Welcome to MilkyWay@home

Memory Leak

Message boards : Number crunching : Memory Leak
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile SuperSluether
Avatar

Send message
Joined: 2 Jul 14
Posts: 15
Credit: 20,991,384
RAC: 46
Message 62157 - Posted: 13 Aug 2014, 13:35:52 UTC

I recently started using Webroot SecureAnywhere as my antivirus program. In a system analysis, Webroot has detected a possible memory leak for milkyway_separation_modified_fit_1.30_windows_x86_64.exe. Additionally, sometimes a possible handle leak is also detected for this process.

Is this error serious, or nothing to worry about? If I need to do something, is there a way to fix the error myself? I have 8GB of RAM on my system, so I don't know if the memory and handle leaks will affect my system because the tasks finish so quickly. This might cause problems on older computers with less RAM though.
ID: 62157 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile arkayn
Avatar

Send message
Joined: 14 Feb 09
Posts: 999
Credit: 74,932,619
RAC: 0
Message 62159 - Posted: 13 Aug 2014, 16:36:12 UTC

We recommend that you tell your AV software to ignore the BOINC data folder as they have all seemed to flag BOINC apps falsely lately.
ID: 62159 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile SuperSluether
Avatar

Send message
Joined: 2 Jul 14
Posts: 15
Credit: 20,991,384
RAC: 46
Message 62160 - Posted: 13 Aug 2014, 20:20:21 UTC - in response to Message 62159.  

We recommend that you tell your AV software to ignore the BOINC data folder as they have all seemed to flag BOINC apps falsely lately.


It's not flagging it as a virus, it's detecting an application error. Either way, it doesn't seem to matter.
ID: 62160 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
swiftmallard
Avatar

Send message
Joined: 18 Jul 09
Posts: 300
Credit: 303,562,776
RAC: 0
Message 62161 - Posted: 13 Aug 2014, 20:53:17 UTC

Modified Fit WUs have been running a long time and this is the first I've heard of a possible memory leak. Positives for Boinc processes we hear of with regularity, and all have been false so far. Exclude the Boinc directory and you'll be happier.
ID: 62161 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
floyd

Send message
Joined: 13 Sep 11
Posts: 17
Credit: 3,263,835
RAC: 0
Message 62170 - Posted: 15 Aug 2014, 8:45:18 UTC

I'm not running Milkyway now but I'm sure there was a memory leak when I last ran Milkyway Separation (not Modified Fit) OpenCL tasks, to the extent where a single task used the whole 2GB I allowed for BOINC and other tasks were suspended for lack of memory. When restarted from a checkpoint, the same task would run with only 90MB, but increasing again. The rate of memory loss depended on the work chunk frequency setting in the Milkyway preferences.
ID: 62170 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3315
Credit: 519,949,340
RAC: 22,061
Message 62171 - Posted: 15 Aug 2014, 10:46:45 UTC - in response to Message 62170.  

I'm not running Milkyway now but I'm sure there was a memory leak when I last ran Milkyway Separation (not Modified Fit) OpenCL tasks, to the extent where a single task used the whole 2GB I allowed for BOINC and other tasks were suspended for lack of memory. When restarted from a checkpoint, the same task would run with only 90MB, but increasing again. The rate of memory loss depended on the work chunk frequency setting in the Milkyway preferences.


At least some of what you are seeing as a 'leak' is by design, as the unit progresses, and finds things, it takes more memory to process that data then at the start of the unit. If it goes back to a checkpoint it should then build up the memory usage again to a similar place it was before.
ID: 62171 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile SuperSluether
Avatar

Send message
Joined: 2 Jul 14
Posts: 15
Credit: 20,991,384
RAC: 46
Message 62173 - Posted: 15 Aug 2014, 13:55:02 UTC - in response to Message 62171.  

I'm not running Milkyway now but I'm sure there was a memory leak when I last ran Milkyway Separation (not Modified Fit) OpenCL tasks, to the extent where a single task used the whole 2GB I allowed for BOINC and other tasks were suspended for lack of memory. When restarted from a checkpoint, the same task would run with only 90MB, but increasing again. The rate of memory loss depended on the work chunk frequency setting in the Milkyway preferences.


At least some of what you are seeing as a 'leak' is by design, as the unit progresses, and finds things, it takes more memory to process that data then at the start of the unit. If it goes back to a checkpoint it should then build up the memory usage again to a similar place it was before.


No, that's not it. A memory leak occurs when a program stores something but doesn't need or use it. All tasks through BOINC should checkpoint to the harddrive at regular intervals to save processed data, avoiding this problem.
ID: 62173 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
floyd

Send message
Joined: 13 Sep 11
Posts: 17
Credit: 3,263,835
RAC: 0
Message 62179 - Posted: 16 Aug 2014, 10:51:16 UTC

Some quick tests show only NVIDIA WUs affected in my case, not CPU or ATI, so this is not by design. Next question, is there a bug in the driver or in the application? Just in case anyone cares, this is for milkyway_separation_1.02_x86_64-pc-linux-gnu__opencl_nvidia and driver version 304.117. No, I'm not installing a later driver, I have no relevant issues with this one.
ID: 62179 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3315
Credit: 519,949,340
RAC: 22,061
Message 62180 - Posted: 16 Aug 2014, 11:18:10 UTC - in response to Message 62173.  

I'm not running Milkyway now but I'm sure there was a memory leak when I last ran Milkyway Separation (not Modified Fit) OpenCL tasks, to the extent where a single task used the whole 2GB I allowed for BOINC and other tasks were suspended for lack of memory. When restarted from a checkpoint, the same task would run with only 90MB, but increasing again. The rate of memory loss depended on the work chunk frequency setting in the Milkyway preferences.


At least some of what you are seeing as a 'leak' is by design, as the unit progresses, and finds things, it takes more memory to process that data then at the start of the unit. If it goes back to a checkpoint it should then build up the memory usage again to a similar place it was before.


No, that's not it. A memory leak occurs when a program stores something but doesn't need or use it. All tasks through BOINC should checkpoint to the harddrive at regular intervals to save processed data, avoiding this problem.


Unfortunately that is NOT the way it works at all projects, I guess some programmers have different agendas. Checkpointing can take time away from crunching by adding lines to the coding, some programmers don't do that, and some are told not to do that.
ID: 62180 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3315
Credit: 519,949,340
RAC: 22,061
Message 62181 - Posted: 16 Aug 2014, 11:22:52 UTC - in response to Message 62179.  

Some quick tests show only NVIDIA WUs affected in my case, not CPU or ATI, so this is not by design.


The coding for the AMD gpu's is NOT the same as the coding for the Nvidia gpu's, Nvidia actually helped Boinc in the beginning while AMD told Boinc 'we are not interested'. That means each project gets their own programmers and does their own coding for both Nvidia and AMD gpu's and some are much better then others. Since Nvidia provided some support, and may still, it tends to be easier and more programmers know that, the AMD stuff is mostly learn as you go and much fewer people are really good at it.

Next question, is there a bug in the driver or in the application? Just in case anyone cares, this is for milkyway_separation_1.02_x86_64-pc-linux-gnu__opencl_nvidia and driver version 304.117. No, I'm not installing a later driver, I have no relevant issues with this one.


Sorry I cannot help you there.
ID: 62181 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
floyd

Send message
Joined: 13 Sep 11
Posts: 17
Credit: 3,263,835
RAC: 0
Message 62183 - Posted: 16 Aug 2014, 11:53:14 UTC - in response to Message 62181.  

Next question, is there a bug in the driver or in the application? Just in case anyone cares, this is for milkyway_separation_1.02_x86_64-pc-linux-gnu__opencl_nvidia and driver version 304.117. No, I'm not installing a later driver, I have no relevant issues with this one.


Sorry I cannot help you there.


Thanks mikey. Actually I don't need help with this as I'm currently not running Milkyway anyway. Just wanted to share some information.
ID: 62183 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Jake Weiss
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 25 Feb 13
Posts: 580
Credit: 94,200,158
RAC: 0
Message 62199 - Posted: 18 Aug 2014, 16:24:36 UTC

To clear up some confusion, both the Nvidia and AMD GPUs run off of the same code in our application (except for one really old version for really old AMD GPUs). All of this is coded in a cross compatible language (OpenCL). Granted how this language compiles down to machine code at run time by the OpenCL interpreter on your device may vary depending on your GPU.

Additionally, we have both of our applications (separation and nbody) built with the ability checkpoint. To balance out the lost computation time from checkpointing we try time limit the frequency of checkpointing to once every few minutes.

As to the reported memory leak, it will not hurt your computer. If it exists, it is probably just memory not being deleted properly by our code before closing. Luckily, when the application actually closes, your operating system will automatically delete everything allocated for the program anyway so there is nothing to worry about. I will take some time to look into whether it exists later this week and if it does try to fix it.

Jake W.
ID: 62199 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : Memory Leak

©2024 Astroinformatics Group