Very informative, thanks. Very thorough "roadmap" of what we have to expect.
Does anyone have an idea about my signal 11s yet? I sure hoped they'd stop when I stopped running Einstein as root (not so much because of the credit, but I'm not sure how many other Linux users are affected... when I have a look at some results, I tend to see quite a few computation errors on Linux boxes). And, as I said, I'd like to provide the debugger output but I'm not quite sure how...
Currently the SSE code can only be compiled with gcc; and when linking gcc code to MSC code the debugging information of the gcc objects get lost. Therefore I find it rather unlikely that there will be a Windows App with an SSE 'hot-loop', say, this year.
"This year" is only 2 weeks Bernd... :-P
Would it be fair to say "not within 2 months"?
If that is the case, then how about extending deadlines again to the range of 18-21 days so as to help out the Pentium II / Pentium III class systems that sometimes miss the 14-day deadline by only a little bit?
Also, if I fire up a VM host, I would eventually be able to delete that host once all results cleared, right?
Thanks for the answers, I think I'll just stick to the 'mass-audience' Application for then and just wait that comes along ;)
Would be really neat to have the Server offer various SIMD optimized Clients, just takes more time per Version to compile for all of the Sets existing (it's gotten quite a load actually). That would be a perfect use of the available BOINC feature (CPU caps detection), I imagine other Projects would quickly go the same road once implemented and proven working.
Thanks for the answers, I think I'll just stick to the 'mass-audience' Application for then and just wait that comes along ;)
Would be really neat to have the Server offer various SIMD optimized Clients, just takes more time per Version to compile for all of the Sets existing (it's gotten quite a load actually). That would be a perfect use of the available BOINC feature (CPU caps detection), I imagine other Projects would quickly go the same road once implemented and proven working.
Well, the problem with it is that it requires BOINC version 5.6 and newer. There are many people out there still happily hammering away with BOINC 4.xx or something in the 5.2.xx or 5.4.xx range... Otherwise, I would guess the host would need to be sent a "generic" application, with only FPU support...
If that is the case, then how about extending deadlines again to the range of 18-21 days so as to help out the Pentium II / Pentium III class systems that sometimes miss the 14-day deadline by only a little bit?
A two weeks deadline is adequate for my PII running Linux 24/7 with 300 share. SETI gives me deadlines extending two months, QMC deadlines varie.
Tullio
If that is the case, then how about extending deadlines again to the range of 18-21 days so as to help out the Pentium II / Pentium III class systems that sometimes miss the 14-day deadline by only a little bit?
A two weeks deadline is adequate for my PII running Linux 24/7 with 300 share. SETI gives me deadlines extending two months, QMC deadlines varie.
Tullio
That's because you're running 24/7, apparently have reasonable cache settings, and because SETI's deadlines are grossly excessive for many of their tasks. See this thread over there. Your 2 month deadlines could be reduced down to 20-30 days. If that were to come to pass, you may find that there is deadline pressure on your host. Your last Einstein result took 3.6 days to complete. The QMC task you have showing took 8.75 days. That is already 12.35 days in CPU clock. Wall clock (actual time) is going to be somewhat longer than that (the CPU clock divided by whatever your cpu efficiency values for the projects are)...
This is somewhat OT for the specific purpose of this thread, but I mention all of this because SETI's excessive deadlines have caused ripple effects out to other projects. The effect here is people complaining about how Einstein is not "playing fair" because BOINC moves the Einstein tasks into EDF/High Priority mode due to not running 24/7 and/or having a lower resource allocation. The complaint then is centered around the fact that they think "Einstein" is "hogging their system" because the local cpu scheduler is no longer marching second-by-second in lockstep with resource allocations.
Anyway, the deadlines at SETI are way too generous in many cases. My gut feeling is that the deadlines here are a bit too low, but I don't have the knowledge or the time to pour into doing the research like Joe, Richard, and Fred are doing over there...
4.21
89905240 35986038 19 Dec 2007 3:16:58 UTC 20 Dec 2007 9:57:45 UTC Over Success Done 42,473.16 235.06 pending
89896177 35981871 18 Dec 2007 21:34:00 UTC 20 Dec 2007 1:53:17 UTC Over Success Done 42,561.25 235.06 235.06
89881033 35975401 18 Dec 2007 14:54:25 UTC 19 Dec 2007 9:38:49 UTC Over Success Done 41,622.94 235.06 pending
4.20
89761403 35921601 15 Dec 2007 9:54:31 UTC 18 Dec 2007 14:31:51 UTC Over Success Done 39,564.02 235.06 235.06
89729984 35907704 14 Dec 2007 14:43:29 UTC 18 Dec 2007 0:50:28 UTC Over Success Done 39,096.87 235.06 235.06
89718595 35902532 14 Dec 2007 5:47:44 UTC 18 Dec 2007 0:50:28 UTC Over Success Done 39,695.32 235.06 235.06
Why is your computers page showing that as a Phenom 9500?
Anyway, keep in mind that there is still a runtime variation that is cyclic and that it could've been worse with 4.20. I think you'd need to have at least 10-15 results or so before saying that it is no better than 4.20...
Why is your computers page showing that as a Phenom 9500?
Anyway, keep in mind that there is still a runtime variation that is cyclic and that it could've been worse with 4.20. I think you'd need to have at least 10-15 results or so before saying that it is no better than 4.20...
My Phenom has been delivered today and I replaced the X2. :-)
Why is your computers page showing that as a Phenom 9500?
Anyway, keep in mind that there is still a runtime variation that is cyclic and that it could've been worse with 4.20. I think you'd need to have at least 10-15 results or so before saying that it is no better than 4.20...
My Phenom has been delivered today and I replaced the X2. :-)
Very cool, will be nice to see how the Phenom is performing. I'm surprised that BOINC isn't considering the upgraded PC as a completely new machine (and I thought you could not merge machines with different CPUs).
Very informative, thanks.
)
Very informative, thanks. Very thorough "roadmap" of what we have to expect.
Does anyone have an idea about my signal 11s yet? I sure hoped they'd stop when I stopped running Einstein as root (not so much because of the credit, but I'm not sure how many other Linux users are affected... when I have a look at some results, I tend to see quite a few computation errors on Linux boxes). And, as I said, I'd like to provide the debugger output but I'm not quite sure how...
RE: Currently the SSE code
)
"This year" is only 2 weeks Bernd... :-P
Would it be fair to say "not within 2 months"?
If that is the case, then how about extending deadlines again to the range of 18-21 days so as to help out the Pentium II / Pentium III class systems that sometimes miss the 14-day deadline by only a little bit?
Also, if I fire up a VM host, I would eventually be able to delete that host once all results cleared, right?
Thanks for the answers, I
)
Thanks for the answers, I think I'll just stick to the 'mass-audience' Application for then and just wait that comes along ;)
Would be really neat to have the Server offer various SIMD optimized Clients, just takes more time per Version to compile for all of the Sets existing (it's gotten quite a load actually). That would be a perfect use of the available BOINC feature (CPU caps detection), I imagine other Projects would quickly go the same road once implemented and proven working.
RE: Thanks for the answers,
)
Well, the problem with it is that it requires BOINC version 5.6 and newer. There are many people out there still happily hammering away with BOINC 4.xx or something in the 5.2.xx or 5.4.xx range... Otherwise, I would guess the host would need to be sent a "generic" application, with only FPU support...
RE: If that is the case,
)
A two weeks deadline is adequate for my PII running Linux 24/7 with 300 share. SETI gives me deadlines extending two months, QMC deadlines varie.
Tullio
RE: RE: If that is the
)
That's because you're running 24/7, apparently have reasonable cache settings, and because SETI's deadlines are grossly excessive for many of their tasks. See this thread over there. Your 2 month deadlines could be reduced down to 20-30 days. If that were to come to pass, you may find that there is deadline pressure on your host. Your last Einstein result took 3.6 days to complete. The QMC task you have showing took 8.75 days. That is already 12.35 days in CPU clock. Wall clock (actual time) is going to be somewhat longer than that (the CPU clock divided by whatever your cpu efficiency values for the projects are)...
This is somewhat OT for the specific purpose of this thread, but I mention all of this because SETI's excessive deadlines have caused ripple effects out to other projects. The effect here is people complaining about how Einstein is not "playing fair" because BOINC moves the Einstein tasks into EDF/High Priority mode due to not running 24/7 and/or having a lower resource allocation. The complaint then is centered around the fact that they think "Einstein" is "hogging their system" because the local cpu scheduler is no longer marching second-by-second in lockstep with resource allocations.
Anyway, the deadlines at SETI are way too generous in many cases. My gut feeling is that the deadlines here are a bit too low, but I don't have the knowledge or the time to pour into doing the research like Joe, Richard, and Fred are doing over there...
My Athlon 64 X2 5000+ doesn't
)
My Athlon 64 X2 5000+ doesn't like it. :-(
4.21
89905240 35986038 19 Dec 2007 3:16:58 UTC 20 Dec 2007 9:57:45 UTC Over Success Done 42,473.16 235.06 pending
89896177 35981871 18 Dec 2007 21:34:00 UTC 20 Dec 2007 1:53:17 UTC Over Success Done 42,561.25 235.06 235.06
89881033 35975401 18 Dec 2007 14:54:25 UTC 19 Dec 2007 9:38:49 UTC Over Success Done 41,622.94 235.06 pending
4.20
89761403 35921601 15 Dec 2007 9:54:31 UTC 18 Dec 2007 14:31:51 UTC Over Success Done 39,564.02 235.06 235.06
89729984 35907704 14 Dec 2007 14:43:29 UTC 18 Dec 2007 0:50:28 UTC Over Success Done 39,096.87 235.06 235.06
89718595 35902532 14 Dec 2007 5:47:44 UTC 18 Dec 2007 0:50:28 UTC Over Success Done 39,695.32 235.06 235.06
RE: My Athlon 64 X2 5000+
)
Why is your computers page showing that as a Phenom 9500?
Anyway, keep in mind that there is still a runtime variation that is cyclic and that it could've been worse with 4.20. I think you'd need to have at least 10-15 results or so before saying that it is no better than 4.20...
RE: RE: My Athlon 64 X2
)
My Phenom has been delivered today and I replaced the X2. :-)
RE: RE: RE: My Athlon
)
Very cool, will be nice to see how the Phenom is performing. I'm surprised that BOINC isn't considering the upgraded PC as a completely new machine (and I thought you could not merge machines with different CPUs).
CU
Bikeman