Message boards :
Number crunching :
AP validation dilemma
Message board moderation
Author | Message |
---|---|
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14654 Credit: 200,643,578 RAC: 874 |
Got the first one of an expected class of WUs - WU 364288117. Ignore the computation error - that's not important. The other two are, to all intents and purposes, identical - both Intel Core2 running Windows XP SP3, BOINC v6.2.19, and the stock application. Fine upstanding SETIzens, the pair of them. They didn't validate because one of them was computed with AP v4.36, and the other with AP v5.00: we thought they wouldn't validate against each other, and here's the proof. Because I've been so involved with the testing, I know how to run my task with either v4.36 or v5.00. So I can validate and grant credit to either one of my fellow communards, but probably not both. Which should I choose, and why? |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
Version 5. 4.36 is superseded by new stock, so we must go by the current science. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Wandering Willie Send message Joined: 19 Aug 99 Posts: 136 Credit: 2,127,073 RAC: 0 |
Hi Richard, I would say let it run and see what the outcome is after all it could go to both, or give it to AP 5.00 as it is the latest. Smacks of semi favourtism otherwise. The decision in the end is yours and yours alone. Michael |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Well, my personal opinion is to suspend "production" AP crunching on main as I already said in Lunatics forums. But note, that it's still my personal opinion only. While "validation" word under suspiction there is not much to change beside credits will you give "validation" to one or to another. |
Leaps-from-Shadows Send message Joined: 11 Aug 08 Posts: 323 Credit: 259,220 RAC: 0 |
Well, my personal opinion is to suspend "production" AP crunching on main as I already said in Lunatics forums. So you think we shouldn't do any v5 work units at all - with standard or optimized apps? Have the work units themselves changed? I ask because if they haven't, we could simply switch back to using v4 apps to crunch them. I just want to make sure I'm not wasting CPU cycles crunching work units that don't add to the science. My RAC would certainly thank me if I stopped crunching Astropulse units, but that's not what this is about... Cruiser has two v5 units in progress on Main, two in progress on Beta (actually, one completed but not reported and one in progress), plus seven in Main's queue and two in Beta's queue. I have suspended the seven queued units on Main until this is cleared up. Cruiser Gateway GT5692 L-f-S Edition -Phenom X4 9650 CPU -4GB 667MHz DDR2 RAM -500GB SATA HD -Vista x64 SP1 -BOINC 6.2.19 32-bit client -SSE3 optimized 32-bit apps |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
So you think we shouldn't do any v5 work units at all - with standard or optimized apps? My response at Lunatics:
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Fred J. Verster Send message Joined: 21 Apr 04 Posts: 3252 Credit: 31,903,643 RAC: 0 |
Hi, so it wouldn't be 'wise' to install either off the optimized AP apps. right now and wait until this is cleared? On 1 rig, a P4 D820, I run NON-OPtimized apps. and crunch MB and AP, also EINSTEIN, this way I can view the graphics (screensaver; boincscr.exe). This is the 'slowest host', I (still) run. |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
It isn't the [Windows] optimised apps in question, they validate fine against stock so far (as far as we can tell). It's some perceived anomalous behaviour from all the apps with regard to introduction of some software radar blanking yielding unexpected signal counts. There is much more analysis to be done IMO, to determine whether these are truly anomalous, or some other unexpected phenomena. "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
1) Opt app now is as good (or as bad) as stock V5, just considerably faster. So if you wanna crunch, do it with opt app. 2) returning to V4.x has no meaning at all cause stock is V5.0 and server validator will mostly refuse validation between v5 and v4 (situation described in first post). So either stop either continue with V5. |
W-K 666 Send message Joined: 18 May 99 Posts: 19100 Credit: 40,757,560 RAC: 67 |
As the vast majority of users never visit here, so any attempt to stop tasks being processed by V5 is futile. So all we can do is to try and contact Eric or Josh and get them to sort out this mess. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14654 Credit: 200,643,578 RAC: 874 |
OK, the dilemma is over. I decided to crunch it with a v5 (optimised) application, and the result is: 1056385822 4589383 13 Nov 2008 13:27:21 UTC 25 Nov 2008 11:44:03 UTC Over Success Done 178,391.50 760.39 0.00 The first task is the stock v4.36, the last is the stock v5.00 A similar thing happened to the first of Geek@Play's jobs (mis)reported in the AK_V8 issues thread: WU 352088249 1065468267 3218838 21 Nov 2008 8:00:49 UTC 26 Nov 2008 19:12:02 UTC Over Success Done 51,769.95 759.31 759.31 The first two are v5.00 (optimised and stock, respectively): the final one is v4.36 stock. I think we can take it that v4.36 results are not even weakly similar to v5.00 results. Under normal circumstances, I would try to urge Eric to grant credit manually to the v4.36 tasks which have lost out, but all of these will be long purged from the database and untraceable before the long holiday weekend is over. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.