climateprediction.net home page
Windows Crashers -- Were you Looking At Graphics?

Windows Crashers -- Were you Looking At Graphics?

Questions and Answers : Windows : Windows Crashers -- Were you Looking At Graphics?
Message board moderation

To post messages, you must log in.

AuthorMessage
old_user1
Avatar

Send message
Joined: 5 Aug 04
Posts: 907
Credit: 299,864
RAC: 0
Message 5963 - Posted: 8 Nov 2004, 10:36:43 UTC
Last modified: 8 Nov 2004, 10:36:52 UTC

I'm beginning to suspect a lot of "-5 errors" could be from the graphics crashing, which may actually leave an "orphan" hadsm3um_* running, which would in turn crash future launches of hadsm3_*. This is because in BOINC the main program does everything - the launching of a model, the graphics etc. So if there is a graphics problem that would crash everything. I'm going to look into this a little more.

ID: 5963 · Report as offensive     Reply Quote
old_user27795

Send message
Joined: 30 Oct 04
Posts: 4
Credit: 0
RAC: 0
Message 5979 - Posted: 8 Nov 2004, 21:51:35 UTC

Hi, The memory failure I had, at least one of them I do remember something about graphics routine, but at the time I was not looking at it. However in the last statement, I may have looked or displayed it for my wife at an earlier time and I wonder if your closing routine or memory release never completes and hangs in background till machine reaches at near memory overload condition and window crashes your primary module as owner of memory block????
ID: 5979 · Report as offensive     Reply Quote
Jord
Avatar

Send message
Joined: 5 Aug 04
Posts: 250
Credit: 93,274
RAC: 0
Message 5989 - Posted: 9 Nov 2004, 9:40:16 UTC

As I said in my email, Carl, I don't use the screensaver and hardly ever look at the graphics. Yet two models crashing in a row and leaving their oprhan running in the background? Also, both models crashed during the weekend, so I would only find out after 26 hours or more.

It's hardly a coincidence, that's why I emailed you, to ask if there were any known problems. :)

But even if the problem would be the graphics, wouldn't it be wise if we told you what videocard we use, which drivers and DirectX verion (if applicable)?
Ati 9600 Pro 128MB (AGP 4x), Catalysts 4.10 and DirectX9c.

I did notice that both hadsm3um and hadsm3um_w are always running. With one of them being the graphics representation, why is it always pre-loaded? Does this also mean it's always running/updating?

I am using the method of unloading the cruncher out of memory when switching to another cruncher, for CPDN, Seti and LHC.

--------------------------------
Jordâ„¢

<img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=2&amp;trans=off">
ID: 5989 · Report as offensive     Reply Quote
old_user27795

Send message
Joined: 30 Oct 04
Posts: 4
Credit: 0
RAC: 0
Message 6001 - Posted: 10 Nov 2004, 4:07:45 UTC

Carl,
I beleive I have found the underlying problem causing your internal crashes. I notice earlier tonight that the system I was running had a high amount of fragmentation, (windows-ME, 768K, 8GIG disk,1.3G CPU), I shutdown all programs and went to another O/S on same system compressed the disk and started up again, I ran for two hours and looked at the disk again from the second O/S and found another 1800 fragments all for CPDN project of various sizes. The system becomes very un-predictable in its operation as the fragments start to crawl over system data area on the disk. YOUR DISK MANAGEMENT ROUTINES NEED FIXING, stats for project environment are , BOINC 4.13, CPDN, 4.04, SETI 4.07, only CPDN produces fragments of this level and also cause system and CPDN stability problems.

David Sr.
ID: 6001 · Report as offensive     Reply Quote
Joe

Send message
Joined: 3 Sep 04
Posts: 5
Credit: 909,025
RAC: 191
Message 6002 - Posted: 10 Nov 2004, 6:21:19 UTC

I took a look at a disk in W2000pro with the file analyzer got the following results:

Volume (D:):
Volume size = 11,530 MB
Cluster size = 4 KB
Used space = 4,761 MB
Free space = 6,769 MB
Percent free space = 58 %

Volume fragmentation
Total fragmentation = 18 %
File fragmentation = 36 %
Free space fragmentation = 0 %

File fragmentation
Total files = 38,159
Average file size = 148 KB
Total fragmented files = 1,577
Total excess fragments = 17,813
Average fragments per file = 1.46

Pagefile fragmentation
Pagefile size = 768 MB
Total fragments = 1

Directory fragmentation
Total directories = 1,952
Fragmented directories = 165
Excess directory fragments = 1,957

Master File Table (MFT) fragmentation
Total MFT size = 41,991 KB
MFT record count = 40,166
Percent MFT in use = 95 %
Total MFT fragments = 2

--------------------------------------------------------------------------------
Fragments File Size Most fragmented files
115 8,136 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.ph13c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.pd13c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd14c10
111 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph14c10
117 8,136 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.ph17c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.pd14c10
115 8,136 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.ph14c10
123 5,000 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.pd17c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd15c10
124 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph15c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd16c10
110 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph16c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd17c10
126 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph17c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd18c10
118 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph18c10
126 8,136 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.ph15c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.pd15c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.pd16c10
132 8,136 KB \Program Files\BOINC\projects\climateprediction.net\3blq_100176566\dataout\3blqaa.ph16c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd21c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd19c10
115 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph19c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd22c10
122 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph22c10
123 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd20c10
117 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph20c10
118 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph21c10
122 5,000 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.pd23c10
119 8,136 KB \Program Files\BOINC\projects\climateprediction.net\2vq6_100155786\dataout\2vq6aa.ph23c10

It looks pretty uniform to me...
ID: 6002 · Report as offensive     Reply Quote
Jord
Avatar

Send message
Joined: 5 Aug 04
Posts: 250
Credit: 93,274
RAC: 0
Message 6006 - Posted: 10 Nov 2004, 11:41:26 UTC

FAT32 fragments earlier than NTFS anyway. I'm running my BOINC from its own partition, only use DIRMS to defrag once in a blue moon, so two CPDN models crashing within a week of eachother is everything but a fragmentation problem. A 2 phase scandisk on my whole system didn't bring up any problems either.


--------------------------------
Jordâ„¢

<img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=2&amp;trans=off">
ID: 6006 · Report as offensive     Reply Quote
old_user1
Avatar

Send message
Joined: 5 Aug 04
Posts: 907
Credit: 299,864
RAC: 0
Message 6009 - Posted: 10 Nov 2004, 14:05:00 UTC - in response to Message 6001.  
Last modified: 10 Nov 2004, 14:05:35 UTC

It could very well be disk fragmentation problems; it's out our hands though. The massive file I/O is done by the Met Office Fortran code. It would probably take 10 people a year to go through and clean things up (about a million lines). I guess they either don't get this problem or don't worry about it on their massive supercomputers.
ID: 6009 · Report as offensive     Reply Quote
old_user27795

Send message
Joined: 30 Oct 04
Posts: 4
Credit: 0
RAC: 0
Message 6013 - Posted: 10 Nov 2004, 23:33:43 UTC - in response to Message 6009.  

&gt; It could very well be disk fragmentation problems; it's out our hands though.
&gt; The massive file I/O is done by the Met Office Fortran code. It would
&gt; probably take 10 people a year to go through and clean things up (about a
&gt; million lines). I guess they either don't get this problem or don't worry
&gt; about it on their massive supercomputers.
&gt;
&gt;Thanks, my guess with looking is where they save the last (data,track loc. and memory), and the routine starts a new file for data at each interupt process..... I will not run this code until it gets fixed I had to rebuild the system from all the damage program area on the disk. I have detached the project for now.

good luck.
ID: 6013 · Report as offensive     Reply Quote
Profile old_user17289

Send message
Joined: 13 Sep 04
Posts: 228
Credit: 354,979
RAC: 0
Message 6015 - Posted: 11 Nov 2004, 3:41:22 UTC

Yes, it's true, CPDN does a bit of fragmenting. I just defagged my C: drive (XP-SP2, NTFS) and an hour later there are more than 50 fragments in the CPDN files.

BUT: I don't see much of a problem with this; fragmented files are normal on Windows, and I don't know of any limitiations. It may slow down IOs a bit (of the CPDN app), but that really doesn't worry me.
ID: 6015 · Report as offensive     Reply Quote

Questions and Answers : Windows : Windows Crashers -- Were you Looking At Graphics?

©2024 climateprediction.net