Message boards : Number crunching : Rosetta 4.1+ and 4.2+
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 34 · Next
Author | Message |
---|---|
bkil Send message Joined: 11 Jan 20 Posts: 97 Credit: 4,433,288 RAC: 0 |
You may try disabling i686 apps as mentioned here to get it up and running: https://boinc.bakerlab.org/rosetta/forum_thread.php?id=13727&postid=94049#94049 |
Michael E.@ team Carl Sagan Send message Joined: 5 Apr 08 Posts: 16 Credit: 1,947,072 RAC: 559 |
I am running Windows 10, BOINC 7.14.2 (x86), with Rosetta preferences at 1 day 12 hours. The Rosetta 4.12 tasks downloaded initially said they would execute in 6 hours but it is taking much longer. How much longer than the Deadline time is allowed? Sorry but I had to abort several of these. The Elapsed time seems to reset on some of these tasks. One task completed (1143119307.) and others had to be aborted. I just downloaded some Rosetta Mini v3.78 tasks and will see how those go. How can I help? I am a retired software writing/support guy who knows what native code means :-). Mike |
Mod.Sense Volunteer moderator Send message Joined: 22 Aug 06 Posts: 4018 Credit: 0 RAC: 0 |
As you first get started, BOINC Manager really hasn't had a chance to see the tasks run, and so the estimate to completion will not be very good. If you set to 36 hours, they will probably take about that long. No worries on deleting the extras, there are other machines clamoring for work. Don't count on any slack on the deadlines. The Elapsed time seems to reset on some of these tasks Do you have the box checked to leave tasks in memory while suspended? Rosetta Moderator: Mod.Sense |
Michael E.@ team Carl Sagan Send message Joined: 5 Apr 08 Posts: 16 Credit: 1,947,072 RAC: 559 |
Mod.Sense asked:"Do you have the box checked to leave tasks in memory while suspended?" On both local PCs, this is is checked/on: BOINC > Advanced View > Options > Computing Preferences > Disk and Memory > Leave non-GPU tasks in memory while suspended On the web pref's, it was unchecked so I checked it. I usually use the local pref's only. To minimize disk writes, I set the BOINC > Advanced View > Options > Computing Preferences > Computing > Request tasks to checkpoint at most every 300 seconds The two PCs both had similar unexpectedly low elapsed time for a short time after they had been processing for a while. They both returned one task so far. Mike |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1698 Credit: 18,172,091 RAC: 24,436 |
Mod.Sense asked:"Do you have the box checked to leave tasks in memory while suspended?"If you're going to have that checked, you might want to increase the amount of memory available to BOINC to avoid out of memory problems (it should make use of page file/virtual memory, but why risk it?). Memory When computer is in use, use at most 95 % When computer is not in use, use at most 95 % Page/swap file: use at most 75 % On the web pref's, it was unchecked so I checked it. I usually use the local pref's only.Local preferences do override the web based ones. To minimize disk writes, I set the BOINC > Advanced View > Options > Computing Preferences > Computing > Request tasks to checkpoint at most every 300 secondsUnnecessary, and will make missing a checkpoint much more likely. With it set to 60 seconds, it doesn't mean it will checkpoint then- it will just ask the application to checkpoint, if it can. If it can, it will. If it can't, it won't. And as for minimising disk writes, that too is unnecessary (unless you are running it on a thumb drive), Even though Rosetta does hammer the disk compared to other projects (and other workloads), the fact is the total amount of writes it does is bugger all of what the drive is capable of dealing with before it eventually dies, in 20 or more years time with much, much heavier use than even Rosetta does with it. And even then, as Annandtech showed many years ago, even consumer based SSDs will outlast their specified life expectancy, often by a huge margin. It took them 18 months to kill the last of their consumer SSDs, using loads that far exceed any regular desktop's extreme disk usage (yes they were SLC drives, and more recent flash technologies (MLC, TLC & QLC) have less total writes before they will fail. But the software to reduce write amplification, combined with capacities much larger than the tested drives, means current SSD would probably last longer, even with a heavier load). The SSD Endurance Experiment Grant Darwin NT |
Michael E.@ team Carl Sagan Send message Joined: 5 Apr 08 Posts: 16 Credit: 1,947,072 RAC: 559 |
Grant - Way cool! Thanks much for the help. I changed the Computing Preferences as you suggested to allow disk writes at 60 seconds and increased the memory allowed to 90%. My PCs have 8 and 16 GB respectively. Are these suggested Preferences written down anywhere? If not, if you could review it, want me to create a PDF-based guide? Send me a message. It needs to be easy to locate as well. Mike |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1698 Credit: 18,172,091 RAC: 24,436 |
Are these suggested Preferences written down anywhere? If not, if you could review it, want me to create a PDF-based guide? Send me a message. It needs to be easy to locate as well.I've just been copying & pasting as people post their questions/issues, although having them somewhere ready to go would speed things up. Grant Darwin NT |
Remarc Send message Joined: 23 Mar 20 Posts: 14 Credit: 302,773 RAC: 0 |
The 64-Bit app for Linux works fine with my machine, i think the 32 bit app for Linux has a bug. Microsoft Windows (98 or later) running on an Intel x86-compatible CPU 4.12 1 Apr 2020, 0:32:45 UTC 76,058 GigaFLOPS yes,delete all of them ) |
James W Send message Joined: 25 Nov 12 Posts: 130 Credit: 1,766,254 RAC: 0 |
Microsoft Windows (98 or later) running on an Intel x86-compatible CPU 4.12 1 Apr 2020, 0:32:45 UTC 76,058 GigaFLOPS Just a touch of irony here?? |
Remarc Send message Joined: 23 Mar 20 Posts: 14 Credit: 302,773 RAC: 0 |
Microsoft Windows (98 or later) running on an Intel x86-compatible CPU 4.12 1 Apr 2020, 0:32:45 UTC 76,058 GigaFLOPS exactly,man offered to drop 50% performance in this project) |
[VENETO] boboviz Send message Joined: 1 Dec 05 Posts: 1997 Credit: 9,739,012 RAC: 10,401 |
Microsoft Windows (98 or later) running on an Intel x86-compatible CPU 4.12 1 Apr 2020, 0:32:45 UTC 76,058 GigaFLOPS The current power on R@H is 22,776.39 TeraFLOPS: 1 - is not 50% 2 - is not a big problem. |
[VENETO] boboviz Send message Joined: 1 Dec 05 Posts: 1997 Credit: 9,739,012 RAC: 10,401 |
The current power on R@H is 22,776.39 TeraFLOPS My fault. There are different values: from home page: TeraFLOPS estimate: 1792.376 from cpu list: 22,776.39 TeraFLOPS from BoincStats: 356.004 TeraFLOPS Despite this discrepacy, the 32 bit systems are marginal |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1698 Credit: 18,172,091 RAC: 24,436 |
Despite this discrepacy, the 32 bit systems are marginalHow can you tell? The numbers of FLOPs are for the applications, it doesn't show the type of system (32bit or 64bit) that was running the application. Even so, one third of the current compute performance is not insignificant. Grant Darwin NT |
robertmiles Send message Joined: 16 Jun 08 Posts: 1233 Credit: 14,320,015 RAC: 3,383 |
My point exactly. There is no need for the 64bit applications, so why produce them?Perhaps the Rosetta Admins should think about removing 32-bit apps for x86 cpu's.Why? They're no slower than the equivalent 64bit application. Haven't you noticed that some operating systems are planning to drop support for 32-bit apps? Do you want Rosetta to be incompatible with new versions of those operating systems? |
robertmiles Send message Joined: 16 Jun 08 Posts: 1233 Credit: 14,320,015 RAC: 3,383 |
I got an error I've never seen before. ERROR: Cannot determine file type. Current supported types are: PDB, CIF, SRLZ, MMTF https://boinc.bakerlab.org/rosetta/result.php?resultid=1146962553 Does this mean an error built into the workunit? Rosetta 4.12 BOINC 7.16.5 Returning it got my first 4.15 task on Rosetta, though. |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1698 Credit: 18,172,091 RAC: 24,436 |
Haven't you noticed that some operating systems are planning to drop support for 32-bit apps?And that is actually a good reason to produce them, unlike the "32bit is old, lets get rid of it" argument for getting rid of 32bit applications- even though they provide a very significant portion of Rosetta's computing power. The same reason for producing 64bit applications is the same reason to continue with 32bit ones- to make use of the computing resources that are available. As it is, all present 64bit hardware & operating systems can handle 32bit applications- and their performance is on par with the 64bit ones. When support for 32bit applications is eventually dropped, then the need for 64bit becomes considerably greater. Grant Darwin NT |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1698 Credit: 18,172,091 RAC: 24,436 |
I got an error I've never seen before.That's how i would interpret it; the application doesn't recognise that file type. It could be the new application was released to process them, but it hasn't been released for Linux on x86 systems at all, nor for Windows 32bit. And if the servers haven't been configured to allocate the new work type only to the new application a lot of Tasks are going to be trashed. Hopefully it was just a one off file corruption glitch. Grant Darwin NT |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1698 Credit: 18,172,091 RAC: 24,436 |
It looks like there's going to be some problems with v 4.15- Estimated completion times are 39min 18sec, that's going to lead to a lot of overfilled caches & Tasks timing out. I can see a lot of work going to waste with this if they run to the default time of 8 hours. Edit- Suspended some Tasks to get the v4.15 ones to run, and it's not looking good. Two are presently running, and their current rate of progress has them running for 8 hours. 3 others ended with a Computation Error within 30 seconds. One running on Rosetta v4.12 windows_intelx86 took 42 seconds to die. Edit- Looks like a batch of dodgy Tasks. All 4 of the errors above were the same as robertmiles had, ERROR: Cannot determine file type. Current supported types are: PDB, CIF, SRLZ, MMTF ERROR:: Exit from: ......srccoreimport_poseimport_pose.cc line: 380 BOINC:: Error reading and gzipping output datafile: default.out Problem Tasks 0930dfdd5b7a8dbc439376d35ff43cd0_dock_ens_20_04_15_28_11_localDocking_0_SAVE_ALL_OUT_910048_18_0 4b344669e2c2754038b4438c4e2e5d64_dock_ens_20_04_15_28_14_localDocking_1_SAVE_ALL_OUT_910066_18_0 c31ed6e33409f8d7b8e76abdd93af3a6_dock_ens_20_04_15_28_13_localDocking_0_SAVE_ALL_OUT_910304_18_0 58975f05ec37c073d52b9aa38b7d1b28_dock_ens_20_04_15_28_13_localDocking_1_SAVE_ALL_OUT_910258_26_0 _globalDocking_ Tasks are processing Ok. Grant Darwin NT |
James W Send message Joined: 25 Nov 12 Posts: 130 Credit: 1,766,254 RAC: 0 |
It looks like there's going to be some problems with v 4.15- Estimated completion times are 39min 18sec; that's going to lead to a lot of overfilled caches & Tasks timing out....The first v4.15 tasks I got were also noted to be in the 39 min. range. However, my BOINC manager revised their completion times to 01:14:54, almost twice as long. Will see if I get same results as you did. |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2130 Credit: 41,407,444 RAC: 15,440 |
Edit- To be fair, at that rate, they should all complete by deadline... ...yeah, sorry. Had to say it... |
Message boards :
Number crunching :
Rosetta 4.1+ and 4.2+
©2024 University of Washington
https://www.bakerlab.org