Posts by Raistmer

1) Message boards : Number crunching : Problems and Technical Issues with Rosetta@home (Message 97181)
Posted 3 Jun 2020 by Raistmer
Post:
Last update broke connectivity with Rosetta@home seems for all my systems.
Even completed results can't upload.

And I have currently no time to implement fixes through whole fleet.
So, Einstein@home now.
Better to think twice before such radical changes.
EDIT: seems whole BOINC issue. Well, saves for electricity then.
2) Message boards : Number crunching : running out of tasks soon (Message 96295)
Posted 9 May 2020 by Raistmer
Post:
Pity that GPU can't be used here....
3) Message boards : Number crunching : Rosetta 4.1+ and 4.2+ (Message 96294)
Posted 9 May 2020 by Raistmer
Post:

This seems counterintuitive. Wouldn't I be better off to increase the expected length and then (I hope) run them in less time than to decrease the time and risk not making the deadlines?

Expected length is the amount of CPU time task will allowed to run. And here is the big difference with SETI and most other projects.
Task doesn't contain fixed number of calculations to complete it. If CPU time allows, new model will be started for same task (slightly different initial atoms configuration or smth alike).
So, if you allow 8 hours per task it will run 8 hours. Only 2h - then it will end in 2 hours.

And yes, to avoid cache overflow in the future better to set BOINC cache size as small as it could be. But changing cache size will not help with already downloaded tasks.
4) Message boards : Number crunching : Rosetta 4.1+ and 4.2+ (Message 96275)
Posted 8 May 2020 by Raistmer
Post:

Am I right in assuming that BOINC will eventually figure this out? In the meantime, what's my best move? Abort all that won't make it? Abort only the unstarted? Let them all go until BOINC figures it out?

Thanks.


This project (instead of SETI you familiar with) allows to reduce lenght of already received tasks.
Best option for your host is to set them to minimal possible length anfd then gradually increase as long as you don't miss deadline.

This can be done in project options here:


https://clip2net.com/s/47qBO85

As you could see I have 2 different sets of options - for powerful hosts (big task length) and for netbooks/smartphones (short length, 4 hours per task currently)

P.S. You need to update project settings (update project from BOINC) and then restart BOINC client itself to update already downloaded tasks length. Newly downloaded will be of new length already.
5) Message boards : Number crunching : How secure is my machine when running R@h? (Message 96274)
Posted 8 May 2020 by Raistmer
Post:
What are you getting at with the phrase?
[quote]...statistics of completions of protein folding by chip type

Perhaps something like this:
https://setiathome.berkeley.edu/cpu_list.php
or
https://setiathome.berkeley.edu/apps.php
ot this
https://setiathome.berkeley.edu/host_stats.php
6) Message boards : Number crunching : Should you run with some threads "idling" or not? (Message 96272)
Posted 8 May 2020 by Raistmer
Post:
Some example when you should.
Not from this project, from E@h (those devcies can't do this project at all):

I have 2 identical Digma Hit phones (cheap smartphone, but 4 ARM cores in it).

See their results:
https://einsteinathome.org/ru/host/12826438/tasks/0/0
and
https://einsteinathome.org/ru/host/12826433/tasks/0/0

One running only 1 core of 4, second - 2 cores of 4.
Results self-explaining. If some other resource limiting (storage space in that particular case) no matter how many cores device has, matters how many it can really use.
7) Message boards : Number crunching : Boinc for Android 7.16.5: Another volunteering app is installed on this device as well as force closing (Message 96208)
Posted 7 May 2020 by Raistmer
Post:
Would be useful to post such issues on BOINC own forums too, not just inside project forum (cause it's BOINC-wide issue, not this project issue).

Regarding "another app"- it founds remnants of own previously running process (some lock file for example or smth similar) and complains.
This behavior I see on all Android versions I used and on all phones. If prev instance terminated abnormaly or even just closed - new one has good chances to complain. But this issue is not show-stopping one, just relauch BOINC app - usually it helps, 1-2 attempts at most.

And regarding terminations - this much more serious issue.
After upgrading to 7.6.15 I lost one of my crunching hosts - Redmi Note 5 phone. It worked OK with prev version but after auto-update BOINC constantly crashes. I had no time to fix so phone currently lost for distributed computing.
More thorough testing required definitely before to force updates!
8) Message boards : Number crunching : minirosetta_database directory in slot (Message 96207)
Posted 7 May 2020 by Raistmer
Post:

What determines the project I get my preferences from? Is it as simple as the most recent timestamp on a change wins?

YEs. The last edited will be used.
9) Message boards : Number crunching : Rosetta on Android device : types of errors and their meaning and reasons (Message 95787)
Posted 2 May 2020 by Raistmer
Post:
Some new phones have assymmetric cores - 4 more powerful ones and 4 less, for example (as in Redmi Note 8).
If phone restricted by available RAM, makes sence to use only powerful cores and leave less powerful for maintenance tasks.
Don't know how many and what size of cache levels in modern ARM architectures (if they have shared L3/2 cache for example, reducing number of cores could improve memory access patterns) but definitely one need all active BOINC tasks be in RAM, not in SWAP, for better performance and less phone degradation.
10) Message boards : Number crunching : minirosetta_database directory in slot (Message 95715)
Posted 2 May 2020 by Raistmer
Post:

That was 2008, things have changed. Mini is no longer newer. It's in fact quite the opposite, now that Rosetta 4.15 is out, with 4.20 coming very soon.

So, all "that time new" features were included into "main" app perhaps?
Ok, that enough for my curiosity, thanks.
11) Message boards : Number crunching : minirosetta_database directory in slot (Message 95713)
Posted 2 May 2020 by Raistmer
Post:
I'm pretty sure it was announced that Rosetta Mini was being depreciated, but now there is a whole bunch of new Rosetta Mini work out there. I was wondering if the Rosetta Mini application will now be updated to reduce the disk usage as well?

BTW, what the difference between Rosetta and Rosetta Mini ? What part is "mini" ?


https://boinc.bakerlab.org/rosetta/forum_thread.php?id=4394#56015

From that short description one could infer that "mini" is newer and has chances to become better too - so why deprecation then?
12) Message boards : Number crunching : Rosetta 4.1+ and 4.2+ (Message 95707)
Posted 1 May 2020 by Raistmer
Post:
I updated the rosetta app version to 4.20 for the arm, android, and mac platforms. I will update the remaining platforms in the next day or so.

Please post issues regarding this updated app version here.

This update includes:

1. extraction of the Rosetta database into the project directory with all following jobs reading from the same database rather than extracting into the slot directory for every job. This significantly reduces the disk usage per job.
2. checkpointing in the Rosetta comparative modeling protocol. This should significantly reduce wasted cpu time if jobs are preempted often, particularly for jobs that take a long time to produce models.


Observed first run of 4 v4.20 tasks simultaneously for my quad host.
First instance extracted archive (big I/O counts) then 3 others just use it (low I/O counts).
No issues and no fallbacks on this host.
Excelent!
13) Message boards : Number crunching : minirosetta_database directory in slot (Message 95704)
Posted 1 May 2020 by Raistmer
Post:
I'm pretty sure it was announced that Rosetta Mini was being depreciated, but now there is a whole bunch of new Rosetta Mini work out there. I was wondering if the Rosetta Mini application will now be updated to reduce the disk usage as well?

BTW, what the difference between Rosetta and Rosetta Mini ? What part is "mini" ?
14) Message boards : Number crunching : Rosetta 4.1+ and 4.2+ (Message 95667)
Posted 1 May 2020 by Raistmer
Post:
Regarding WinXP and memory:
I'm runnign Vista x86 and it does Rosetta work OK. So, definitely not a 32-bit restriction to available memory.
Most probably app was built against new MS SDK. And they broke (they do this regularly) process creation function for old OS. PErhaps app should be built with XP-compatibility flag turned ON.

And regarding "unpatched" and so on. Clear your minds from marketing bul...t.
Talented hacker will enter in modern Win10 just as easy :)
On each old closed backdor each new update opens dozen new ones...
Old OSes just never had exposed SO many network services as modern OSes do.
15) Message boards : Number crunching : Rosetta 4.1+ and 4.2+ (Message 95666)
Posted 1 May 2020 by Raistmer
Post:
WoW, this one seems more algorithm-depended than all others...

<core_client_version>7.6.33</core_client_version>
<![CDATA[
<message>
- exit code 1 (0x1)
</message>
<stderr_txt>
command: projects/boinc.bakerlab.org_rosetta/rosetta_4.15_windows_x86_64.exe -abinitio::fastrelax 1 -ex2aro 1 -frag3 00001.200.3mers.index -in:file:native 00001.pdb -silent_gz 1 -frag9 00001.200.9mers.index -out:file:silent default.out -ex1 1 -abinitio::rsd_wt_loop 0.5 -relax::default_repeats 5 -abinitio::use_filters false -abinitio::increase_cycles 10 -abinitio::rsd_wt_helix 0.5 -beta 1 -abinitio::rg_reweight 0.5 -in:file:boinc_wu_zip ZN5I2D63_RiAFP_fiber_solenoid_Peralta_published_data.zip -out:file:silent default.out -silent_gz -mute all -nstruct 10000 -cpu_run_time 28800 -boinc:max_nstruct 20000 -checkpoint_interval 120 -database minirosetta_database -in::file::zip minirosetta_database.zip -boinc::watchdog -boinc::cpu_run_timeout 36000 -run::rng mt19937 -constant_seed -jran 2934427

ERROR: Error in core::scoring::methods::RamaPreProEnergy::residue_pair_energy(): The RamaPrePro term is incompatible with cyclic dipeptides (as is most of the rest of Rosetta).
ERROR:: Exit from: ......srccoreenergy_methodsRamaPreProEnergy.cc line: 148
BOINC:: Error reading and gzipping output datafile: default.out
06:09:44 (1076): called boinc_finish(1)

</stderr_txt>
]]>

Result: http://boinc.bakerlab.org/rosetta/result.php?resultid=1166025891
ZN5I2D63_RiAFP_fiber_solenoid_Peralta_published_fold_SAVE_ALL_OUT_925453_948_1
16) Message boards : Number crunching : BOINC for Android and R@H web-site computing preferences (Message 95658)
Posted 1 May 2020 by Raistmer
Post:
If you set project share of 0 BOINC should fetch work for such project only when there is idle computational device available.
That is, no more than one task per CPU,
17) Message boards : Number crunching : RALPH@Home ? (Message 95657)
Posted 1 May 2020 by Raistmer
Post:
4.20 is out
This version includes a fallback to the original method of extracting into the slot directory for each job if extracting into the project directory fails.

And there were such cases? Worth to investigate why extracting into slot not fail while extracting into project dir fail.
18) Message boards : Number crunching : The most efficient cruncher rig possible (Message 95623)
Posted 30 Apr 2020 by Raistmer
Post:
Thanks for comments.
Do you aware of smth similar for Windows world (I know that Win10 includes Superfetch, but besides of that, maybe some third party apps/drivers/services) ?
19) Message boards : Number crunching : The most efficient cruncher rig possible (Message 95595)
Posted 30 Apr 2020 by Raistmer
Post:
If you would use the fastest zram compression possible (LZ4?), you may get 6GB out of a 4GB rPi4.

I've done a dirty patch myself on BOINC on a low-memory PC so it would schedule as much work as the zram'med memory amount would allow, instead of limiting to the physical amount. This solves the "waiting for memory" messages. I had to patch it because it caps memory allowance to be under 100% when requesting jobs, although, I guess after you have the jobs the scheduler may allow > 100% settings (TODO).

It has been working pretty good for weeks now.

Could you estimate how big share of compression in processing power? How many cycles goes to maintain compressed data versus to retrieve it directly from memory?
20) Message boards : Number crunching : Access violation exception R@h 4.15 (Message 95593)
Posted 30 Apr 2020 by Raistmer
Post:
Peak working set size 1,413.90 MB
Peak swap size 1,417.56 MB

Reason: Access Violation (0xc0000005) at address 0x007B0450
A fairly RAM hungry Task there- how much RAM & how many Tasks running on that particular system?


yep, most probably Out of memory-related, but it's not Out of memory - it's access violation.
That is, some alloc doesn't have check for success - that's the problem.


Next 20



©2020 University of Washington
https://www.bakerlab.org