Never seen before: only 2 results are valid?
30 Mar 2005 23:20:04 UTC
Topic 188809
(moderation:
any idea, what happened here, anybody?
http://einsteinathome.org/workunit/638564
the last one is mine, #70831
Never seen before: only 2 results are valid?
)
> any idea, what happened here, anybody?
>
> http://einsteinathome.org/workunit/638564
> the last one is mine, #70831
That's the same thing that all the threads about linux results being invalid are talking about - only this was one of those rare exceptions where the linux/mac results got turned in first so they got validated and the windows results didn't.
> any idea, what happened
)
> any idea, what happened here, anybody?
>
> http://einsteinathome.org/workunit/638564
> the last one is mine, #70831
Yes, yours was the last one in. If you had just beaten the last unix user then the windows boxes would have prevailed and the poor old minority unix users would have lost out yet again.
Someone has really got to fix this in a big hurry, particularly for the sake of the unix community out there. They are being punished with an inefficient science app and they are also unjustly losing credit!! Also look at how low their credit actually was on this one.
I've seen a few messages asking about progress in fixing the validator. Someone really needs to bring us up-to-date on this one.
IS ANYBODY LISTENING!!!
Cheers,
Gary.
Running Linux. The vast
)
Running Linux. The vast majority of my recent results are 0.00.
I really hate to do this, but, it looks like i'm going to have to detach my boxes from this project until the Linux thing is rectified.
The cycles would certainly be better spent on projects that can actually use them.
No, WINE is not an option (for various reasons). Nor should it have to be. Don't even mention it.
Don't take this the wrong way, please. I'm not the least bit upset. I'm just being realistic.
> Running Linux. The vast
)
> Running Linux. The vast majority of my recent results are 0.00.
>
have you tried the Boinc CC 4.27, I don't know if that'd fix it for you.
just a thought
tony
[edit] you can get it here
second edit: this might not help. If you're discussing the validation process of einstein excluding results returned On time, but after the granting of credits. Oh well
sorry
> Running Linux. The vast
)
> Running Linux. The vast majority of my recent results are 0.00.
>
> I really hate to do this, but, it looks like i'm going to have to detach my
> boxes from this project until the Linux thing is rectified.
This is strange. The majority of my machines run Linux. When I look at my results, then I see only 2 results which got the magic 0.00 credit.
But in the team there are some other guys which stoped working for the same reason as you do. So what is the reason? It does not seem to be the Linux. Maybe it has something to do with stoping/restarting the work. Maybe some special glibc, some distribution?
> This is strange. The
)
> This is strange. The majority of my machines run Linux. When I look at my
> results, then I see only 2 results which got the magic 0.00 credit.
Until very recently, I was in the same position. It's quite odd.
> But in the team there are some other guys which stoped working for the same
> reason as you do. So what is the reason? It does not seem to be the
> Linux. Maybe it has something to do with stoping/restarting the work. Maybe
> some special glibc, some distribution?
I don't really think it is distro specific. I run Red Hat 9.0, Fedora Core 3, Slackware 9 and Slackware 10.1. The problem seems, well, agnostic.
I will say that I had an early bout of this on Predictor, but, they seem to have resolved the problem. Climate Control and SETI just work and always have (except for when they don't :) ).
> > Running Linux. The vast
)
> > Running Linux. The vast majority of my recent results are 0.00.
> >
> have you tried the Boinc CC 4.27, I don't know if that'd fix it for you.
I run 4.27 on all boxes.
This seems to be Einstein specific as the other projects I am attached to are running flawlessly.
> second edit: this might not help. If you're discussing the validation process
> of einstein excluding results returned On time, but after the granting of
> credits. Oh well
Nope. This has to do with not only the zero credit thing, but, with the slow computation times of the Linux client as well.
Definite problems exist where Einstein's Linux client is concerned. I'm inclined to believe it is a gcc-related problem as I experience those in my own code from time to time. Its' optimization tends to be less than...well...optimal.
Perhaps someone from Einstein will try Intel's compiler at some point?
I also got the majority of my
)
I also got the majority of my Systems running Linux.
It's really a waste, especially when those double-time Results are additionally even discarded.
I'm losing about 30% of my Linux Results due to that, yet I haven't seen or heard of any updates concerning the issue...
Within the last 3 weeks, that accounted for some 300 CPU hours wasted.
In that time, I could have otherwise accomplished :
~100 SETI WorkUnits
~125 PAH WorkUnits
~65% of a CPDN Model
> Within the last 3 weeks,
)
> Within the last 3 weeks, that accounted for some 300 CPU hours wasted.
> In that time, I could have otherwise accomplished :
> ~100 SETI WorkUnits
> ~125 PAH WorkUnits
> ~65% of a CPDN Model
>
Both parties lose in this. Linux users are wasting their time and Einstein is wasting their bandwidth.
While we treat this as an informal competition sometimes, it *is* research. Using a flawed tool is counter-productive. I'd make more sense to me if Einstein were to simply pull the Linux client out of circulation until it is fixed or drop it altogether.
Of course, my being totally off base is not outside the realm of possibility.
> > Within the last 3 weeks,
)
> > Within the last 3 weeks, that accounted for some 300 CPU hours wasted.
> > In that time, I could have otherwise accomplished :
> > ~100 SETI WorkUnits
> > ~125 PAH WorkUnits
> > ~65% of a CPDN Model
> >
>
> Both parties lose in this. Linux users are wasting their time and Einstein is
> wasting their bandwidth.
full ack
it really is a pitty, to say the least.
I stopped running Einstein on linux - its not worth the effort - as FalconFly already pointed out.
>
> While we treat this as an informal competition sometimes, it *is* research.
> Using a flawed tool is counter-productive.
it is a big mess.
some rumours say, there is at least some progress with the performance, validation and exact matching of results I do not know.