Good comment already given from Raistmer and Paul.

Another view in reply:

Lynn W. Taylor wrote:
> This will be my final post on the subject.
> 
> To take the clock analogy one step further.
> 
> You're saying that your cesium clock is much better than the cheezy $1 
> westclox wind up I got at the local bargain shop.
> 
> ... and I'm saying that the cobblestone standard explicitly calls for 
> the use of a wind-up clock.
> 
> I agree with everything you've said.  I agree that Whetstones and 
> Dhrystones are not the best predictors of performance.

Indeed so, we agree on that as the present situation.


> But the standard says "you put the dhrystone number for your machine 
> here, and you put the whetstone number for your machine here, and you 
> put the number of CPU seconds here, and that's your credit, in 
> cobblestones."

And that is the "religious" what 'is', is. Amen.

Or we can be scientific and say that was a good first approximation that 
served it's purpose well enough. We can now move on and do better.


> Martin wrote:
>> Lynn W. Taylor wrote:
[...]
>>> .... starting with the fact that we aren't directly calculating 
>>> cobblestones.
>>
>> A little of the architectural influence can be read from a recent 
>> comparison article:
>>
>> http://www.tomshardware.com/reviews/athlon-l3-cache,2416.html
[...]
>>
>> Note in the benchmarks the identical dhrystone scores and yet some of 
>> the tests can show as much as a 20% real world performance difference.
>>
>> So... Do we put +/- 20% error bars on the credits cobblestones 
>> "depending" on 'whatever'... And that's before allowing for any 
>> approximations made for the FLOPs count for the s...@h-wu AR.

Is +/-20% near enough even for a bit of "fun"?

I suspect various (non-FLOPs) Boinc projects will see even worse than 
that...


I'll close on this well thrashed thread that I don't care about 
credits-as-bragging-rights, at all.

I will admit that my scientific sensibilities are badly irked by various 
non-FLOPs projects needing to award arbitrary credit according to such 
occultism as to how tea leaves might settle in the bottom of a morning 
brew. Credits or scientific discredit?

However, a useful measure of performance for comparison would be, well, 
useful. If only to be able to authoritatively claim that Boinc is 
equivalent to X supercomputers for a specific task.


We can't reliably assume processing effort as "credits = f( CPU time )" 
for unknown hosts. Especially so when those hosts are multitasking, or 
"hyper threaded", or even a virtual host lost in virtualisation. Some 
OSes do not even reliably report time. Further, the benchmarks are 
completely non-representative for GPUs.

Hence, I still put forward my vote for awarding credits for a WU by 
hierarchical calibration back to physical known hardware, our "Etalon" 
computer.


Happy crunchin',
Martin

( Good thread :-) )

-- 
--------------------
Martin Lomas
m_boincdev ml1 co uk.ddSPAM.dd
--------------------
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to