Updated: Possible fix for the "Bad Allocation" error

Oh that´s good news!
Testing it right now! Fingers crossed!

Sharandra, any results?

Quote
Yeah, unfortunately still getting the bad allocation error.

Dang, it looks like we've fixed the wrong sh#t...
Well, that's something at least. Gonna keep tryin'.

Thank you for testing and feedback, Sharandra!

Filter was Flagstones, first Preset at 4096x4096,
Windows 7 Home Premium SP1 64bit, Intel Core I5 760, 16GB RAM, GForce GTX 460.
I did send a bugreport too, after the testrun.

Great!

Just did a test and was able to render a 4096x4096 image of the Flagstones filter without a bad allocation error!

Hmm, I´ll have a look through my filters, maybe I have a simple filter that I could use for a test at that huge size.

I've started to render blurred checker 64k x 64k. Not sure it'll be done by the morning though

I started a blurred checker too, but it didn't finish -- I forgot to set the swap location in FF options to a larger HDD

Now I'm trying to render a simple 64000x64000 perlin noise, should be a bit faster than a night.

I didn't forget to move the Filter Forge cache to the large and slow HDD.
Anyway, it has been rendering for 15 hrs and still no sight of the first block

Sharandra, same here -- I had to cancel mine too. The intermediate results, however, were pretty encouraging. When I cancelled it, it was about 30% complete, with the first pass of progressive preview fully covering the preview area. The GUI was responsive to zooming and panning, and the machine in general remained responsive as well -- I worked with my Excel spreadsheets just fine while FF rendered in background (though it must be noted that I have Windows installed on an SSD, and FF swap is directed to a separate mechanical HDD).

Anyway, I'll attempt another approach as soon as I have an appropriate time window

I'm postponing my 64000x64000 tests until the next week, but I can say that we rendered your Flagstones filter in 4096x4096 countless times under various conditions without any problems.

I'm reserving my final verdict until I have the results of 64000x64000px tests, but it does seem that we've fixed the Bad Alloc bug. The executable that will be rolled out as a v3.0 update will include some additional fixes that should be helpful as well. For example, if you still encounter the Bad Alloc bug somehow, you'll be able to dial down the memory consumption slider in FF options to reserve some address space for fragmentation (this is not yet available in the current build linked in the original post).

We're now hunting the last remaining rendering bug, this one:
http://filterforge.com/forum/read.php?FID=8&TID=10882

Regarding the beta: we won't update the current beta but we'll include the fix into the Beta 2, which should be out soon.

I have just downloaded the newest upd ate and will test it more and see if I still get the "Bad allocation" or not.

In the first two tests, one with a 6000 x 6000 image it did WORK without any problem, I have to say that before this was not possible

AND much more important with a 14500 x 8500 image and the Dropshards filter when it has reached 99% after rendering it finally gave the "Bad alocation" error

I have not changed any of the settings, and it keeps to be RAM = 85%

Since the last fix I have not had as much of "bad allocation" errors, BUT is also true that most of the things I have been doing are with less than 6000x5000 and when I made the 10.000x10.000 test for StevieJ I did not get the "bad allocation" error but this does not mean that does not happen, it only happens less than before, but still exists and is there sometimes.