Message boards :
Number crunching :
Yafu 16t
Message board moderation
Author | Message |
---|---|
fzs600 Send message Joined: 2 Sep 11 Posts: 7 Credit: 25,785,325 RAC: 4 |
Hello Therefore I will never be sent an YAFU-16t Version What should I do to receive? thank you |
yoyo_rkn Volunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message Joined: 22 Aug 11 Posts: 736 Credit: 17,612,101 RAC: 76 |
You should have at least 16 cores. |
fzs600 Send message Joined: 2 Sep 11 Posts: 7 Credit: 25,785,325 RAC: 4 |
You should have at least 16 cores. OK. thank you |
firstomega Send message Joined: 19 Apr 12 Posts: 1 Credit: 12,829,493 RAC: 1,455 |
You can also run it with fewer cores, but it is not really suitable when also running other CPU pojects. You need to change cc_config.xml to 16 cores with the following: <options> <ncpus>16</ncpus> </options> Then you must create an app_config.xml file in the projects folder of yafu with the following: <app_config> <app_version> <app_name>yafu-16t</app_name> <plan_class>16t</plan_class> <avg_ncpus>16</avg_ncpus> <cmdline>--nthreads 8</cmdline> </app_version> </app_config> The 16 is what BOINC will think it uses and 8 are the real cores of the machine. You have also to realise that the deadline may be to short when using not enough cores. |
scole of TSBT Send message Joined: 25 Sep 14 Posts: 2 Credit: 83,225,031 RAC: 1 |
EDIT: Never mind. The system running the 16t WU appears to be using 16 cores now. |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
My two 16t work units are using just a single core each. I did tell BOINC that I had 16 cores but in a cc_config.xml file but this made no difference. I have just added the app_config.xml file posted above. However I am unable to restart BOINC Client at this stage to make it read the file due to me running a QMC work unit. Despite it having checkpointing, this appears to not be working as it will start from 0.00 if I stop BOINC for any reason. It is currently at 33 hours with around 50 still to go so I will have to wait. The yafu work units also went back to 0.00 along with the QMC WU when my PSU gave a loud POP and started issuing smoke 2 days ago. They are also at 33 hours and about a quarter done. Conan |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
My two 16t work units are using just a single core each. Well that QMC job was telling fibs, it said it was at 33% but it was not, when it passed 34 hours run time it moved to 1% completed. With that first checkpoint I have been able to restart BOINC Client and it has read the app_config.xml from firstomega and is now running multiple cores as it should of been doing in the first place. After 21 min 11 sec it has moved to 0.312% completed, so we will see how it goes (it restarted again from 0.00 after the BOINC restart). Conan |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
My two 16t work units are using just a single core each. After almost 20.5 hours the WU is now at a bit over 16.5%. So progressing along but it is going to take at least 100 hours to finish this, they are long work units. What is the normal run time for these 16t's? Conan |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
Well the work unit finished after a run time of 335205 seconds and a cpu time of 874320.70 seconds. This says it ran for 93 hours and received 2096 credits. As it did not run this long when I finally got all the cores working on the problem it must of been adding up all the time it ran when using only a single core, which was about 30 hours then restart, 30 hours then restart and finally another 30 hours with all cores. As such the credit turned out to be miserable at just 2 cr/h counting the CPU time and 22 cr/h counting the RUN time. I will see how the second work unit runs using all cores from the start, hopefully a bit beter run time and therefore credit. Conan |
yoyo_rkn Volunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message Joined: 22 Aug 11 Posts: 736 Credit: 17,612,101 RAC: 76 |
Many of the 16t workunits are running short. But there are some which needs to run over the full time. I would assume that they run roughly 25h on 16 cores. yoyo |
marmot Send message Joined: 5 Nov 15 Posts: 33 Credit: 53,531,496 RAC: 0 |
Are they all going to be that low? Maybe I should terminate these 2 16t's and go back to 8t? |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
Just finished this 16t work unit after 14 days WU 715200 It ran the whole time in single CPU mode and would not switch to to use 16 cores, which is why it took so long. Conan |
yoyo_rkn Volunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message Joined: 22 Aug 11 Posts: 736 Credit: 17,612,101 RAC: 76 |
Thanks for finishing a 16t task. Those are realy needed and not many user are running them. It's curious that this workunit runs only on a single core the whole time. I agree, there are phases where it runs only on a single core, especially at the end. But his shouldn't happen the whole time if you don't play arround with boinc settings and app_config. yoyo |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
I have another 16t work unit and it also is running on just the single core (over 3 days so far). I run Amicable Numbers and it runs on the number of cores set for it (was 16 but have reduced to 8). When running Milkyway it also uses the full 16 cores. So I don't know why YAFU wont use the cores I have, nothing is at high priority which usually will stop a "mt" WU from running. It is only the 16t wu's as the 8t and 4t work units on this host run the required number of cores. Conan |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
I have another 16t work unit and it also is running on just the single core (over 3 days so far). This one finally finished after about 13 days WU 728864. Again it ran just in single mode and would not use all the available cores. Conan |
marmot Send message Joined: 5 Nov 15 Posts: 33 Credit: 53,531,496 RAC: 0 |
I have another 16t work unit and it also is running on just the single core (over 3 days so far). All the WU types (MT, 4T, etc) have a consolidation period in between ECM and GNFS sections where YAFU runs as a solo process with 1 to 16 cores. At the very end, YAFU.exe runs in single thread, NCI mode with high disk usage as it compiles the data into it's final result. Maybe it's possible that every time you were viewing the WU that it was in consolidation mode? The consolidation can take 15 minutes to an hour on the 16T and the final database collection can take some hours also. If that's not what happened, check your app_config.xml for typos. It's always a typo that gets me so I switched to Notepad++ and it color codes the tags and turns them red when the tags don't match up. (BTW, the credit you received for that WU was horrible!) |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
G'Day marmot, The work unit only ever ran on one cpu. This I know as the other 15 cpu's were all running other projects and did so for the nearly 2 weeks that the work unit ran. I don't have an "app_config.xml" file, don't recall ever having one or ever creating one, for any project (also not too sure what to put in one either, even using the wiki as a guide). The 16t work units have run OK on this computer before but the last 2 that have run have both only used a single core. Originally only 1 cpu core was used till I changed "cc_config.xml" to use all 16 then things ran OK for awhile. 4t and 8t work units run without issue and use the required number of cores (can't recall if normal YAFU uses all 16 cores or not, as it has been awhile since I have had one). Conan |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
G'Day marmot, Have to take back the bit about not ever having an app_config.xml file as I copied one earlier in this thread, but it seems to have not worked and I no longer have it. Other projects don't need one so why here? I have another 16t download, again only using a single core. Conan |
Conan Send message Joined: 5 Sep 11 Posts: 46 Credit: 7,351,043 RAC: 2,864 |
G'Day marmot, Anyone know how to get these things to use more than 1 core? 4t and 8t work fine but 16t does not. Conan |
PAK-FA Send message Joined: 19 Oct 15 Posts: 4 Credit: 10,875,085 RAC: 0 |
I can confirm that I'm also receiving 16t tasks that running only on 1 thread, may be all 16t tasks have this issue? may be some program fix needed? |