| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Cure Muscular Dystrophy - Phase 2 Forum
Thread: How many would be OK with a "Let it Rip" [OPTION] for HCMD2? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 62
|
|
| Author |
|
|
poppageek
Advanced Cruncher Joined: Nov 16, 2004 Post Count: 99 Status: Offline Project Badges: 
|
I have 30 cores 24/7 and have completed 2 of the large WUs. Works for me.
Let her rip! |
||
|
|
Hypernova
Master Cruncher Audaces Fortuna Juvat ! Vaud - Switzerland Joined: Dec 16, 2008 Post Count: 1908 Status: Offline Project Badges: 
|
After some thoughts I suppose that Movieman meant that he would have liked to have the WU's being multi-threaded. That would make sense for very long WU's that would then be crunched in parallel by the multiple cores.
----------------------------------------But here comes the dilemma: Either one large WU crunched in parallel by many cores, or smaller WU's running one per core. I suppose that it is much easier to code the software as one WU per core then a multithreaded WU. My past experience in software (I agree it was 20 years ago) is that we as human beings have a lot of difficulties in thinking parallel software except in very specific cases where it is an evident thing, like vectorial, or matrix calculation, and graphics which lead to these massively parallel gpu's like the ones of Nvidia and ATI. This is why it took so long to start having general purpose multicore CPU's and it was easier to just increase frequency. Now we must go the multicore solution as frequency increase is not any more an option due to heat, with the existing chip technology. Making best use of 8, 12, 16, or a 100 cores is no easy task at all. Massively parallel general purpose software is probably the reason why the birth of Larrabee is so difficult.  |
||
|
|
Jack007
Master Cruncher CANADA Joined: Feb 25, 2005 Post Count: 1604 Status: Offline Project Badges: 
|
I like Mysteron's solution. I'm glad you understood it, I've read it like 10 times and I feel dumber every time. But LET ER RIP I can understand  Edit: Mysteron please don't take offense, I only mean what I said about me, if the techs can understand (i'm sure they do) and if they can use it, Huzzah! And no need to break it down for me, I don't really have the need to know. I just like taxing my 12 GIGS of RAM on my overclocked PC, not my overtaxed brain...  P.S. I used to calculate those prime numbers on a single core machine, I think one was taking me up to 2 or 3 MONTHS! So BRING IT ON WCG, I laugh at your A type units(Dengue), 1 GIG of RAM apiece,  And if you do get some wunits that need more power, I'll have a good excuse to go shopping! Come on how bout a multithreaded app that I can run just ONE that takes like 4 GIGS by itself, or 8 even... How bout CEP2 being designed for the overclocking/nerdy/bored crowd.  Make it crazy highly powered so that we have to upgrade some more, then you have all that power when CEP2 is done for use on the regular projects. A I7980x is about $1900 from HP, 1999 from Dell Go ahead Make me SHOP, I dare you, I double dare you, I triple double dog dare you!  All your WorkUnits are belong to us! whoa I just reread that, gotta lighten up on the coffee   [Edit 1 times, last edit by Jack007 at Apr 1, 2010 2:51:46 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Maybe you've missed the word option in the title? So if it is an option, why the vote. If I can opt out for my slow machines (or any machine) then what is the big deal? If this is an attempt to get a groundswell of support for the Techs to add an option to the system, then I understand. If this reduces the actual total fpops needed to complete a whole WU (children included), then I agree with having an OPTION.If we do not need this option to complete the HCMD2 project, it seems to me that asking the Techs to provide NEW features to the system that CHANGE or ENHANCE the WCG system is far more important. Otherwise, where's the fire? |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
That I wrote [Feb 22, 2010 6:17:44 PM], 6 weeks ago. Think we need not further clarify what e.g. the 6 hour meets 6 hours quorum does with part of the work, depending on the speed/efficiency differential of the coupled devices, a smaller or larger part.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges:
|
If we do not need this option to complete the HCMD2 project, it seems to me that asking the Techs to provide NEW features to the system that CHANGE or ENHANCE the WCG system is far more important. I think that was the point.I have not taken Sekerob's question as a vote but as a survey to see how many people would use it. Personally I have not answered Sek because I would not use it on my P4: that would make HCMD2 WUs last much longer than I wish. And I would have checked it on the quad "just in case" only, because I have one WU stopped at 6 hours only once in a while, and I remember only one stopped at 12 hours. The quad completes almost all its HCMD2 WUs. Bottom line: I don't really need to have the techs spending time on this feature.  ---------------------------------------- [Edit 1 times, last edit by JmBoullier at Apr 1, 2010 5:22:28 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Jean, anytime a wingmen to your speed-daemon results does cut-off at 6 hours or 12 hours, part of your task, be it run in less or more than the standard 6 hours, but doing all positions, is discarded. I see it frequently, the non-perfect match.
----------------------------------------
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Apr 1, 2010 5:38:21 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Jean, anytime a wingmen to your speed-daemon results does cut-off at 6 hours or 12 hours, part of your task, be it run in less or more than the standard 6 hours, but doing all positions, is discarded. I see it frequently, the non-perfect match. Are you saying that you have seen WU being classed as an Error in this case? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I have not taken Sekerob's question as a vote but as a survey to see how many people would use it. So you agree with my identifying this as seeing if the users want to generate a ground swell of support. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Sorry astrolab, but I thought it was general knowledge, or I totally misunderstand you, so apologies if you know this already: Extremely simplified, 2 HCMD2 tasks meet, one does 400 positions in 6 hours, the other 600 positions in 6 hours, neither 2 doing the full load nor has reached 60% of total positions in the task. 400 are used, the part for which quorum exists. Extra credit is given also to the faster machine for the extra 200, BUT these other 200 are repackaged and send out again in a child result as absolute quorum for each position is required.
----------------------------------------WCG tries to match devices on speed/efficiency to reduce the position overage with one of the 2... get it as close to both doing the same number of positions, some harder than others. The object of this suggestion was to let it run without time limit so there will never be a need to discard a single position in a pair.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|