Posts by Mr. Kevvy


log in
41) Message boards : Number crunching : Are some gpu tasks longer now? (Message 1794720)
Posted 9 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
Thats definitely not the case.

Speaking from my own experience with NVidia cards and CUDA applications. GBT 'guppi' VLARs are settling around 53 minutes (two at a time) on GTX 750Ti. Will start GTX 970 test in about two hours' time.


980 11-13 mins., 970 13-16 mins. with 0.2 core one at a time. Does about the same regardless of the CPU or PCIe revision. Non-VLArs are/were 3-4 mins.
42) Message boards : Number crunching : Are some gpu tasks longer now? (Message 1794711)
Posted 9 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
Perhaps the powers to be will change the programs and not schedule vlars in the near future.


Dr. Korpela indicated that it was tried on Beta to add a setting to the Project Preferences to allow/disallow GPU VLAR, but that this resulted in no GPU work at all. I'm puzzled by this, as the code with no option to disallow them was in placed and worked just fine for a while; something must have been done incorrectly so I'm still hopeful that they are working on it and a solution is forthcoming. I keep seeing that it's a non-issue because non-VLAR Arecibo work is going away, but I don't see any evidence of this (edit: Richard H. posted some above... damn) and the team has indicated otherwise (about equal amounts of work will be distributed from each source.)

I'm now finding that the newer blc6 and blc7 VLARs now run about 2x slower on my GTX 750Ti with cuda50 - it was closer to your 4x at the start of the run.


I noticed today that my RAC started rising after over a month of fall. It should have bottomed out around 260K. So it does appear that some (welcome) change was made. (Edit: maybe it was just the recent AP run.)
43) Message boards : Cafe SETI : Thor is the Designer Cafe! (Message 1794532)
Posted 8 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
P.S. There's a chance that a personal appearance may also happen this evening.


Nice... a perthornal appearance. :^D
44) Message boards : Number crunching : Tried to free up all tasks, but I screwed up; how to fix? (Message 1794495)
Posted 8 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
They are now "ghosts" and will have to time out. In future, you can abort the tasks then update the project and then they will be immediately reassigned.

(Edit: Ensure "No new tasks" is on when this is done of course!)
45) Message boards : Number crunching : Invalid Host Messaging (Message 1794470)
Posted 8 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
computer 7374749 would be:
based on RAC in the top-users would indicate this user:
Name: fbi
RAC: 74,728
country:Germany

the only german @ ~74K RAC


Detective work is good, but the result is ambiguous. For example, here is an Austrian with enough RAC and hidden computers.

Since the admins. here (ie Dr. Korpela) can see who it actually is all we really can do is advise them.

Didn't know that - sorry.


No worries... at least you are trying to help! :^)
46) Message boards : Number crunching : Invalid Host Messaging (Message 1794436)
Posted 8 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
. . This host is not very helpful.

http://setiathome.berkeley.edu/results.php?hostid=7940844


Too minor to worry about really. Also it would be suggested to confine multiple short replies into one posting.

@Richard et al: 7374749 is now the #2 host showing 1572 trashed GUPPI work units. Nothing appears to have been been done so far.
47) Message boards : Cafe SETI : The joke thread Part 4. (Message 1794264)
Posted 7 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
(Warning: NSFW language in a large font if clicked!)

(2016 June 7) Today's joke is the User of the Day on the main page, bottom-left. As well, click his profile (if no one's around) and check his team name. Interesting random pick...

It may not last, so if it's someone nondescript now, it's been changed. :^)
48) Message boards : Number crunching : Panic Mode On (103) Server Problems? (Message 1794102)
Posted 6 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
You don not have to restart BOINC, just suspend/resume the task and it will run fine on next run.


Didn't work. Um... Linux!
49) Message boards : Number crunching : Invalid Host Messaging (Message 1794047)
Posted 6 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
I've written jointly to Dr Anderson and Dr Korpela about this host...
I included Dr Anderson, because some of the temptations leading to this sort of behaviour arise from the BOINC platform design, at least as much as from SETI's use of it.


Excellent and thanks; may carry more weight from a developer. This thread has numerous examples of the same, including a few posts down (Cody Johnson?) with >32K tasks In Progress.
50) Message boards : Number crunching : Panic Mode On (103) Server Problems? (Message 1794017)
Posted 6 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
No, I am using a SSE2 client.


Thanks... that was the only thing I did differently from this machine than several others so no idea why it's the only one with this problem.
51) Message boards : Number crunching : Panic Mode On (103) Server Problems? (Message 1794008)
Posted 6 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
...but CPU only tasks have a glitch. Sometimes they do not make any checkpoint and seem to go on forever, so I have to abort them.
Tullio


If this is on one of your Linux machines, I have the same problem only on this computer. However you may not have to abort the the stuck work units. When I quit the BOINC Manager including terminating apps. and relaunch BOINC they complete normally; I checked them and they are valid and get credited. I need to do this a couple of times a day to clear one or two stuck ones. I was wondering when I would see someone else with the same issue.

Are you using the AVX MB client?
52) Message boards : Number crunching : Invalid Host Messaging (Message 1793980)
Posted 6 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
Host 7374749

This one should be reported to Dr. Korpela(?) The owner may have developed code that locks all the GUPPI work unit files so that they error out but allows the non-GUPPIs to complete normally (or maybe doing it manually ie by deleting them, just trying to avoid the "Aborted".)

In progress (871) · Validation pending (1670) · Validation inconclusive (87) · Valid (861) · Invalid (0) · Error (925)

All of those 925 error units appear to be GPU GUPPI work with the same -5 Can't Open File. Also shows as "Anonymous" indicating the owner has hidden their computers.
53) Message boards : Number crunching : vlar data and breakthrough listen (Message 1793917)
Posted 6 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
In short - both lists most probably intersect.


Only slightly... there are (edit: were prior to the May announcement) 460 stars with Kepler object planets, the furthest being 8kLy away, but there are 600M stars within 5kLy from earth, so less than one millionth contain Kepler planets.

Thus, if not looking at confirmed planetary systems we may as well stay with random searches as the search has been essentially randomized.
54) Message boards : Number crunching : vlar data and breakthrough listen (Message 1793910)
Posted 6 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
The Greenbank telescope does mainly close observations of suns that may have planets around them (therefore more likely to have aliens).


That was the initial idea (to observe Kepler objects) but doesn't seem to have carried through to the Breakthrough Listen observation plan:

1. Green Bank Radio Telescope

World’s deepest searches for artificial signals in five key samples (Northern Hemisphere):

All 43 stars within 5 parsecs, at 1-15 GHz. First-ever complete SETI survey within 5 parsecs. Sensitive to “Earth-leakage” levels of radio transmission.
1000 stars of all spectral-types (OBAFGKM). Within 50 parsecs. 1-15 GHz.
One Million Nearby Stars. In 2016, first 5,000 stars; 1 minute exposure (1-15 GHz)
Centers of 100 nearby galaxies: spirals, ellipticals, dwarfs, irregulars (1-15 GHz)
Exotic Stars: 20 White Dwarfs, 20 Neutron stars, 20 black holes

2. Parkes Radio Telescope

World’s deepest searches for artificial signals in six key samples (Southern Hemisphere):

All 43 stars (at south declinations) within 5 parsecs, at 1-15 GHz. First-ever complete SETI survey within 5 parsecs. Sensitive to “Earth-leakage” levels of radio transmission.
1000 stars (south) of all spectral-types (OBAFGKM). Within 50 parsecs. (1-4 GHz)
One Million Nearby Stars (south). In 2016-2017, first 5,000 stars; 1 minute exposure (1-4 GHz)
Galactic plane and Center (1-4 GHz)
Centers of 100 nearby galaxies (south declinations): spirals, ellipticals, dwarfs, irregulars (1-4 GHz)
Exotica: 20 White Dwarfs, 20 Neutron stars, 20 black holes
55) Message boards : Number crunching : No more guppi's=vlars on the gpu please (Message 1793667)
Posted 5 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
Each major iteration in the MB application has seen an increase in the amount of calculation performed. With that an the increased complexity of the calculations required for the guppi data it is hardly surprising that these take longer to run.


That isn't the issue here... GUPPIs don't do any more calculation or take any longer to run on CPUs than Arecibo VHARs; they usually end up a little faster. They only run very slowly on GPUs due to architecture issues.
56) Message boards : Number crunching : No more guppi's=vlars on the gpu please (Message 1793566)
Posted 5 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
. . Now a magic app that would turn a GPU VLAR WU into a CPU WU and a CPU based nonVLAR WU into a GPU WU would go down well :)


Thank you for the below. :^)

I thought about writing this as a practice app, but quickly realized that with it, all the existing MB work units would be shuffled to the GPU and complete quickly, while the work would still be randomly assigned as half-and-half GUPPI VLAR. Soon, the machine would have nothing but GUPPI VLAR work units so would be in the same situation, unless the owner starting aborting them. And this even assumes that the validators allow a CPU-assigned work unit to be completed by a GPU or vice versa. That I don't know. Raistmer mentioned a rescheduling app so I think he does...
57) Message boards : Number crunching : No more guppi's=vlars on the gpu please (Message 1793363)
Posted 4 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
If the option arrives to disallow that, I'll be in something of a dilemma - to switch or not to switch.


Having initiated the fundraiser for the receiver, I'd probably be the last person one would expect to throw the switch, but I might. That decision comes not from credit or technical issues, but from the observation plan

All 43 stars within 5 parsecs, at 1-15 GHz. First-ever complete SETI survey within 5 parsecs. Sensitive to “Earth-leakage” levels of radio transmission.
1000 stars of all spectral-types (OBAFGKM). Within 50 parsecs. 1-15 GHz.
One Million Nearby Stars. In 2016, first 5,000 stars; 1 minute exposure (1-15 GHz)
Centers of 100 nearby galaxies: spirals, ellipticals, dwarfs, irregulars (1-15 GHz)
Exotic Stars: 20 White Dwarfs, 20 Neutron stars, 20 black holes


The bill of sale on this new data was that it would be focusing on the habitable planets discovered by the Kepler mission. I don't see that anywhere in the BL list.

My reasoning behind wanting it is that, unfortunately and I am hoping temporarily, we only have a sample size of one. We know that life can form and evolve to intelligence on a planet large enough to have a magnetic field in the >273K <373K "goldilocks zone" in a system with a "good jupiter" around a class G2V star. As we know nothing about other possibilities, this is what we should be looking at. If the response is that because we don't know what else is out there then we are limiting our chances of discovery, then we must stick to doing random searches a la Arecibo as we have until now; anything else is making a decision as to what we expect to find thus limiting the possibilities. The furthest Kepler object is 8K Ly away, and there are six hundred million stars within 5K Ly. There's just too much to look at without narrowing it down to the best possibilities of what we currently know in order to maximize our odds (which is why we got interested in GBT in the first place, right?)

There isn't much chance of life around an "exotic" because it fried any planets around it on its way to becoming "exotic" so this is designed to look for deliberate beacons (is it?), but as we don't know what logic alien intelligence follows, we may as well look anywhere... they may decide to put beacons around habitable planets, because that is where we should be looking, except unfortunately they're from a Jupiter-sized world with oceans of liquid that to us would be gas, so we don't bother looking there because it isn't habitable to us and miss it by doing targeted searches.

I hope to be proven wrong. And I would still probably add a multi-CPU Xeon box or two to the mix to work on the GUPPIs.
58) Message boards : Number crunching : No more guppi's=vlars on the gpu please (Message 1793346)
Posted 4 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
The problem is this isn't just "RAC-obsession". VLARs were kept off the NVidia GPUs for a long time because they can cause the machine to stutter or hang or the work unit to crash. ie my wife's machine starting generating piles of errors as every time she watched any sort of streaming video ie YouTube or Facebook the VLAR work unit in progress would crash.

Even if the volunteer chooses not to accept them, their machine(s) will still get them on their CPUs where they will run just fine and cause no issues (for some, even faster than non-VLARs) rather than 2.5x or more slower, at least while there are some non-VLARs. (I keep reading that they are going away, but I haven't seen any indication when. The indication from the team was that it was going to remain about half and half.) If the more efficient GPUs stick to the non-VLAR work, this frees up more CPUs out there for the VLARs (even on my own machine I see this with the GPUs choking on a stack of GUPPIs while the CPUs have all the regular non-VLAR MB work they should be getting. Random scheduling is inefficient.) It's possible that the net result of allowing them on NVidia GPUs is hardly any improvement in compute capacity, just an increase in grief and people leaving the project... yes, some have over it.

I'm still puzzled that the code to keep the GUPPIs away was already in place and worked just fine for months(?), but apparently when it was put back on Beta with a setting in the project prefs. to allow or deny them, the result was no GPU work at all. Something must have not been done the same, because the code was already in place and working. I hope it is retried, so this issue can finally be put to rest.
59) Message boards : Number crunching : Average Credit Decreasing? (Message 1792943)
Posted 2 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
1 Credit or 1,000,000 RAC. Wat's Da Bigggeee?


I can live with it (though it's kind of a pain) but there are volunteers either leaving the project over it or aborting GUPPI work units by the thousands, neither being good for the long-term. I can empathize as they are/were spending not only thousands on hardware, but for many similar amounts on power to do a certain amount of work. And now suddenly that is cut in half...

As well, as SETI@Home has been tapped to do the processing for BreakThrough Listen, there is now too much work available, but this work is being done at roughly a third the speed it should/could by volunteers' GPUs. If this bottleneck could be reduced or eliminated, we may be able to keep up.

OK, got the obvious out of the way. :^)


I don't care if it's me, Petri, Raistmer, or someone else entirely. Everyone involved in some small way will have to do the 'jiggly worm EUREKA dash'


So it turns out to be a software issue after all. :^D
60) Message boards : Number crunching : Average Credit Decreasing? (Message 1792891)
Posted 2 Jun 2016 by Profile Mr. KevvyCrowdfunding Project Donor
I had a look at 30-40 AMD GPU machines' tasks, and as far as I can tell, it looks as though new AMDs are similarly affected. I was unable to find any that didn't take 2-4x as long to complete a GUPPI VLAR work unit. This includes with and without SoG. Maybe I just didn't look in the right place, so if anyone has any URLs otherwise they would be welcome.

Old AMDs I saw were very slow but they were consistently very slow with the VLARs, and this would reinforce that it's an issue with the VLAR results exceeding cache size, as the old cards would have hardly any cache so both VLAR and non-VLAR work would exceed it so have about the same runtimes.

If correct, this would seem to imply that the only resolution for this is going to be a workaround redesign of the pulsefinding algorithm, if this is even possible.


Previous 20 · Next 20

Copyright © 2016 University of California