.
Both the Cortex-A8 (ARMini/BeagleBoard) and the ARM1136 (RaspberryPi) have comparable but different debug hardware. It's worth noting that the A9home version of Aemulor only offers the slower 'ARM610 mode' because there is no suitable debug hardware for it to exploit. |
| |
[ Log in to reply ] |
| |
|
Chris Evans |
Message #118599, posted by CJE at 10:31, 8/9/2011, in reply to message #118595 |
CJE Micros chap
Posts: 228
|
Thanks the update Adrian. |
| |
[ Log in to reply ] |
| |
|
Steffen Huber |
Message #118601, posted by hubersn at 12:05, 8/9/2011, in reply to message #118595 |
Member
Posts: 91
|
I don't want to make any further comments/commitments just yet, but I should correct one point....Aemulor can be made to work on both processors (the screenshot at the top of this thread shows that, and it in fact started life on a StrongARM RiscPC), but it uses features of the XScale hardware to gain a substantial speed increase.
Both the Cortex-A8 (ARMini/BeagleBoard) and the ARM1136 (RaspberryPi) have comparable but different debug hardware. It's worth noting that the A9home version of Aemulor only offers the slower 'ARM610 mode' because there is no suitable debug hardware for it to exploit. Hi Adrian,
I just wanted to reinforce a point that I have made previously.
I know that you are some kind of perfectionist. You are always striving for maximum performance. But please keep in mind that Joe Sixpack (or Otto Normalverbraucher as we Germans say) would be very happy to use a 90%-performance-solution instead of not using anything at all. |
| |
[ Log in to reply ] |
| |
|
Chris Evans |
Message #118609, posted by CJE at 10:26, 9/9/2011, in reply to message #118595 |
CJE Micros chap
Posts: 228
|
... It's worth noting that the A9home version of Aemulor only offers the slower 'ARM610 mode' because... Does that mean Aemulor on the A9home will run software not StrongARM compatible?
Or am I reading too mucj into 'ARM610 mode'? |
| |
[ Log in to reply ] |
| |
|
Adrian Lees |
Message #118617, posted by adrianl at 14:27, 9/9/2011, in reply to message #118609 |
Member
Posts: 1637
|
Does that mean Aemulor on the A9home will run software not StrongARM compatible?
Or am I reading too much into 'ARM610 mode'? It has that name because it implements all of the ARM610's instructions and most closely models that processor; a lot of pre-SA software should work. If it makes heavy use of code modification, however, it's less likely to work, and may require use of the 'ARM3 mode' which is slower still but more capable of coping with such nastiness |
| |
[ Log in to reply ] |
| |
|
Martin Bazley |
Message #118624, posted by swirlythingy at 21:53, 9/9/2011, in reply to message #118617 |
Posts: 460
|
If it makes heavy use of code modification, however, it's less likely to work, and may require use of the 'ARM3 mode' which is slower still but more capable of coping with such nastiness To be fair, the ARM3 processor wasn't exactly known for its speed. If Aemulor runs faster than an ARM3 does (bearing in mind this is on an ARM9, never mind an XScale or OMAP), technically it's already more than fast enough! |
| |
[ Log in to reply ] |
| |
|
Chris Evans |
Message #118647, posted by CJE at 12:26, 12/9/2011, in reply to message #118624 |
CJE Micros chap
Posts: 228
|
If it makes heavy use of code modification, however, it's less likely to work, and may require use of the 'ARM3 mode' which is slower still but more capable of coping with such nastiness To be fair, the ARM3 processor wasn't exactly known for its speed. If Aemulor runs faster than an ARM3 does (bearing in mind this is on an ARM9, never mind an XScale or OMAP), technically it's already more than fast enough!
Martin I don't understand. You seem to say ARM3 is slow "ARM3 wasn't known for its speed." the it's fast enough "it's already more than fast enough!"
I remember being amazed how fast the ARM3 was, one Mandelbrot program was 10 times faster IIRC! |
| |
[ Log in to reply ] |
| |
|
Andrew Rawnsley |
Message #118649, posted by arawnsley at 13:04, 12/9/2011, in reply to message #118647 |
R-Comp chap
Posts: 600
|
It is also worth remembering that most 26bit apps (that haven't been 32bitted) were designed to run on legacy hardware - typically ARM 250 as a baseline in the classroom (A3020). Programs like Eureka and Impression are still perfectly usable on a 30Mhz ARM3 / 33Mhz ARM 610 machine for ordinary tasks. |
| |
[ Log in to reply ] |
| |
|
Martin Bazley |
Message #118650, posted by swirlythingy at 13:04, 12/9/2011, in reply to message #118647 |
Posts: 460
|
Martin I don't understand. You seem to say ARM3 is slow "ARM3 wasn't known for its speed." As I'm sure you're aware, the ARM3 came out 20 years ago. It was clocked at 25MHz. The OMAP3530, on which a future version of Aemulor will likely be running, is clocked at 720MHz. That's a heck of a lot of spare cycles to get lost in the emulation overhead.
If the aim is to run software written for the ARM3, then it is safe to assume that software was written for a much slower processor than the one it finds itself running on, and (particularly in the case of games) will likely be unusably fast. Your Mandelbrot program was also written 20 years ago, and no matter how fast it may have seemed running on the technology of the day, chips these days are much, much faster. If Aemulor achieves an emulated clock rate of greater than 25MHz, it's already more than fast enough. Clear? |
| |
[ Log in to reply ] |
| |
|
Chris Evans |
Message #118660, posted by CJE at 12:46, 13/9/2011, in reply to message #118650 |
CJE Micros chap
Posts: 228
|
Martin I don't understand. You seem to say ARM3 is slow "ARM3 wasn't known for its speed." As I'm sure you're aware, the ARM3 came out 20 years ago. It was clocked at 25MHz. The OMAP3530, on which a future version of Aemulor will likely be running, is clocked at 720MHz. That's a heck of a lot of spare cycles to get lost in the emulation overhead.
If the aim is to run software written for the ARM3, then it is safe to assume that software was written for a much slower processor than the one it finds itself running on, and (particularly in the case of games) will likely be unusably fast. Your Mandelbrot program was also written 20 years ago, and no matter how fast it may have seemed running on the technology of the day, chips these days are much, much faster. If Aemulor achieves an emulated clock rate of greater than 25MHz, it's already more than fast enough. Clear?
I see what you mean, yes a lot of the time you wouldn't notice it being slow, though some things like printing would be a pain... |
| |
[ Log in to reply ] |
| |
|
Adrian Lees |
Message #120647, posted by adrianl at 01:33, 20/6/2012, in reply to message #118617 |
Member
Posts: 1637
|
Getting there. Quick snapshot of RPi running the ARM610 engine, which is now working on my ARMini(/BB) too. Still some stuff to tidy up, unsurprisingly....
[Edited by adrianl at 02:33, 20/6/2012] |
| |
[ Log in to reply ] |
| |
|
Trevor Johnson |
Message #120648, posted by trevj at 05:44, 20/6/2012, in reply to message #120647 |
Member
Posts: 660
|
Oooh! Great news, Adrian  Will it still be sold by Spellings? How does Blu fare (apart from the display output, which I expect is an issue at this stage)?
[Edited by trevj at 06:45, 20/6/2012] |
| |
[ Log in to reply ] |
| |
|
David R. Lane |
Message #120653, posted by DaveLane at 20:01, 20/6/2012, in reply to message #120648 |
Member
Posts: 4
|
Great news! When can I get it! I have got a few applications I want to run on my BB like the spread sheet software, Eureka, and the game FreeCell. |
| |
[ Log in to reply ] |
| |
| Pages (3): 1
> >|
|
