|
21)
Message boards :
BOINC client :
Benchmarks. Linux v Windows
(Message 7525)
Posted 16 Jan 2007 by ohiomike Post: With out repeating my whole rant- http://boinc.berkeley.edu/dev/forum_thread.php?id=1382 from the Manager forum. There does need to be some work done on the benchmarking code. The thread/priority handling is much different between windows and Linux. I think if we launched the benchmarks as a free-running task (not a thread) and let them run for a set period and report (ie- start and run for 30 seconds then dump results to a file and suicide) that we might get better results. One of the biggest problems with running a benchmark from the client is that the client wants to be low priority, where as the science app wants to be as high of a priority as can be run without affecting user interaction. Windows and Linux once again handle this differently. What priority do we want to give to the benchmark- full speed (no other tasks) or something approximating "real-world" usage. And what is "real-world" usage? It is much difference between this machine I use for email, etc and my Linux development machine, which generally just crunches 20 hours a day or so. |
|
22)
Message boards :
BOINC Manager :
BOINC_5.5.0_SSE2
(Message 7524)
Posted 16 Jan 2007 by ohiomike Post: Lets keep in mind 1) That an over-clocked machine often gets cheated- It finishes to quickly and the slower machines win the vote. 2) All optimized apps are not just increasing benchmark scores- compiling the SETI app with the Intel ICC compiler, using in the performance math libs cuts the WU processing time by better than 30% in my experience. 3) The benchmark scores are fairly bogus. A) The Windows version starts the benchmark thread running and then drops it's own priority. The Linux version uses a "fork" command to start the thread which gives it the same priority as the parent. B) The Windows clock that they are using is not that good, it rounds everything off to the nearest 30 mSec period, where as Linux is about 1 mSec in most cases. C) I you compare the Boinc benchmark scores to Sandra (Windows) or UnixBench (Linux) you will find all sorts of different answers to what the computing power is. 4) That being said, I did download an optimized 5.9.0 Client from the web that reported crazy high (2x-3x) benchmark scores. If that is what the scores should be , fine, but they are "out of line" with the main-stream so far that I deleted the apps and built my own, optimizing the compile of the Berkeley code. That reports just slightly higher than the "stock" version, which is what I would expect. |
|
23)
Message boards :
BOINC Manager :
Question about V5.9.0 build
(Message 7522)
Posted 16 Jan 2007 by ohiomike Post: Thanks, I'll give that a try. PS- The client appears to be working nicely on Linux (over 24 hours so far, I just built the client not the Manager). |
|
24)
Message boards :
BOINC Manager :
Question about V5.9.0 build
(Message 7491)
Posted 15 Jan 2007 by ohiomike Post: Anyone know why I am getting an undefined reference to 'CViewProjectsGrid::CViewProjectsGrid(wxNotebook)' when I try to compile the boinc client version 5.9.0? Is this a problem with wxWidgets-2.6.3? I'm getting to tired to track it down tonight. |
|
25)
Message boards :
BOINC client :
Ubuntu Init Script
(Message 7356)
Posted 9 Jan 2007 by ohiomike Post: It is better to start the task using "nice". This will cause all client tasks to also be run in the nice mode. That way it only hogs resources when they are not needed else where. Un-attended both my CPUs peg on 100%, but when I am doing something, the tasks back-off to allow good user performance. See "man nice". |
|
26)
Message boards :
BOINC client :
Will someone Please fix the Linux client benchmark!
(Message 7132)
Posted 26 Dec 2006 by ohiomike Post: This gets irritating if you run several different computers/OS's. If I set preferences to get enough files to keep the Linux machines running (twice the buffer it would require if the client would calculate the estimated time correctly), my windows machines get buried in tasks. I know I can "work around this" by declaring different locations and preferences for the Windows and Linux machines, but the question is- Why should I have to??????? It seems that everyone knows that the Linux benchmark is about 1/2 of the real operating speed, is it that hard to fix? |
Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.