Available now: AmigaOS 3.1.4

[indigolemon]

I have the same with my Blizzard, exec.library and expansion.library always end up in Chip Ram. The best explanation I read was:

The sad truth is that your accelerators RAM is NOT AUTOCONFIG. I mean it is not recognized automagically by AmigaOS during the system initialisation procedure. When your system boots, it doesn't see any FAST RAM. That's why it uses CHIP RAM for jump tables of two very first libraries - exec and expansion. After this stage of initialisation, it calls procedures on all autoconfig boards it finds - and a piece of code situated somewhere in your accelerator board just adds its memory to the global system memory list. But it makes it AFTER exec and expansion are created, that's why they are not in FAST RAM.

[graffias79]

Originally Posted by indigolemon

I have the same with my Blizzard, exec.library and expansion.library always end up in Chip Ram. The best explanation I read was:

Well that's discouraging.

I think I'm just going to leave the system at 3.1. At least it ran faster that way.

[Daedalus]

Okay, maybe I'm missing something (somewhat sleep deprived today), but I wouldn't expect Autoconfig RAM to be available until after exec and expansion are already open, since exec is needed to open expansion, and expansion is needed to run Autoconfig. I know Blizzards have some funky Autoconfig behaviour, but if they add fast RAM to the pool at the same point that Autoconfig does, then surely there's little real difference, as exec and expansion are already opened in chip RAM and can't be moved without a reboot?

[desiv]

Originally Posted by Daedalus

Okay, maybe I'm missing something (somewhat sleep deprived today), but I wouldn't expect Autoconfig RAM to be available until after exec and expansion are already open..

Interesting that for me, EXEC loaded into FAST in WinUAE... Just EXPANSION was in CHIP.

Man, I have to clear out my desk and try this on my A1200 now.. ;-)

[Toni Wilen]

Yes, Blizzards (and many others) have non-autoconfig RAM that gets added by board's ROM code.

Originally Posted by Daedalus

Okay, maybe I'm missing something (somewhat sleep deprived today), but I wouldn't expect Autoconfig RAM to be available until after exec and expansion are already open, since exec is needed to open expansion, and expansion is needed to run Autoconfig. I know Blizzards have some funky Autoconfig behaviour, but if they add fast RAM to the pool at the same point that Autoconfig does, then surely there's little real difference, as exec and expansion are already opened in chip RAM and can't be moved without a reboot?

Exec gets re-created in "best" RAM after autoconfig RAM has been added but before autoconfig ROMs are executed. This is only possible if RAM is autoconfig because after autoboot ROMs have been executed, exec can't anymore move or change.

EDIT: Autoconfig or hardwired RAM location that exec checks, like $c00000 or $08000000.

[Daedalus]

Originally Posted by Toni Wilen

Exec gets re-created in "best" RAM after autoconfig RAM has been added but before autoconfig ROMs are executed. This is only possible if RAM is autoconfig because after autoboot ROMs have been executed, exec can't anymore move or change.

Ah, that explains it then, thanks for clearing it up. I guess it makes sense that it can be moved relatively easily before any "foreign" code has been executed.

[graffias79]

Well, I have an A4000 currently being worked on by Acill and it has an A3640 accelerator. I'm wondering if I'm going to run into the same issue if I try to upgrade that to 3.1.4? If so I may not even bother.

[Toni Wilen]

Originally Posted by graffias79

Well, I have an A4000 currently being worked on by Acill and it has an A3640 accelerator. I'm wondering if I'm going to run into the same issue if I try to upgrade that to 3.1.4? If so I may not even bother.

No. A3640 (and all other Commodore boards) do not have this problem.

If board does not have any boot ROM or no autoconfig boards and RAM is detected automatically: it is fine.

If board has boot ROM and has IO only autoconfig board (size much smaller than RAM size) and it does NOT have built-in HD controller, it might be "bad".

If board has IO autoconfig board, boot ROM and HD controller (and no RAM autoconfig board): only way to know is to boot it.

Originally Posted by desiv

Interesting that for me, EXEC loaded into FAST in WinUAE... Just EXPANSION was in CHIP.

Man, I have to clear out my desk and try this on my A1200 now.. ;-)

You get same behavior in WinUAE if you configure Blizzard accelerator without any other type of fast RAM. (It would be quite useless emulation if it didn't..)

[JLPedro]

I have an ACA 1221 and using the acatune maprom to load kickstart from 3.1.4 , expansion.library is loaded to "slow ram" with is the 1MB maprom memory from the ACA, all the rest is in Fast Ram, don't know if it's good or bad.

[graffias79]

It seems to me like this would affect a large number of A1200 accelerators?

[desiv]

Originally Posted by graffias79

It seems to me like this would affect a large number of A1200 accelerators?

Q: Is this new?
What happens with a standard 3.1 ROM and this Blizzard; where is EXEC?
Is this a new "feature" of the new release, or expected and just noticed?

[rockersuke]

Originally Posted by desiv

What happens with a standard 3.1 ROM and this Blizzard; where is EXEC?

Just checked and... according to sysinfo 4.0 in my A1200 with 3.1 ROM and a BLizzard 1230 MKIV 68030 50 MHz with FPU, everything in the libraries section is in "32bitRAM" except... Exec and Expansion, which are in CHIP!

edit: jumpers are configured to map ROM into RAM

--

[graffias79]

Originally Posted by desiv

Q: Is this new?
What happens with a standard 3.1 ROM and this Blizzard; where is EXEC?
Is this a new "feature" of the new release, or expected and just noticed?

I'll put my 3.1 ROMs back when I get home from work and see.

[Chain]

Originally Posted by Nightfox

Is anyone able to create a ROM image that includes the new intuition.library? That would make the LoadModule step unnecessary

Yes, it simply works, no reboot needed

[Madk71]

Hi, I am having some problems partitioning my 32gb CF. The OS says that there are not enough memory despite the fact that my A1200 has an aca1221ec with 8mb. But I don't find any support for the OS. Can someone point me where I am wrong, please?
Any incompatibilities with hardware like Elbox 4xeide?
Thanks

[Daedalus]

Is it telling you it can't validate the volume? It takes a lot of memory to validate a large partition (IIRC about 1MB per GB). There are a couple of ways around this:

First, you can split the drive into smaller partitions, which was traditionally the way Amigas had their hard drives set up anyway.

Second, you can use a larger block size on FFS partitions, which also gives a nice speed boost in some situations. Try 4096 or 8192.

Third, you can use a different filesystem that doesn't require a validation, such as PFS3AIO.

Bear in mind that you should reboot after changing partition sizes, block sizes or filesystems to get the OS to recognise the new setups.

[Madk71]

@Daedalus
Thank you. I will try.
But there is still a mystery about the elbox 4xeide. It starnge to sell an os without any support (it sounds like win*ows) :-).
Do you know if the installer need to be replaced because some software told me that the version is old or they doesn't work at all.
Thanks

[Daedalus]

You mean drivers-wise? You won't need drivers for hard drives connected to the first port of the interface. Only the second port needs a driver, and for that you can use the Elbox one, IDEFix or any other software. That interface also came with software that patched the 3.0/3.1 drivers to break up large drives into 4GB virtual drives, which worked around the drive size limitation in those drivers. But that's not needed under 3.1.4 - large drives can be used without any problems.

If you had an older version of Installer installed on the hard drive before you tried to install 3.1.4, it's possible you need to update it. A version comes with 3.1.4, and is the version you should use.

[Madk71]

Hi, I can't see hard drives or cd attached at the second port(using 4xeideprefs it shows you all the physical attached drives) after I put the rom with the new kickstart (I don't know if it is a fortuity). I will try to swap the kickstart again.
About the installer, I made a new installation from scratch. And some software seems to compare the current installer version with the one on OS3.9, based on the number it seems older than the one on the OS3.9, but actually it may be newer. For this reason the software that allow me to continue are ok. Other software fails, like the installer for the elbox 4xeide.

[Daedalus]

Ah yes. The Installer will report a lower version number since it doesn't support some of the newer features of 3.9, but it's strange that things like drivers require the 3.9 version...

It's possible that the EIDE driver patches the old scsi.device, but deosn't know howto patch the new scsi.device in the new ROMs. I don't know... If you try IDEFix '97 (you can download a trial version from Aminet), it comes with tools for querying the IDE port. See if that can find all the devices - if so, you can just use that driver instead.