IRC logs of #boinc for Sunday, 2012-07-29

00:21 *** BadBarbarian has joined #boinc

00:42 *** Simone1 has joined #boinc

00:45 *** Simone2 has quit IRC

00:50 *** paul_l has joined #boinc

03:00 *** Simone1 has quit IRC

03:20 *** efc has quit IRC

04:02 *** pppingme has quit IRC

04:19 *** yoyo[RKN] has joined #boinc

05:25 *** Caterpillar has joined #boinc

05:48 *** desti_T2 has joined #boinc

05:52 *** desti has quit IRC

07:33 *** si has joined #boinc

09:08 *** mapreri has joined #boinc

09:58 *** Tank_Master has quit IRC

09:59 *** Tank_Master has joined #boinc

09:59 *** BadBarbarian has quit IRC

10:15 *** AKDM has quit IRC

11:47 *** AKDM has joined #boinc

11:53 *** _Turf has joined #boinc

11:54 *** Turf has quit IRC

11:59 *** _Turf has quit IRC

11:59 *** Turf has joined #boinc

14:14 *** mapreri has quit IRC

14:37 *** _Danilo_ has quit IRC

14:49 *** _Danilo_ has joined #boinc

15:59 *** si has quit IRC

16:50 *** whynot has joined #boinc

16:50 *** yoyo[RKN] has quit IRC

16:57 *** efc has joined #boinc

17:51 *** AKDM has quit IRC

17:52 *** whynot has quit IRC

17:57 *** Caterpillar has quit IRC

18:02 *** efc has quit IRC

18:40 *** JamesB192 has quit IRC

19:37 *** BadBarbarian has joined #boinc

19:47 *** pppingme has joined #boinc

19:58 *** pppingme has quit IRC

20:03 *** pppingme has joined #boinc

20:41 <paul_l> what do you people do in here?

20:46 <paul_l> been idling for a day and nothing

20:46 <paul_l> yes, I boinc, have a farm with gpu's etc.

20:49 *** JeroenDeDauw has joined #boinc

20:49 *** JeroenDeDauw has left #boinc

20:53 <Tank_Master> !wx hio

20:53 <Tank_Master> &wx hio

20:53 <Romulus> Tank_Master: Temperature: 76F / 24C | Humidity: 46% | Pressure: 30.09in / 1019hPa (Falling) | Conditions: Partly Cloudy | Wind Direction: North | Wind Speed: 0mph / 0km/h ; Sunday - Mostly cloudy in the morning, then partly cloudy. High of 79F. Winds less than 5 mph.; Sunday Night - Mostly cloudy in the evening, then partly cloudy. Low of 52F. Winds less than 5 mph.; Monday - Mostly cloudy. High of (1 more message)

20:53 <Tank_Master> &more

20:53 <Romulus> Tank_Master: 75F. Winds from the NNW at 5 to 10 mph.;

20:54 *** _Danilo_ has quit IRC

20:54 * Tank_Master pushes xcamel into a crevasse

20:56 <paul_l> Romulus, what do these people do in here?

20:56 <Romulus> Bugger all, I dunno.

20:56 <paul_l> me either

20:56 <BadBarbarian> now thats funny

20:57 <Tank_Master> sometimes bots just nail it

21:37 <RomW> :)

21:38 <RomW> I've been working on next gen UI perf inprovments

21:38 <RomW> improvements

21:53 <paul_l> Rom Walton?

21:59 <paul_l> I'll be darn...

21:59 <RomW> Yeah

21:59 <RomW> :)

21:59 <RomW> I show up every once and a great while

22:01 <paul_l> a pleasure to meet you. Been using your work since the beginning.

22:02 <RomW> Please to meet you as well

22:02 <RomW> Hmmm,  actually, I think there are those in IRC that predate even me...

22:03 <RomW> :)

22:06 <paul_l> good thing we age well in IRC..

22:06 <RomW> lol

22:08 <paul_l> man, I remember beinng one of the handful of people on a large team to beta test the original BOINC, something like .1x

22:09 <RomW> Chuckle, IIRC, BOINC had a very long Beta test cycle

22:19 <RomW> @paul, How far back was that? 03? 04?

22:21 <paul_l> I'd be lying if I said, used to keep all the boinc versions I used on a data drive, but I accidently wiped it a while back

22:22 <RomW> :)

22:22 <RomW> It happens

22:22 *** efc has joined #boinc

22:23 <RomW> Well, I don't think I started with the project until the open beta was announced

22:23 <RomW> Which was at the end of 03

22:40 <paul_l> just tried to search some really old forum posts for a date, got a headache

22:42 <RomW> no worries

22:42 <RomW> Things went crazy during the initial S@H launch

22:42 <RomW> some things were retired

22:43 <RomW> They may not even be publically available anymore

22:44 <efc> there were a lot of tools that worked with the old architecture, the original circa '99 client

22:45 <paul_l> oh yeah was just looking at posts about SetISpy, SetiDriver

22:45 <paul_l> fun times

22:46 <efc> some of the management tools still work. think i have one of those.

22:46 <efc> BOINCView. Still works.

22:47 <PovAddict> the new generation is busy making iPhone apps and Facebook games

22:48 <efc> i get the impression that BOINC is harder to deal with, too.

22:48 <efc> at any rate it took away or made unnecessary some of the options those tools used

22:48 <PovAddict> I have no idea about the old SETI client, so I dunno

22:49 <efc> Queue management being a big one

22:49 <PovAddict> did the classic client have a queue at all? :)

22:51 <efc> i can't remember. I do remember groups grabbing large numbers of WUs, and keeping the results, then dumping them en masse

22:51 <efc> it wasn't the most helpful thing for the network..

22:51 <RomW> Nope, SETI Classic didn't have a queue

22:52 <RomW> That was one of the big design changes BOINC introduced

22:52 <paul_l> SetiQue was one were you could grab a bunch and host your own queue

22:53 <RomW> Yeah, there were addons to do that...  Some of the addons let you cherry pick workunits too

22:53 <efc> yeah, i remember there were "sweet" WUs

22:53 <efc> certain slew angles or something

22:54 <RomW> yeah

22:54 <efc> the queue deal was, the servers were unreliable, so teams would grab mass quantities so they'd have work, which tended to perpetuate the problems

22:55 <RomW> IIRC, somebody had written a fancy queue management tool that you could use to process a wu to 98% complete and then re-farm out to all of your machines to complete the last 2% to increase wu counts

22:57 <PovAddict> omg

22:57 <PovAddict> that's blatant cheating

22:57 <RomW> lol

22:57 <PovAddict> hmm

22:57 <paul_l> oh yeah, then y'all caught a bunch of them and outed their accounts

22:58 <RomW> Well, S@H classic didn't have any of the safegards that are in place for BOINC

22:58 <PovAddict> I wonder how much time on the K Computer would equate to a year of SETI@Home volunteers

22:59 <PovAddict> wow

23:00 <PovAddict> apparently 1 year of the SETI@Home grid (769 TFLOPS as of Nov 2009, according to Wikipedia) is equivalent to 34 days on the K computer

23:01 <RomW> IIRC that task assignment mechinism was just the client during a 'dir' on a certain web server and randomly choosing a wu to process

23:02 <RomW> every 30 minutes or so a background daemon would change out the files

23:02 <PovAddict> 17 days on the IBM Sequoia, damn

23:02 <RomW> Chickle

23:03 <RomW> Chuckle, add in the recent numbers, with GPUs

23:03 <RomW> nevermind

23:03 <RomW> its gone down

23:04 <PovAddict> where can I see the project's FLOPS?

23:04 <RomW> http://boincstats.com/en/stats/0/project/detail

23:04 <Romulus> Title: BOINCstats/BAM! | SETI@Home - Detailed stats (at boincstats.com)

23:05 <RomW> http://boincstats.com/en/stats/-1/project/detail

23:05 <Romulus> <http://tinyurl.com/c9v5mdl> (at boincstats.com)

23:05 <PovAddict> so a S@H-year is 13.3 days on the Sequoia

23:05 <PovAddict> and the Sequoia will be primarily used for nuclear weapon simulation

23:06 <RomW> willy is da man when it comes to BOINC stats

23:06 <PovAddict> I think I'll go cry myself to sleep

23:06 <efc> I sort of wonder if the s@h endeavor makes sense now; if the same money went into a cluster, would it be faster..

23:07 <efc> We could piece together some of ILLIAC IV and set it to work.

23:07 <PovAddict> efc: how much of a cluster can you make with $100k?

23:08 <efc> 150 GPUs maybe? And your software is a lot simpler

23:09 <efc> Don't have to do all the work 3 times to find cheaters

23:09 <RomW> I'm not sure S@H uses the 3 times rules anymore

23:09 <PovAddict> afaik it's 2x now

23:10 <RomW> I think they turned on adaptive replication

23:10 <PovAddict> initial_replication = min_quorum = 2

23:10 <RomW> ah

23:10 <RomW> okay, so maybe not

23:11 <PovAddict> no, maybe they do

23:11 <PovAddict> maybe they go for 3x if one of the clients is untrusted

23:12 <PovAddict> efc: how much money would you need to make a 700 tflops cluster? :)

23:12 <RomW> Oooo BOINC is floating around 6 PetaFLOPS as a whole

23:13 <PovAddict> well I only care about the order of magnitude of those numbers...

23:13 <PovAddict> they're based on credits, I see big reasons for them to be BS :)

23:13 <efc> I dunno. Also dunno how much money really goes into s@h (maybe 100k of cash, but they get some equipment donations)

23:13 <PovAddict> efc: S@H's budget is 400k

23:14 <efc> For projects not so floating point bound, i think FPGAs could really do great

23:14 <RomW> Somebody has started to work on that

23:14 <PovAddict> indeed

23:15 <efc> i did some work on one, but gave up, also after the GPU s@h came out it wasn't such a win

23:15 <RomW> I would like to see how we are going to handle benchmarks on one of those things

23:15 <RomW> lol

23:15 <PovAddict> efc: you spent most of the time looking at how the existing app worked

23:15 <efc> yeah

23:15 <PovAddict> efc: I think you should write documentation for it

23:16 <PovAddict> you don't get to make a FPGA app but at least something useful to the project comes out of your efforts :)

23:17 <efc> i wonder how that would work. I spent a lot of time on similar documentation for another open source project, and it was like pulling teeth to get them to take it.

23:17 <RomW> Wiki

23:17 <RomW> :)

23:17 <PovAddict> depends on the project...

23:17 <RomW> You want create page access to the wiki?

23:18 <RomW> oh wait

23:18 <efc> the s@h code would be pretty decent if they would just label variables with units

23:18 <RomW> I can't do anything about S@H

23:18 <PovAddict> a few months ago, I wondered how the multiple servers behind anongit.kde.org worked, and how they synced from git.kde.org

23:18 <RomW> lol

23:18 <PovAddict> so I read some scripts, asked some questions to the KDE sysadmins

23:19 <PovAddict> I somehow ended up writing 700 words :P

23:21 <efc> i should start getting my findings on various stuff on the web somehow. I figure out many things and am just to lazy to post it.

23:22 <PovAddict> another admin said it was "the best documentation he had seen in a while"... and then people wonder why I spend time on free software instead of, say, studying for exams

23:23 <efc> as I read once.. Documentation is like sex, even when it is bad, it is better than nothing.

23:24 <RomW> roflmao

23:25 <RomW> Depends on what is being documented

23:25 <RomW> If the documentation is on a function, and how the function works has totally changed, then that is worse than not having any documentation

23:27 <efc> yeah, hence I think the best place for docs is in the source. Anything else is always out of date.

23:27 <RomW> :)

23:27 <PovAddict> I have ideas for a documentation markup format with embedded 'unit tests'... I have yet to write anything though

23:28 <efc> We have that problem at work. Its the kind of place that produces multi-hundred page design and requirements documents. Ugh.

23:28 <RomW> oooo

23:29 <RomW> speaking of which, I need to run doxy against some source

23:29 <PovAddict> Click on Attach Ptoject <% assert mgr.menubar.hasItem("Attach Project") %>

23:31 <PovAddict> if the menu item changes, you'd be told of the documentation assert failures, which ideally would point you to everything you have to update to match

23:40 <PovAddict> unfortunately it's not very high in my todo/ideas list (which is long enough for a lifetime already)

23:42 <efc> Testing/Docs, not very much fun.

23:42 <PovAddict> efc: what units does the S@h code use? time, frequency and power shouldn't be that much of a mess...

23:43 <efc> Yes, those would be nice. power is in dBm? who knows.

23:44 <efc> Also, intermediate quantities are hard to figure out. They tend to be in funky unit combinations.

23:44 <PovAddict> I saw a C++ library for units, supporting all base SI units and all possible combinations of them (some of them with defined names)

23:49 <PovAddict> the base class is a template taking seven parameters (the exponents of each base unit)

23:51 <RomW> @Pov, Have you ever stumbled across an XML parser that uses static hash maps before?

23:57 <PovAddict> meaning a hash map that isn't generated at runtime from dynamic data?

23:57 <PovAddict> I think I have seen it for other kinds of parsers

23:57 <RomW> Well, the name/value pairs are already known and generated

23:58 <RomW> Talk about some huge gains

23:58 <RomW> wow

Generated by irclog2html.py 2.4 by Marius Gedminas - find it at mg.pov.lt!