crash after StarAlignement with large data sets in 01.08.04.1171 (x64, Win7)

[Juan Conejero]

I have just finished a series of tests with version 1.8.4.1171 and StarAlignment on Windows 7, Windows 8.1 and Windows 10 (clean installation). No problem at all with 250 debayered DSLR images (4290x2852 pixels RGB each). No slowdown at all. No noticeable final freeze. I cannot reproduce this problem on any of our working Windows machines.

[msmythers]

Well I guess I will just have to stay with 1.8.3.1123 for large sets. I have all updates, no viruses or any other thing I know to do.



Mike

[Juan Conejero]

Update: I have managed to reproduce the final freeze (5 minutes) after registering 200 images on a Windows 8.1 virtual machine (but no slowdown or crash). Definitely this is a memory management problem, but StarAlignment is not causing it. I'll investigate this further.

[msmythers]

Thanks Juan.

I have not had a crash but always a freeze on my computer.



Mike

[Philippe B.]

Hi Juan
This is not a crash.


This is just a kind of freeze of PI during several minutes. And closing PI process is the only solution to speed up.


It depends of the amount of data we process.
For example, with my 12Mpix sony A7s, problem comes around 220th image. With my nikon D810A, it is around 50 of so.
Also, when I am doing this with a small amount, there is sometimes a freeze (1-2 minutes) and PI go back to life.
 
Cheers

[lucchett]

I think I had the same issue with 35 4kx4k pixel images under Fedora 22.

the process completed quickly, but than PI froze.

I had to kill PI.
I noticed the Cpu was driven max speed with turbo mode on (3.8Ghz in my case).
killing the process stooped turbo mode

andrea

[Juan Conejero]

I have good news: I have managed to find, understand and solve this problem. This bug has been fixed in the next 1.8.4 version, which will be released for Linux and Windows very soon. Sorry for the inconvenience.

[tommy_nawratil]

 8)       
thank you Juan!

today we had a PI workshop with some 20 people here in Vienna, they all had no problems to stack a 15pic set in BPP without freeze.
Good to know it will work again on large sets too!
I noticed the project and xisf data created in 1171 were not readable for the guys running 1123 ("no xosm 1.0"),
but could manage by converting the data to fits format.
There is no way to have project data compatible by some "save as xosm 1.0" I guess?
Will have to think how to handle next weeks session, the guys are pretty slow in updating 

Tommy

[Juan Conejero]

There is no way to have project data compatible by some "save as xosm 1.0" I guess?

No, this is not possible. 1.8.4 can read XOSM 1.0 and 1.1, but only write 1.1. Version 1.8.3 of PixInsight will become obsolete during this month, once we release the next 1.8.4 version.

[Philippe B.]

Thank you Juan !!! You're the best  8)

[kuhomo]

Hi, I have a similar problem. I'm using W8-64 on a Lenovo T430 with 8GByte Memory. Image Integration for bias and darks only works for less then about 90 raw files (6000x4000, files from a Nikon 5300). With flats and lights the problem even occurs earlier. For my tests I always have rebooted the computer.
In some test runs PI (1.08.03.1123) gave the message that there is not enough memory available, but PI was not usable anylonger, in other cases it simply rebooted the computer.
I have attached a screenshot, which documents the memory usage before and after the above mentioned message appeared. Commit Charge steadily increases until PI refuses to work.

Any ideas?

Greetings from Cologne, Germany.

[Philippe B.]

Juan
Do you know when you could have a new version which solve this problem ?
Thank you
Cheers

[jkmorse]

Juan,

Your note seems to suggest a fix in Linux too, but no one has mentioned it directly.  I am running 1171 in Linux (Linux Mint 17.2) and, while my freezes are significantly less than described in this post, even with small alignments of 25 images I am getting a short (5-10 second) freeze at the end.  Just making sure you are aware that the issue exists in Linux as well.  I am running this on a desktop with a core i7 chipset.  For swaps I point to a 16Gb ramdisk (half of my installed 32Gb of Ram).  PI is installed on a 512Gb SSD (Samsung TVO Pro).  Let me know if you need any other details.

Best,

Jim

[Juan Conejero]

Hi Jim, Philippe,

Yes, this bug exists on both Linux and Windows. It is completely fixed in the next 1.8.4 version.

The new version is very very close to be ready for release. Just a few more tests have to be made, especially on Windows, so it is a matter of days. See the benchmarks for new build 1185:

http://pixinsight.com/benchmark/

[jkmorse]

Wow!!!