Aemulor on the Beagle/Panda boards
manuel (1438) 23 posts |
I just found what an awesome app this was on the Ionyx, so I was wondering if it’s ever going to be released with Panda/Beagle compatibility. I know Adran Lees, it’s author, is around here, so maybe he can answer :) thanks |
Chris Hall (132) 3554 posts |
The rumour was that the Xscale processor on the Iyonix had a set of registers (unused by RISC OS) which greatly facilitated the Aemulor software but that such registers are not present on the ARM7 architecture making it a difficult ‘start from scratch’ task. But I may have misunderstood. Also ARM7 introduces a new set of constraints (such as unaligned 8 bit load/store ops). |
Chris Evans (457) 1614 posts |
Adrian in another thread mentioned in passing that he has run Aemulor or the Raspberry Pi! |
Trevor Johnson (329) 1645 posts |
|
Bryan Hogan (339) 592 posts |
AIUI the Xscale has some hardware features that helped Aemulor do its fixes quickly, but that the software method was more portable. Didn’t Adrian get the A9 version working during the afternoon of the show at which the A9 was launched? As it is fixing software designed to run on an 8MHz chip so that it runs on a 1GHz chip, I think the “slow” version will be fine! |
Keith Dunlop (214) 162 posts |
From research that we have done it would seem that a lot of people are hanging on to their Iyonixes just because of Aemulor… |
Bryan Hogan (339) 592 posts |
Probably still a lot of Impression users out there :-( What other popular software was never 32 bitted that will be forcing people to keep hold of their RiscPC or Iyonix? I’m sure RComp would love to see Aemulor available for the ARMini as it would probably result in a nice boost in sales! |
Steffen Huber (91) 1953 posts |
Bryan, unfortunately, much of the “aemulored” software was used on 200/233/287 MHz machines. The Aemulor miracle on the IYONIX was to be able to run the old 26bit software at StrongARM speed. The purely interpreted Aemulor was more like an ARM610/30 MHz speed-wise. On the other hand, the OMAP3/OMAP4 seems to be a lot faster overall than the IYONIX XScale, so maybe the interpreted Aemulor would be all we really need – so Adrian, stop teasing us and just release it ;-) |
Jess Hampshire (158) 865 posts |
I think there are a couple of other reasons too. (they cost a load of money and they have better display support and faster disk access than any of the new machines). I don’t have aemulor, but I think it will be a pretty vital piece of software if the RISC OS Pi is to get into education. As I understand it the Pi has the same generation of ARM as the Iyonix, which means that a lot of Iyonix compatible software will run. However a lot of (almost all?) educational software will be designed for A5000 or RiscPCs with no chance of a port being made. Without aemulor, the only other software that would run would be BBC BASIC. |
Chris Evans (457) 1614 posts |
If I remember and understand correctly what Adrian told me at the London show Aemulor should only be (back of an envelope quessimate) 20 to 25% slower on BB/RPi than on an Iyonix! The target market I think would be very happy with half the speed or lower of an Iyonix just so they could use their old software. |
Bryan Hogan (339) 592 posts |
Running slowly is infinitely better than not running at all ;-) |
Adrian Lees (1349) 122 posts |
You wouldn’t prefer it working first? ;) There are some (I suspect small) issues with the ARM610 engine on both the RaspberryPi and the ARMini at present. Since it is basically ARMv3 code, though, I suspect it won’t require much work and I’ve already handled the alignment issues of the Cortex devices. It in fact runs on ARM610/StrongARM processors too, and this is the reason that it took little work to do an A9home version, though I didn’t have a ‘proper’ day job then ;) The StrongARM engine will not work at all at the moment, because it’s XScale-specific. I believe there is comparable hardware in both the ARM11 of the RaspberryPi and the Cortex-A8 of the ARMini such that it should be possible to deploy that too after some more extensive work….however, I’m not promising that until it’s working. My priority is to get the ARM3/610 engines running properly and then worry about the StrongARM engine later after an initial release. |
Martin Bazley (331) 379 posts |
No, it doesn’t. The Iyonix is ARMv5, the Pi is ARMv6, and the OMAP is ARMv7. The issue with unaligned memory accesses was introduced in ARMv6, so I’m afraid that’s purely wishful thinking on your part. (Although I believe the ARMv6 has a compatibility flag which allows it to emulate the old ‘rotated load’ behaviour, which was removed in ARMv7, so that might be worth looking at…) |
Adrian Lees (1349) 122 posts |
Progress or more teasing, depending upon your perspective. |
manuel (1438) 23 posts |
Awesome!! Would you test some games? A simple one, like Accorn Lemmings, would do :) I know I’d probably have t set custom resolutoins etc, but I’m all in for old Accorn games on the Rpi and the Pandaboard! (So many incredible Amiga conversions) |