Out of memory

[sundown]

I've done the test myself and I can confirm the problem exists on X1000 at least on plain Update 6 and on up to date Update 6... However the problem does not show on a beta system.

I'll have to re-check on beta. My test was to copy large video files to RAM: till it said "out of memory", then delete the files. Last I tested, the memory in RAM: was not cleared & crashed soon after. I filed a bug report (7880) many months ago.

There is a hack to "hide" the problem, I won't post it. Under normal use, 1.6GB of memory is more then enough to run apps.

By the way, I don't use the hack myself, it was only to test.

Ken

[zzd10h]

My problem is not about use of the OS with 400mb remaining but
why copy 5mb of data waste 1gb of memory on x1000.

I don't copy use very big files

[abalaban]

@sundown

Sorry I should have specified that I was referring to zzd10h's problem.

[sundown]

@sundown

Sorry I should have specified that I was referring to zzd10h's problem.

Understood, I did another test last night. Ram disk reports 1.5GB free on bootup, if I copy a 1.6GB file to RAM:, it now just stops with no message, I can do a soft re-set. If I copy the same file to ram from a shell, the system just locks up with no message, needs a hard re-set. I'm inclined to believe the developers have bigger fish to fry before they worry about this problem.

Ken

[Raziel]

Quite an old thread

May i add my findings of the last days?

I alos go the same error, yet for me it's crash in ramlib.support (anyone needs a log, i can provide some)?

Could someone proabably confirm that this has been fixed (beta testers?) and will be no more with FE?

Also there is some mixing in of a similar Notepad on "copy and paste crash" in here aswell, would be awesome if those two bugs could be squashed with one update

[danbeaver]

I guess that out of the 2 GB of ram, none is used for system addresses, pointers, USB and network stacks, and the OS (operating system); these locations should come from some sort of free "cloud" memory that is entirely out of the range of addressable system locations. Therefore, the whole 2 GB should be available. To the best of my knowledge, only 256K is allocated out of Graphic ram as accessible in the current OS. I could be wrong now....

[Raziel]

I guess that out of the 2 GB of ram, none is used for system addresses, pointers, USB and network stacks, and the OS (operating system); these locations should come from some sort of free "cloud" memory that is entirely out of the range of addressable system locations. Therefore, the whole 2 GB should be available.

Uh, you might be up to something here.

The new memory system "can" address memory beyond the 2 GB range, yet all programs that want to access it, needs to be recompiled (wonder if that will be a big change in the source?)

I don't know if the whole range of another 2 GB could be used with this method or only lets say 512 MB, would be nice to hear a statement

If enough memory could be addressed beyond the barrier than we would have some kind of memory protection, wouldn't we?
Given the fact that only system programs (the ones you posted) and specifically written programs (which in turn would have to follow strict rules) could adress them..

To the best of my knowledge, only 256K is allocated out of Graphic ram as accessible in the current OS. I could be wrong now....

256 MB
Yes, it still is, but maybe that will change too (not with the new memory management) but maybe with an updated gfx driver?

[danbeaver]

The "new" system will allow access to >2 GB of memory as "extended Memory" (i.e., LIMM) that is paged in and out; it requires a rewrite of the OS to use 64-bit addresses rather than the 32-bit it currently uses -- this is why Ubuntu/Linux can address more than 2 GB.

I haven't read that the new drivers are written to use the GPU's routines to access the full amount of memory on the graphics card, but this would make sense to me if they are able to do all the things the promo's say.

[Severin]

Ok. lets clear a few things up here...

OS4.x can address around 1.65 Gb of ram, with the older systems where you have 256mb, 512mb or 1gb of ram it's not a problem and you had the pager to swap things in and out of the other available memory locations and stored on the swap partition. A swap partition is pointless on the X1000 as it can address all the available ram space directly having 2gb of real physical ram. Don't forget that kickstart takes ram that isn't reported as used on OS4. So when it says you have 400mb free you don't, it's reporting the 2gb physical ram it can see minus what it roughly thinks it's used. it doesn't know anything about the unaddressable space or what kickstart is using.

OS4.1+ memory reporting is not very accurate. with the new ram system ram is freed/flushed when needed and things like avail, workbench titlebar, sysmon etc only give a vague estimate of what's really there.

256mb graphics card... 256mb is a limit within Picasso96. It has nothing to do with the amount of memory you can use on the card, warp3d, gallium etc will be able to use all the memory of the card. 256mb for 2D workbench & own screens is more than enough so stop worrying abnout it.

As for loading a 4mb picture into notepad... probably not something the developers ever thought anyone would be stupid enough to try. it's more likely the data itself rather than the size that's causing the problem. an awful lot of non ASCII in there for a program only designed to eidt small ASCII files. might as well try loading an 800mb mp4 file into gui prefs then complain when it fails.

IF you want to do tests like that grab something like lord of the rings from project gutenberg and use the join command to append it several times to get a large ASCII file for testing with. eg.

join lotr.txt lotr.txt lotr.txt as big-ascii.txt

[danbeaver]

Well I'm glad that was cleared up!