University of Maryland Mike P. Cummings  
Center for Bioinformatics and Computational Biology
HomeResearchPublicationsPersonnel

The Lattice Project
About Lattice
Applications
Client Activity
Create Account
Message Boards
Participant Profiles
Questions & Answers
Research Projects
Rules and Policies
Statistics
Teams
Top Computers
Top Participants
Top Teams
Your Account

BOINC Logo



Forum Thread

New Work
log in

Advanced search

Message boards : News : New Work

1 · 2 · Next
Author Message
Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4445 - Posted: 22 Sep 2010, 18:15:51 UTC

Some new work that requires at least 3G of memory (RAM) is slowly becoming available. The runtime on these jobs is uncertain; a lower bound is perhaps 15 hours, an upper bound in the hundreds of hours for slower machines. We'll learn more as we go along, but if the deadline seems too short, don't panic! We're trying to improve our accuracy but have a ways to go yet.

Trotador
Send message
Joined: 7 Aug 09
Posts: 66
Credit: 4,172,231
RAC: 0
Message 4446 - Posted: 22 Sep 2010, 19:09:25 UTC

I've got one, it errored in 15 seconds like two previous instances.

http://boinc.umiacs.umd.edu/workunit.php?wuid=1263196

Anyway, 3Gb RAM, LOL, is it checking the available RAM? My computer has 4 GB RAM but it is XP and besides there are other three boincs projects running...so it not just physical RAM, but of course you know it.

Trotador
Send message
Joined: 7 Aug 09
Posts: 66
Credit: 4,172,231
RAC: 0
Message 4447 - Posted: 22 Sep 2010, 19:14:15 UTC

Second one in Linux is progressing so far..

Have these ones checkpoints?

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4448 - Posted: 22 Sep 2010, 19:14:17 UTC - in response to Message 4446.

You're right, I'm seeing a lot of out of memory errors. Maybe for the next batch I'll increase it to 4G or 4.5G and see if the errors go down. Sorry about that...

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4449 - Posted: 22 Sep 2010, 19:14:42 UTC - in response to Message 4447.

Second one in Linux is progressing so far..

Have these ones checkpoints?


Yes, they should checkpoint.

dividedbymyself
Send message
Joined: 22 Jul 08
Posts: 3
Credit: 25,858
RAC: 0
Message 4461 - Posted: 23 Sep 2010, 10:08:08 UTC

Why a deadline of only a few days when you estimate some of them can take hundreds of hours?

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4462 - Posted: 23 Sep 2010, 14:23:32 UTC - in response to Message 4461.

The estimates are being computer generated right now; it's possible mine is no better than the computer's. The deadlines should eventually get better. For now, we hope they're decent, but if they seem to be consistently too short or too long we'll make adjustments.

Trotador
Send message
Joined: 7 Aug 09
Posts: 66
Credit: 4,172,231
RAC: 0
Message 4464 - Posted: 23 Sep 2010, 18:16:49 UTC

Whe I've got home I've foud this one cancelled by server, the one in Linux I said yesterday was progressing nicely or that seemed

http://boinc.umiacs.umd.edu/workunit.php?wuid=1263213

why?

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4465 - Posted: 23 Sep 2010, 19:36:00 UTC - in response to Message 4464.

Sorry, I went through and canceled all the troublesome workunits from yesterday, primarily because I wanted Mac and Linux clients to be able to focus on the new work (with updated settings). Hope you understand there's a bit of experimentation going on, but you should be able to crunch this new work now without any fear (and many thanks)

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4466 - Posted: 23 Sep 2010, 19:38:49 UTC

Just an update: the reason for all the errors yesterday is because it was hitting up against the Windows 32-bit per-process memory limit, which is somewhere in the vicinity of 2G. The solution is to create a 64-bit version of GARLI for BOINC - while I'm working on that I'll have to exclude Windows machines from these large memory jobs, sorry...

Ironworker16
Avatar
Send message
Joined: 3 Jul 05
Posts: 79
Credit: 1,871,962
RAC: 164
Message 4468 - Posted: 23 Sep 2010, 21:55:49 UTC - in response to Message 4466.
Last modified: 23 Sep 2010, 21:59:32 UTC

I'm looking forward to a 64-bit version of GARLI for BOINC. I did not realize I was limit to 2G. I curious will the 64 bit ver. crunch faster than the 32 bit ver. I also have another question I have 12G of ram and a i7 processor and normally crunch 8 work units at a time so if the work units use x amount of memory let’s say 3g I would be limit to 3 maybe 4 work units. Will boinc manager try to use more memory than I have installed or will it limit its self. If so do you have to flag the units so that boinc knows the memory requirements?
____________

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4472 - Posted: 24 Sep 2010, 4:46:10 UTC - in response to Message 4468.

I'll take a crack at your questions; others, feel free.

I'm looking forward to a 64-bit version of GARLI for BOINC. I did not realize I was limit to 2G. I curious will the 64 bit ver. crunch faster than the 32 bit ver.


I'm not sure, but certainly GARLI itself runs faster with more memory, up to a point, for most data sets.

I also have another question I have 12G of ram and a i7 processor and normally crunch 8 work units at a time so if the work units use x amount of memory let’s say 3g I would be limit to 3 maybe 4 work units.


Yes, I would think so.

Will boinc manager try to use more memory than I have installed or will it limit its self. If so do you have to flag the units so that boinc knows the memory requirements?


I don't think you need to worry about this.

Pepo
Avatar
Send message
Joined: 16 Nov 07
Posts: 33
Credit: 13,982
RAC: 0
Message 4474 - Posted: 28 Sep 2010, 9:12:49 UTC - in response to Message 4472.

Adam Bazinet wrote:
I'll take a crack at your questions; others, feel free.

Ironworker16 wrote:
I also have another question I have 12G of ram and a i7 processor and normally crunch 8 work units at a time so if the work units use x amount of memory let’s say 3g I would be limit to 3 maybe 4 work units.

Yes, I would think so.


Ironworker16 wrote:
Will boinc manager try to use more memory than I have installed or will it limit its self. If so do you have to flag the units so that boinc knows the memory requirements?

I don't think you need to worry about this.


As I know BOINC, I'm expecting to see memory problems here and am especially curious to see, how it will behave on such host (especially if it would crunch for multiple projects, but with significant Resource Share towards Lattice, what is probably not the case of Ironworker16's machine).

Ironworker16
Avatar
Send message
Joined: 3 Jul 05
Posts: 79
Credit: 1,871,962
RAC: 164
Message 4476 - Posted: 28 Sep 2010, 21:31:01 UTC - in response to Message 4466.

The solution is to create a 64-bit version of GARLI for BOINC - while I'm working on that I'll have to exclude Windows machines from these large memory jobs, sorry...


I was wondering what's the status of the 64 bit verison? Also if you need to run some test work units I'm ready.
____________

Profile Adam Bazinet
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 18 Feb 05
Posts: 1448
Credit: 334,567
RAC: 1
Message 4477 - Posted: 28 Sep 2010, 22:56:09 UTC - in response to Message 4476.

The status is I haven't started on it yet, I haven't had a chance. But I'll keep you posted...

Professor Ray
Send message
Joined: 15 Feb 08
Posts: 61
Credit: 24,874
RAC: 28
Message 4512 - Posted: 14 Oct 2010, 16:21:48 UTC
Last modified: 14 Oct 2010, 16:27:04 UTC

Memory Limits for Windows Releases

Address space limits for 32-bit Windows:

User-mode virtual address space for each 32-bit process: 2 GB; Up to 3 GB with IMAGE_FILE_LARGE_ADDRESS_AWARE and "4GT"

Kernel-mode virtual address space: from 1 GB to a maximum of 2 GB with "4GT"

Paged pool, Nonpaged pool, System cache virtual address spaces are limited by available kernel-mode virtual address space or constrained by registry key values for the respective memory type. The physical size of either Nonpaged pool and System cache are limited by physical memory.

Address space limits for 64-bit Windows:

User-mode virtual address space for each 32-bit process: 2 GB (default); 4 GB w/ IMAGE_FILE_LARGE_ADDRESS_AWARE set

Kernel-mode virtual address space: 8 TB

Succinctly put: If Lattice is amenable to a minimum of 3GB of user-mode VA (with the appropriate OS memory configuration), then Lattice WU's should be processable on either 32 or 64-bit Windows systems. IF 3GB is insufficint user-mode VA, then such WU's can ONLY be processed on 64-bit windows. However, if 4GB is insufficient, then Lattice needs to be developed for native 64-bit execution. The foregoing notwithstanding, it doesn't address the user-mode requirments for multi-cored CPUs running multiple instances of Lattice concurrently (nor system VA requirements that non-BOINC processing may entail); determining whether system VA space is being exhausted requires the use of a kernel debugger.

Ironworker16
Avatar
Send message
Joined: 3 Jul 05
Posts: 79
Credit: 1,871,962
RAC: 164
Message 4513 - Posted: 14 Oct 2010, 23:22:10 UTC - in response to Message 4512.

Ray if you look at this thread GARLI v5.01 you will see there is a windows 64 bit ver.
____________

Professor Ray
Send message
Joined: 15 Feb 08
Posts: 61
Credit: 24,874
RAC: 28
Message 4514 - Posted: 15 Oct 2010, 10:18:25 UTC

Oh my bad. I didn't realize I was posting to an ancient thread. Looks like he got that worked out in the last three weeks.

Profile Michele ANSALDI
Send message
Joined: 4 Jun 07
Posts: 3
Credit: 60,253
RAC: 0
Message 4598 - Posted: 11 Nov 2010, 15:29:40 UTC

This is my desktop running only Boinc 24 hours:

09/11/2010 23:24:39 | | Starting BOINC client version 6.12.4 for windows_x86_64
09/11/2010 23:24:39 | | log flags: file_xfer, sched_ops, task
09/11/2010 23:24:39 | | Libraries: libcurl/7.19.7 OpenSSL/0.9.8l zlib/1.2.5
09/11/2010 23:24:39 | | Data directory: F:\Boinc\ProgramData
09/11/2010 23:24:39 | | Running under account Michele ANSALDI
09/11/2010 23:24:39 | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7 CPU 975 @ 3.33GHz [Family 6 Model 26 Stepping 5]
09/11/2010 23:24:39 | | Processor: 256.00 KB cache
09/11/2010 23:24:39 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 syscall nx lm vmx tm2 popcnt pbe
09/11/2010 23:24:39 | | OS: Microsoft Windows 7: Ultimate x64 Edition, (06.01.7600.00)
09/11/2010 23:24:39 | | Memory: 11.99 GB physical, 23.98 GB virtual
09/11/2010 23:24:39 | | Disk: 465.76 GB total, 461.71 GB free
09/11/2010 23:24:39 | | Local time is UTC +1 hours
09/11/2010 23:24:39 | | NVIDIA GPU 0: GeForce GTX 285 (driver version unknown, CUDA version 3020, compute capability 1.3, 998MB, 708 GFLOPS peak)
09/11/2010 23:24:39 | | ATI GPU 0: ATI Radeon HD5800 series (Cypress) (CAL version 1.4.838, 2048MB, 2720 GFLOPS peak).

The Cpu is OC at 3.915 GHz so I think that is fast.
For a long period I abort the wu's that I catch (sometimes after 150 hours) because the deadline was too short. The 11/04 I choosed to see what happened if I finished a wu and so the 11/11 this was the result: http://boinc.umiacs.umd.edu/result.php?resultid=2999880 after 293h 35'.
So I know now that the wu's can be finished. I'm at the 98,615% of the result 2999879 after 294h 45' now.
I think that you must explain that the deadline is not important in a better way or set longest deadlines.
____________

Trotador
Send message
Joined: 7 Aug 09
Posts: 66
Credit: 4,172,231
RAC: 0
Message 4737 - Posted: 5 Feb 2011, 13:24:14 UTC

Any work for today Adam?

thanks,

1 · 2 · Next
Post to thread

Message boards : News : New Work

.......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... .......... ..........

University of Maryland     UM Home | Directories | Search | Admissions | Calendar
Copyright © 2017 The Lattice Project
Direct questions and comments to Lattice Admin