Message boards :
Number crunching :
app_config.xml - the bit
Message board moderation
Author | Message |
---|---|
Guy Send message Joined: 17 May 99 Posts: 69 Credit: 8,183,090 RAC: 0 |
Hello, I'm looking for explicit examples of how to use <app_version> in app_config.xml I have two HD6950's but only a dual core CPU. So in the interest of 'keeping one CPU core free' I need to control which CPU apps run depending upon which GPU apps are running. Is <app_version> even the right approach for this fine control? Anyone help please? Thanks. Guy |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
You could try a diferent aproch, if you know how to manualy install the builds. There are a new set of AP optimized app who practicaly uses almost no CPU to do the GPU work (like the MB GPU job), there are avaiable at Mikes site: http://mikesworldnet.de/download.html This builds combined with some tweaks on the app_config.xml file allow you a lot more control of what you crunch and how. Of couse YMMV since each hosts is unique but with them running you actualy will see your 2 GPU´s could be feeded with less than one core leaving the second core free to do other jobs even when you crunch AP. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Hello, Depending on exactly what you want to do the app_config.xml can help get you there. gpu_usage: The number of GPU instances (possibly fractional) used by GPU versions of this app. <gpu_usage>1</gpu_usage> = 1 per GPU, <gpu_usage>0.50</gpu_usage> = 2 per GPU, <gpu_usage>0.33</gpu_usage> = 3 per GPU. cpu_usage: The number of CPU instances (possibly fractional) used by GPU versions of this app. <cpu_usage>1</cpu_usage> = use/reserve 1 CPU/core, <cpu_usage>0.50</cpu_usage> = use/reserve half a CPU/core. So 2 instance would be 1 core I'm not sure how many instances a 6950 can handle, but this would run 2 MB or 2 AP per GPU & reserve a CPU core per AP instance. <app_config> <app> <name>setiathome_v7</name> <gpu_versions> <gpu_usage>0.50</gpu_usage> <cpu_usage>0.04</cpu_usage> </gpu_versions> </app> <app> <name>astropulse_v6</name> <gpu_versions> <gpu_usage>0.50</gpu_usage> <cpu_usage>1</cpu_usage> </gpu_versions> </app> </app_config> SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Cruncher-American Send message Joined: 25 Mar 02 Posts: 1513 Credit: 370,893,186 RAC: 340 |
Since you want to use only 1 CPU core max, you should also use the <max_concurrent> parameter in app_config to limit the (possible) concurrent GPU APs to 1: <max_concurrent>1</max_concurrent> |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
If you use the newer builds (r2180) you could try something like this and run 2 AP or MB WU at the same time on each one of your GPU´s and still leaving one of your 2 cores free. Be aware don't try to do use this file with AP stock cruch, AP stock need 1 core per crunched WU - <cpu_usage>1</cpu_usage> <app_config> <app> <name>setiathome_v7</name> <gpu_versions> <gpu_usage>0.5</gpu_usage> <cpu_usage>0.1</cpu_usage> </gpu_versions> </app> <app> <name>astropulse_v6</name> <gpu_versions> <gpu_usage>0.5</gpu_usage> <cpu_usage>0.1</cpu_usage> </gpu_versions> </app> </app_config> You could see on this host for example http://setiathome.berkeley.edu/show_host_detail.php?hostid=6269362 it runs 8 AP WU at the same time and it´s GPU´s are feeded by one core of an old and slow I5 with only 4 cores, leaving the other 3 cores free to do other jobs. |
petri33 Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 |
If I had 2 cores I'd definitely try HAL9000's instructions. MB GPU tasks would crunch happily along with CPU and whenever there would be AP tasks they would get automagically one CPU core per AP task. If your GPU is more happy with 2 AP's then You could try cpu_usage 0.5 for the AP section. One core can service 2 AP's or more especially with the newer builds. Hello, To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
Mike Send message Joined: 17 Feb 01 Posts: 34380 Credit: 79,922,639 RAC: 80 |
Running 2 instances of AP is risky with only 2 cores. If you get 2 high blanked units one will go into stall on a AMD CPU because it cant feed the GPU anymore.. With each crime and every kindness we birth our future. |
Guy Send message Joined: 17 May 99 Posts: 69 Credit: 8,183,090 RAC: 0 |
Thanks Guy. I'm kinda relieved to hear <app_version> isn't the way... For your interest; my HD6950's crunch 3 x 7.03 each quite happily while a 7.00 CPU app is crunching as well. The curious thing about this set-up is that 75% of my CPU cores are utilised - despite having set the local client to use 50%, or one core. Well, at least I know I'm not going to be barking up the wrong tree. So thanks. And thanks for other answers/suggestions tendered. It's a lot to digest but I'm still looking for a 'breakthrough' for when the 6.06 apps start up! Or a rare 6.01 CPU app!! If I 'balance' my set-up I'll post. Thank you all. Guy |
OzzFan Send message Joined: 9 Apr 02 Posts: 15691 Credit: 84,761,841 RAC: 28 |
The curious thing about this set-up is that 75% of my CPU cores are utilised - despite having set the local client to use 50%, or one core. The CPU-limiting preference setting is for CPU apps only. Part of a CPU will always be used to "feed" the GPU(s). |
Guy Send message Joined: 17 May 99 Posts: 69 Credit: 8,183,090 RAC: 0 |
Ah! Thanks. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.