RedSquirrel RO 5.19

Have been running Red Squirrel for a while with RO5.19.
Loading RO from ‘romsets’ – this works up to RO version 5.19 (19-Jun-2012).

But by RO version 5.19 (12-Nov-2012) it stops working.

Any ideas – are there any RO roms between those dates available for checking?

Thanks

are there any RO roms between those dates available for checking?

On a public server: not that I know

On an archive here:
IOMDHALDev.5.19-2012-07-22.zip
IOMDHALDev.5.19-2012-07-23.zip
IOMDHALDev.5.19-2012-08-07.zip
IOMDHALDev.5.19-2012-08-08.zip
IOMDHALDev.5.19-2012-08-10.zip
IOMDHALDev.5.19-2012-08-19.zip
IOMDHALDev.5.19-2012-08-24.zip
IOMDHALDev.5.19-2012-08-26.zip
IOMDHALDev.5.19-2012-08-27.zip
IOMDHALDev.5.19-2012-08-29.zip
IOMDHALDev.5.19-2012-08-31.zip
IOMDHALDev.5.19-2012-09-01.zip
IOMDHALDev.5.19-2012-09-02.zip
IOMDHALDev.5.19-2012-09-03.zip
IOMDHALDev.5.19-2012-09-06.zip
IOMDHALDev.5.19-2012-09-07.zip
IOMDHALDev.5.19-2012-09-09.zip
IOMDHALDev.5.19-2012-09-10.zip
IOMDHALDev.5.19-2012-09-12.zip
IOMDHALDev.5.19-2012-09-13.zip
IOMDHALDev.5.19-2012-09-16.zip
IOMDHALDev.5.19-2012-09-17.zip
IOMDHALDev.5.19-2012-09-19.zip
IOMDHALDev.5.19-2012-09-20.zip
IOMDHALDev.5.19-2012-09-22.zip
IOMDHALDev.5.19-2012-09-26.zip
IOMDHALDev.5.19-2012-10-01.zip
IOMDHALDev.5.19-2012-10-08.zip
IOMDHALDev.5.19-2012-10-10.zip
IOMDHALDev.5.19-2012-10-11.zip
IOMDHALDev.5.19-2012-10-13.zip
IOMDHALDev.5.19-2012-10-14.zip
IOMDHALDev.5.19-2012-10-15.zip
IOMDHALDev.5.19-2012-10-17.zip
IOMDHALDev.5.19-2012-10-19.zip
IOMDHALDev.5.19-2012-10-21.zip
IOMDHALDev.5.19-2012-10-22.zip
IOMDHALDev.5.19-2012-10-24.zip
IOMDHALDev.5.19-2012-10-27.zip
IOMDHALDev.5.19-2012-10-29.zip
IOMDHALDev.5.19-2012-11-02.zip
IOMDHALDev.5.19-2012-11-05.zip
IOMDHALDev.5.19-2012-11-07.zip
IOMDHALDev.5.19-2012-11-12.zip

Which would you like to go for as a starter?
Binary search technique and take the mid-point? i.e. 2012-09-17

Yes – seems like a good choice :-)

Yes – seems like a good choice

Mail? (No FTP server setup¹, but I could FTP to a server)

¹ Keep meaning to slip something into the DMZ at work, but people keep finding more work than the hours in the day

On an archive here:
IOMDHALDev.5.19-2012-07-22.zip

I think you should start with this one. I’m prepared to bet £1.50 that the problem is that RedSquirrel isn’t emulating accurately enough to pass the POST, which was (re)added in July 2012.

From the looks of the RedSquirrel website it was last updated in 2002, so I suspect the chances of getting the emulator updated are slim to nil and it’d be simpler to switch to RPCEmu – indeed posting the question in the RPCEmu forum should have hinted at that, no doubt someone will have done benchmarks and be able to show it’s faster too (as it includes a re-compiler).

RedSquirrel seems pretty stable – if there is an error – you seem to get thrown out of full screen – a couple of Alt/Rtns and its back working.

RPCEmu – seems to just buckle under – restart needed :-(

Would it be possible for a RO5.21 ROM be produced without the POST – surly only needed for a actual Hardware ROM?

I’d be interested in getting the latest RO working Red Squirrel as well. I did eMail Graeme about the source available recently, as I wanted to add instruction tracing to it.

Why Red Squirrel? Well, out of all the available emulators, its still the most accurate emulation of a StrongARM. RPCEmu’s MMU doesn’t generate Permission Faults
There’s also a CMP instruction mismatch

Admittedly Red Squirrel does have a few issues, but they’re minor in comparison. As well as the RO 5 POST issue, it doesn’t generate Page Faults on 1KB subpages.

RedSq seems to work ok here – with RO 5.19 – but having a problem with XOS_ClaimDeviceVector – I think – with a module loading from PoduleROM.

Module works ok when loaded from Predesk with RO5.19 but not with RO5.21 – must have made a mistake. (RedSq crashes out)

Did send a email to Graeme – about which version of VA would be best to get, for running RO5 on [ no reply so far] :-(

Is there anyway of getting VA to load a different version of ROM – other than softloading it? (are the various versions any different in this?)

IOMDHALDev.5.19-2012-07-22.zip

Steve, could you eMail me this one as Sprow suggests, I’ll see if it still works under Red Squirrel. I’ve run into some odd issues with RPCEmu’s IOMD emulation and need to see what happens under Red Squirrel.

jon at jaspp dot org dot uk

IOMDHALDev.5.19-2012-07-22.zip

Steve, could you eMail me this one

Should be thudding into the mailbox about now.

Looks like I need to go back further, 2012-07-22 doesn’t POST on Red Squirrel.

Looks like I need to go back further, 2012-07-22 doesn’t POST on Red Squirrel.

IOMDHALDev.5.19-2012-03-18.zip is the earliest I have (disc crash zapped various earlier)

However, the Internet Archive has earlier like this from the 19th of January 2012 – late 2011 and the files for 5.17 are on the page.

Thanks Steve, I downloaded the few that were available via Wayback:

15-01-2012
09-05-2012

15th Jan works under RS, 9th May doesn’t

Version 5.19 (19-Jun-2012) is working here & now (sending this) -loading from ROM dir.
Anyone out there that rolls their own OS – able to do a RO 5.21 version for emulators -without POST?

Colin, could you eMail me a copy please (jon at jaspp dot org dot uk).

What did you do about the HostFS module being 26bit? I’m about to 32bit it, but you can save me the trouble if you already have a 32bit version.

Whilst playing around with Red Squirrel, I noticed that the POST fails if more than one RAM module is configured past the 15-01-12 build (only have the 09-05-2012 build to test, so could be sometime between the two builds)

Could the POST issue be down to the memory detection and unrelated to the full POST being added?

25-02-2012 is when I rewrote the RAM detection code in order to try and fix some RAM related issued people had been reporting with previous ROMs.
However my new code did contain a couple of bugs that would have caused issues with some setups – those were fixed on 19-06-2012.
15-07-2012 is when the POST was added.
Kinetic support (which is another likely cause of incompatibilities) was in 28-03-2013

I just looked at the POST source and immediately spotted the problem. The emulation detection only detects RPCEmu. I tested this by modifying the 22-07-12 build and sure enough it boots, I also tested with the 26-01-15 build which also boots.
I’d suggest changing the emulation detection to:

LDRB    r4, [r5, #IOMD_VERSION]
TEQ     r4, #0   ; RPCEmu has version ID of zero
TEQNE   r4, #&9c ; Red Squirrel has version ID of 9C
MOVEQ   pc, lr

Thanks for finding this out.
I notice that with RO5.21 the screen memory can be dragged out to 32Mb.

Since the * command VIDCBandwidthLimit is not implemented – the screen resolution/16M colours is very limited – bit like a RPC with no VRAM.
(RO5.19) can go up to at least 2048×1536).
[edit]
After trying RO5.19 (12Nov 2012) it seems to behave the same as RO5.21 – so perhaps the change happened when the POST added.

It would be nice if the BandwidthLimit could be removed for use of RO5.21 with respect to the emulators – without upsetting others. Should allow max screen res – with emulation of SArm processors.
What causes the writing to be partly offset – to the right – when f12 is pressed?
It seems to vary with ROM versions.

Did anything happen about removing the ‘BandwidthLimit’ from later versions of RO5 – for use with emulators?

Does the latest versions of RO5 – for emulators – have all the refs to zero page removed?

So running a prog like ‘Protector’ to catch peek/pokes to zero page – could be used to catch programming errors.

Did anything happen about removing the ‘BandwidthLimit’ from later versions of RO5 – for use with emulators?

Nothing’s happened with the bandwidth limit yet – but (since 12th Aug 2012) there’s support for raising the bandwidth limit to 608MB/s when 8MB of VRAM is detected. That’s just enough for 1080p @ 60Hz. That works fine with RPCEmu, but it sounds like for some reason the OS isn’t detecting any VRAM being present when running on Red Squirrel – there’s only enough space in the physical memory map for 8MB of VRAM, so for screen memory to go up to 32MB it must be using regular RAM.

Does the latest versions of RO5 – for emulators – have all the refs to zero page removed?

No. Support for high processor vectors was only introduced with ARMv5, so on IOMD we need to at least keep the processor vectors located at &0.

We could provide some protection by moving kernel workspace away and making page zero completely inaccessible in user mode, but I don’t have any immediate plans to make that change.

It seems that if VRAM is set to zero – then you get full screen memory from regular RAM.

If the Bandwidth limit was lifted – with VRAM=0 – then later versions of RO5 could be used/tried.