Posts by William


log in
1) Message boards : Cafe SETI : Real Pictures of Real Seti Friends (Message 1502593)
Posted 11 days ago by Profile William
Nobody, nobody... :)

Richard?

You maybe?


You're not making sense. However, it being you, maybe it doesn't have to make sense. It being you, maybe the only sense is in trying to provoke. Be careful who you provoke, dear.
Wrong thing to say, you'll try now ;) Good thing I'm going for a two week break.
2) Message boards : Cafe SETI : TLPTPW 210 - Make Me Laugh! (Message 1502591)
Posted 11 days ago by Profile William
*sob* I didn't make Angela laugh... There goes my box of cookies...

Winning before disappearing on Easter break.
3) Message boards : Number crunching : Running Stock Experiment (Message 1502082)
Posted 12 days ago by Profile William
RAC has a half-life of one week - asymptotic curve.

IOW after 2 weeks you still have ~1300 points from before the switch contributing to RAC. It may take a few weeks to level off, not counting pendings.

Your GPU tasks take roughly 1/10 of the time of CPU, so relative contribution should be around 500 from CPU and 5000 from GPU (not counting AP) - that brings us to at least 4 weeks before old RAC has decayed enough for 'MB CPU only' RAC to be predominant. Aybody mentioned you need patience for this?

edit: I forgot of course about it being 8 core. so maybe half half. more around 2 weeks then.
You ain;t looking at a smooth system of course, so might be worth running extra to get a better idea of where it levels of (and don't talk to me about error margins).

this host - to save me having to look it up again each time ;)

BTW I hit my head pretty hard yesterday, if I'm talking BS you can take it as a sign that cognitive functions suffered.
4) Message boards : Cafe SETI : TLPTPW 210 - Make Me Laugh! (Message 1501644)
Posted 13 days ago by Profile William
Passing through after having given permission.

Home made cookies?

Sorry, I'm german. Ve don't do zer humour zing.
5) Message boards : Number crunching : Running Stock Experiment (Message 1501330)
Posted 14 days ago by Profile William
Renaming app_info.xml to something else is a quick and easy way to go stock and preserve the app_info.xml for future reuse. Yes, I'd recommend that, especially if you did tweaks to the app_info.xml.
Just make sure you have run down your cache before you rename it - you should get everything back, with sent lost results enabled, but it's neater to start with an empty cache.

ad 3) stock MB on GPU will be a slight bugger with the multiple cuda versions. Ah, I see you already have established APRs on that host. Not a problem then.
6) Message boards : Number crunching : I'm falling, I bought a parachute. From 100% AP, to 100% MB. (Message 1500998)
Posted 15 days ago by Profile William
scaling. best stock MB is AVX, best stock AP is what, SSE2?
higher SIMD, less credit, if you want it put simply.

but do by all means run an experiment. Data is always good.

The implementation of CreditNew does not match the whitepaper.

over and out
7) Message boards : Number crunching : Issues with Lunatics site ? (Message 1500911)
Posted 15 days ago by Profile William
Same procedure as last year?

Anybody told Tobias (?) yet? Remember Lunatics is just a part of kwsn - any queries as to why there is a problem need to be addressed by the KWSN Admin - and I think he hasn't been very active for quite a while. The other kwsn sites are down for me as well, so it is a global KWSN problem.
Nope this isn;t my two cents, it's just half a penny.
8) Message boards : Number crunching : Letter to Professor Stuart Bale, Director SSL (Message 1498545)
Posted 20 days ago by Profile William
I was asked to be diplomatic about this.

First of all I tend to agree with James. Prof. Bale in all likelyhood knows nothing about BOINC and credit. He should know something like SETI@home exists within the premises, but that probably is about it. So on receipt of that letter he probably will say 'WTH?' and pass it on to Eric and David, since they were mentioned. If we are very lucky he'll just bin it as incomprehensible.

If we are less lucky Eric and David get it and read it.

Now put yourselves into their shoes. Something you should always do when writing such letters. Always look at the thing from the other persons POV. Always look at how you would feel when you get such a letter. And then think again if you found a good way not only to make your point, but more crucially to make the other person do what you would like him or her to do. Somebody who ends up antagonised is very unlikely to respond in the way you want. It's not only what you say it is how you say it. I've been trying to teach that lesson for some time now. The only effect so far has been for me to be promoted to spokesperson...

So, Eric first. Eric has stated several times, that he does not like the low credit and the credit imbalance either. And it is not his fault. It is a systemic design problem of the way CreditNew was coded.

You are right, Lionel, the only way to deal with this is to recode CreditNew.

Now, Eric doesn't want the low credit, but there's nothing he can do about it. There are workarounds but they are clutches at best. Plus, the BOINC side of the project is David's domain and Eric has very little say in that.
So, Eric might feel you (and you made your letter sound as if you were speaking for a lot of people) blame him for something that is not his fault and that he would love to change.
How does somebody react if you blame them for something that is not their fault and that they would like to change? Right, there you go. One person antagonised. If you are lucky just annoyed. But certainly not in the mood to move mountains to change things.

David? This is a bit tricky.
David is not a bad person. David is a person out of their depth with a piece of code which has a complexity that has long since exceeded easy understanding of the whole picture. Like an artist that has a whole mural to do and doesn't see that the piece in front of him is connected to the other side of the room.
David knows about volunteer retention problems. David has been told various times about credit. It's just that David is no engineer. He cannot understand that the way he coded CreditNew just has to have the outcome we see. That the drop is inherently built into the design. It's probably not, as we often tend to think, that he does no believe that credit is screwed. It is not a matter of belief. It is a matter of hard data and showing the problems in the code and how that way of coding has the effects we see. I tend to believe that he genuinely does not understand why that code grants credit far below the cobblestone scale. Why that code HAS to grant below the cobblestone scale as soon as SIMD apps are involved.
So David has no clear understanding of the problem. It is not a matter of belief. It is the moment you step back from the tapestry and realise you've been using the wrong type of blue for that past two square metres.
So what do you do? The sensible thing is to unravel and redo.
The human thing is to say sod it and carry on.

So how will he react to such a letter? Positively? Really???

Now, Lionel, you may not be aware of that, but a few of us have been trying to correct this matter for some time now, though it is slow going, due to various contraints on our collective time.
We did walk the code. There are now two people on this planet who know exactly how the CreditNew code works. There may be a couple more that walked the code and didn't say 'sod it' and understood. But they have kept silently to themselves, if they exist.

I approached Eric and Bernd Machenschalk from Einstein with our conclusions. I approached a few more projects, trying to figure out if they use CreditNew at all (many don't for various reasons) and if they would be willing to partake in a beta testing of 'NewCreditNew'. Jason did show the graphs in some other thread here. When Einstein have finished their infrastructre overhaul, they have agreed to test on their beta server. So we are making progress.
In the end we hope to have code that can be used by boinc as is. Code David can trust. Code David will use.

Do you have any idea how difficult it is to get David to change course?
It can be done, with a lot of patience and goodwill and diplomacy.
[he doesn't like sarcasm, btw]

When an opportunity presented itself I even approached David and told him we were doing this. The response I got can only be classed as mixed. But it was not at that point a complete rebuff. It looked like he could be persuaded to be openminded about the issue and look at what we have (or rather will have).
You have no idea how big a step forward that was.

And now, dear Lionel, you come and drop that bombshell into our laps.
You may have endangered the whole operation that was almost silently working towards a resolution.
David may end up so antagonised, that he'll put up all the barriers people have when they feel attacked. And then we are going to have one hell of a job to get our code in.
It was never going to be an easy thing. I was never sure I could pull this off. And when I say I, I mean I. Jason does the code, Richard does consultancy and proofreading but I am the spokesperson. I write the letters. It is my job to phrase what I write such in a way so the person reading it will NOT put up barriers. It is my eloquence and diplomacy and frankness and bluntness and occasional rudeness that has so far managed to get things done as far as they have.
So, I was never sure I would be able to pull this off and get David to accept the rewritten code. I just hoped I could. Because I am fairly sure there will only be one shot at this. If we fail, nobody else will ever do it again. Not the least because the combined skillset is so rare.
People always like to complain. Very few people do the next step and do something about it that REALLY helps.
And what really helps is to code. Nothing else. David cannot be expected to rewrite that code. He doesn't understand what's wrong with it. You just have to go and do it and hope that he accepts it.

So as I mentioned at the beginning, I hope that Prof. Bale chucks the letter into the bin. Or that David has a brief moment of annoyance before he chucks it in the bin. Anything but a reaction that endangers our operation.
It is a long way to go still. It would be nice if nobody was throwing spanners into the works.

Or as somebody else likes to put it 'Patience, grasshopper'.
9) Message boards : Cafe SETI : Shannon/kenzieb (Message 1491932)
Posted 33 days ago by Profile William
As to the thread topic, my response can be summed up as 'So what?'

Everybody else may I remind you to keep it civil and do NOT attack other posters for their opinions.

Oh and Batter Up, on a quick count at least 6 posters on these boards can be confirmed as female by other people here having met them in the flesh.
Out of those 6 five have children and at least five are or have been married.

We may be a minority, but we do exist. No, I won't be posting a picture of myself. That doesn't prove anything anyway.
10) Message boards : Number crunching : Invalid WU question (Message 1491913)
Posted 33 days ago by Profile William
he shows 2 gaussian and a few spikes before stderr bit the dust.

If it's signal ordering it is still a known issues, that may have been addressed since. And it's bad luck to have so few signals coincide that you get an invalid out of it.

Anyway, same over all message - nothing to worry about. The odd invalid just happens. If you start to get lots of inconclusives and subsequent invalids on a host, then that needs attention and possibly some TLC.
11) Message boards : Number crunching : CLOSED - SETI/BOINC Milestones [ v2.0 ] - XXVI - CLOSED (Message 1491907)
Posted 33 days ago by Profile William
1500 posts :)

ok, 1501 now

and had an anniversary and I've got 6M too - how did that happen ever so quietly?
12) Message boards : Number crunching : Invalid WU question (Message 1491898)
Posted 33 days ago by Profile William
http://setiathome.berkeley.edu/workunit.php?wuid=1451643950

I don't understand why I got an invalid on this wu when the other two crunchers got credit. And it's the only one sense I restarted crunching again.

made link clickable

looks like stderr is truncated. Known issue. Don't worry about it, just bad luck.
13) Message boards : Number crunching : Mix and match GPU? (Message 1491810)
Posted 33 days ago by Profile William
Thanks Hal. You mean a separate project for each card?

No, he means two separate instances of BOINC, each driving only one of the GPUs.

Forget it, it is WAY above your paygrade.

Hal, you are welcome to try and explain, but while I am probably one of the very few persons who know how to set that up properly I also know how complicated it is, and given Miklos' prowess with computers I strongly advise against anything that involves more than clicking a few buttons. You've not read 'How to debug' Hal have you?

A 580 and a 780Ti? Might work, don;t think they are too far apart in speed.

Um.

Ok, Miklos, I'm afraid you will NOT get both cards in one host to work. Because to get them to work you need a working cc_config.xml to tell boinc to use all the GPUs...

Do you have a third host you could put that card in? Else, maybe just chuck out the 460 and put the 780Ti in there (if the rest of the system is up to it).
14) Message boards : Number crunching : Panic Mode On (87) Server Problems? (Message 1490630)
Posted 35 days ago by Profile William
No Beta site traffic either.

Well, I just got work from Beta and reporting to main worked too.

Actually the stuck SSP shows more than 300k to send - we do know that a stuck SSP can lead to runaway splitters, when stuck below the highwater mark. I guess if it got stuck _above_ the highwater mark now, splitters are probably not gettting the signal to fire up or at least not working flat out.

Either way, I doubt it will get sorted before maintenance.

edit: small correction, my beta UPload seems to be stuck too.
15) Message boards : Number crunching : Need some help here please (Message 1488851)
Posted 39 days ago by Profile William
or see my post on how to get boinc to dump a full cc_config.xml into the right place.

the cc_config needs to be in the same folder as client_state.xml

when you have the file, right-click it and choose 'edit' this will open it in notepad (as opposed to just clicking which will open an xml editor and mess things up)

find the right option and change to 1
16) Message boards : Number crunching : how to debug? (Message 1488847)
Posted 39 days ago by Profile William
No offtopic or unhelpful posts please.
I neither want to have to look for the rolling pin nor do I want to search for a newspaper.
17) Message boards : Number crunching : how to debug? (Message 1488842)
Posted 39 days ago by Profile William
Ok, so, now let us try to get BOINC to dump the cc_config.xml file.

in Advanced View (that is the one that shows you the list of your tasks and has tabs like projects, tasks and transfers) go to

Tools -> Computing preferences

go to the tab 'exclusive applications'

click on 'Add'

you'll get a file selection screen looking at your Program directory.
Pick a folder from some program you use - any .exe will do, we are going to remove the entry again.

Pick a folder, double click tpo open if there are more folders inside open until you see an executable file. Doubleclick it.

The window should now have the name of the executable in the top box.

click 'ok' - the window will close and BOINC will generate a cc_config.xml in the right place.

now, we didn;t really want to have an exclusive app, so

tools -> computing preferences

it should open on the right tab.

click on the .exe entry you made and click 'remove'. click 'ok'

Ok, I think that's far too much in one go, so, start at the top and when you get stuck, holler.

NB if all goes well you now have a fully fledged cc_config.xml with every tag under the sun in it. Much more than you'll ever need, but that way you get a working copy in the right location.
18) Message boards : Number crunching : GPU work only? (Message 1488818)
Posted 39 days ago by Profile William
That's what different 'venues' are for.

on your host page http://setiathome.berkeley.edu/hosts_user.php you can see an entry 'location'.
click on the 'details' link for the one host that you want to be different.
at the bottom of the page there is an entry 'location' select 'work' 'school' or 'home' - doesn't matter which, or if that computer is actually at such a location. click 'update'

in your host list 'location' for that host should have changed to the location you picked.

Now, go to project preferences http://setiathome.berkeley.edu/prefs.php?subset=project and click on 'add separate preferences for <location>' make sure to pick the extra location you just assigned the host to.

In THAT subset of preferences, untick CPU.

Isssue 'update' from the host in question to get it to see the new location and preferences.


In addition, Notebooks benefit from using 'TThrottle' for temperature controlled running - it allows you to specify a maximum temperature for CPU and GPU and it will thrittle the boinc threads so that that temperature is not exceeded. Do NOT use boinc's 'use only x% of CPU time'.
19) Message boards : Number crunching : how to debug? (Message 1488811)
Posted 39 days ago by Profile William
'fascinating'

Ok, I need to switch boinc version, then I'll try to lead you through getting boinc to do the work for you.
20) Questions and Answers : Unix/Linux : CPU or GPU when running SETI@home on systems with CPU & GPU's ?? (Message 1488785)
Posted 39 days ago by Profile William
As far as I know there is no way to directly get information via the CLI whether a task was assigned to GPU or CPU.
You have to either check the website task list against the local task list and look which were assigne to GPU or you look in client_state.xml and check which tasks were assigned a (GPU) plan_class.

If a GPU was detected and if your preferences are set to ask for GPU tasks, the yes, both CPU and GPU should be running tasks.


Next 20

Copyright © 2014 University of California