@TMTisFree

Quote:

I have some in the range of 10-50MB that decompresses up to 600MB. I though that 2GB of memory would be enough.

The jpeg decompressor needs to allocate a block of memory big enough to contain the entire bitmap, ie a single 600MB block. It is quite possible, especially if you have other programs running at the time, that there isn't a single contiguous free memory block that size (ie you may have 10 separate 60MB blocks, but no single >=600MB block). Try opening the file in multiview after rebooting, and see if it opens. If it does, the problem is probably what I just wrote. That is the Achilles heel of the Amiga shared memory system, other OSs with a separate address space per task are a lot less susceptible to this.

Last edited by CodeSmith on 23-Aug-2008 at 10:35 PM.

@CodeSmith

in that case the picture should be decompressed in to virtual memory (MEMF_VIRTUAL).

MEMF_SHARED is only needed on messaging, don't use it on private data like pictures and text strings.

MEMF_CHIP and MEMF_FAST are obsolete.

Last edited by NutsAboutAmiga on 23-Aug-2008 at 11:15 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

I don't think that helps any. AmigaOS can address a maximum of 2GB shared among all tasks. MEMF_VIRTUAL simply allows you to "fake" having that much memory even if you don't (so you can successfully AllocMem 500MB even if you only have 256MB installed).

@CodeSmith

iirc memf_virtual is for virtualised address space, not necessarily swap. it should give a contiguous block even if memory is fragmented. afaik this is the default, as is swappable mem. you need another tag to force real ram (memf_physical?)

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

@Chris_Y

Quote:

it should give a contiguous block even if memory is fragmented.

But how? Remember there is one single address space shared among all tasks, and AmigaOS doesn't compact memory like MacOS < X did. The only possible way I can think of that MEMF_VIRTUAL would prevent fragmentation is if there isn't in fact one single shared address space for MEMF_VIRTUAL and the MMU changes its contents with every context switch (ie if address 0x123456 contains 10 when task #1 is running, 20 when task #2 is running, and 10 again when task #1 gets the CPU again). This would break the Amiga message passing system, unless of course messages are not allowed to live in MEMF_VIRTUAL (IIRC in 3.x they could live anywhere, even on the stack). Mind you, if that *is* the case, then OS4 will have partial process memory isolation, and that's a good thing. Is this how OS4's MEMF_VIRTUAL works?

@all

Just to be sure, I attempt to load a 59MB JPEG file in Multiview:
the crashlog is here.

Bye,
TMTisFree

_________________
The engineering approach to our non-problems: "build a better washer".
The scientific approach to our non-problems: "find a new energy source".
The environmentalist approach to our non-problems: "stop washing your shirts".

@TMTisFree

When using PicShow, it just freezes the machine (cold reboot needed).

Bye,
TMTisFree

_________________
The engineering approach to our non-problems: "build a better washer".
The scientific approach to our non-problems: "find a new energy source".
The environmentalist approach to our non-problems: "stop washing your shirts".

@CodeSmith

As i remember it, the application gets a virtual address which maps to a physical address. This virtual address space is in one block, however the physical memory it maps to is not necessarily.

If I could find the document I've just been looking for, I would be able to say for sure, but I appear to have lost it.

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

@TMTisFree

Quote:

When using PicShow, it just freezes the machine (cold reboot needed).

Quote:

Temporary lockups can happen when the picture is scaled to another size in the background. If the picture needs more than 10 seconds to load, it will probably need a similar amount of time to be scaled to screen size. More if interpolation is used. So just wait a little, perhaps it unlocks again. These lockups are Picass96's fault. P96 locks the entire computer instead of just the bitmap which is accessed when using LockBitMapTags(). It does not happen with Cybergraphics. PicShow was never meant to display such huge images. However, a complete redesign of it is currently being planned. It can only be a matter of years until it's ready Quote: Layout failed This happens on low-memory conditions. If you don't manage to quit PicShow, there usually is no other way out but a reboot. This is a known problem (bad design). Bye, Thomas

source #16

#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

@TMTisFree

Quote:

I have a very good one, I just miss an USB driver for the scanner.

Most of the combination printers (fax,scan,print) and some of the seperate scanners scan to memory now. You have to go to the printer to place the original on the scan window, so you just push the scan button on the scanner at the same time. The scanned image is then available on your computer through the USB connection. It's actually quicker for me to place the original, push the scan button and remove the original than to place the original, return to the computer to start the scan from a program and remove the original. Dedicated scan programs are pretty useless with modern equipment in my view.

Quote:

I have some in the range of 10-50MB that decompresses up to 600MB. I though that 2GB of memory would be enough.

I just did an experiment to save a jpeg at the same compression ratio you mentioned (600MB to 50MB) and had to save the jpeg at 20% quality to match that ratio. The resulting picture has obvious blocks of color and looks like the images in some old low-res games. Such a picture would be pretty worthless in my view. Besides, how would you print such a large picture accurately (printing all the pixels)? Wouldn't the print software need to reduce the size to print it anyway?

The biggest downside for OS4/OS4.1 is the lack of a modern Internet browser. I've already preordered OS 4.1 but I'm going to be forced to switch to a PC if we don't get a new browser soon. I don't have the room for a clutter of different computers, so the PC will replace the µA1. I can't take much more of the scrambled up WEB pages, links that don't work etc. My Yahoo page is almost unreadable and is barely usable. I can't use Mapquest any more. I can't even look for a new apartment at Realtor.com because it won't work. I can't find the best interest rates because bankrate.com doesn't work. And on and on etc.

_________________
X1000 with 2GB memory & OS4.1FE

@CodeSmith

Correct you can’t use MEMF_VIRTUAL for the message system, and you can use it whit in forbids but you can use semaphores instead I guess, so it has it set of limitations, but should be prefect for images and text, and stuff like that you are never going to share whit an other program.

Anyway manually allocating memory for the message structure is also obsolete in AmigaOS4.0, if you follow the programming guidelines.
Quote:

OS4 will have partial process memory isolation

You are completely correct OS4.0 has partial memory protection.

Read about memory system her:
http://os4.hyperion-entertainment.biz/index.php%3Foption=content&task=view&id=22&Itemid=.html

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@number6

Thanks for the link, wasn't aware of the pb.

Bye,
TMTisFree

_________________
The engineering approach to our non-problems: "build a better washer".
The scientific approach to our non-problems: "find a new energy source".
The environmentalist approach to our non-problems: "stop washing your shirts".

@Xenic

Quote:

modern Internet browser

If viewing CSS pages is your definition of modern Internet browser, then you can just download OWB 2.6 and be happy.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@TMTisFree

It’s possible that datatype is using chip memory, chip and fast memory is limited to 64 MB max, I think, but let’s say its 24mb or more is already used, then you will run out of space for you large image quickly, memory allocation fails program crashes, because the program does not have a fault handling routine for failures.

I think setpatch has addchip switch you might won’t to try that se if you can increase max available chip/fast memory.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@Xenic

You cannot have those all-in-one machines in a professionnal environment. Imagine if a part of it is out of order...
My scanners all have multisheet charger. I don't have to push whatever button.

Concerning the compression ratio, I suppose it depends of what you have in the data to be compressed (for ex. redondant data, large part of the same color, black and white color, etc). I also have some JPEG in the range 50-100MB I can display but not convert on a PC with 4GB of RAM: I have to use a special program to convert them raw format then in pyramidal TIFF and then use a web image server (à-la googlemap) to be able to view them.

About browsers, well, I use a mix of AWeb (scriptable), IBrowse (rarely), and of course OWB (v1.24 and v2.6). OWB is progressing nicely, I think. Otherwise it's Firefox.

Bye,
TMTisFree

_________________
The engineering approach to our non-problems: "build a better washer".
The scientific approach to our non-problems: "find a new energy source".
The environmentalist approach to our non-problems: "stop washing your shirts".

Just a thought is OS4.1 going to have the startup-music we were told about years ago?

_________________
Rage quited 29th May 2011

@Stephen_Robinson

Yeah i cant belive Hyperion failed on that annonuncement......first time mesa thinks

_________________
On Planet Boing Trevor is God

@Hondo

Quote:

Yeah i cant belive Hyperion failed on that annonuncement......first time mesa thinks

Unfortunately, they did not: from what I heard, Hirasawa's *is* in AOS 4.1.
(I say "unfortunaltely" because I hate that useless, annoying, resource-wasting kind of stuff.)

saimo

_________________
RETREAM - retro dreams for Amiga, Commodore 64 and PC

@CodeSmith

Quote:

@Chris_Y Quote: it should give a contiguous block even if memory is fragmented. But how? Remember there is one single address space shared among all tasks, and AmigaOS doesn't compact memory like MacOS < X did. The only possible way I can think of that MEMF_VIRTUAL would prevent fragmentation is if there isn't in fact one single shared address space for MEMF_VIRTUAL and the MMU changes its contents with every context

That would be what MEMF_PRIVATE is for. Memory allocated with the MEMF_PRIVATE flag can be
per process. If this is implemented however, is another question entierly.

I seem to recall that for compatibility reasons, all memory is shared (MEMF_SHARED) unless explicitly declared private (MEMF_PRIVATE).

Hmmm according to Amigakit, this isn't out till next month now Argh!!!

One more week to go....

_________________
Rage quited 29th May 2011