John seems to have a fleet of old slow computers as well ...

I am pretty sure Jack at Collatz has a reference machine at that  
project as well ...

But the bottom line is that if you do set-up the calibration and get  
the data then the dependence on the primary references can be broken  
(with a minor loss of accuracy) by substituting a later generation  
machine.

In that, contrary to rumor, the interest is in getting a stable and  
reasonably accurate system ... not a perfect system that 100 years  
from now we can prove a deviation of 0.0000001% or less ...

If there was one weakness in the proposal I put up was that I did not  
do the part that Dr. Anderson did in his ... in part because I  
couldn't get anyone to read the base proposal, adding in obtuse math  
would not have made it more attractive.  But the core of what he  
suggested is pretty much what I had in mind ...

On Sep 29, 2009, at 10:09 AM, Martin wrote:

> Martin wrote:
> [...]
>> The main aim is to eliminate the need for adding extra code into the
>> clients just for the sake of adding up FLOPS or IOPS or whatever.
>> Further, the aim is to also eliminate the need for adding /any/
>> 'performance instrumentation' into the clients. (Optimisations are
>> inherently allowed for also.)
>>
>> Referencing back to known hardware looks to be a lot easier, and also
>> offers the chance to be "scientific" about the credits. We can then  
>> do
>> meaningful performance comparisons for hardware, algorithms, or
>> whatever... Who knows what new Computer Science can be then  
>> uncovered.
>>
>> The credits freaks then also get a 'currency' that is stable and
>> referenced back to known hardware/performance.
>>
>> Better still, all the calibration can be coordinated purely server- 
>> side.
>>
>> Just eliminating all the arguments and extended discussions has  
>> just got
>> to be a good plus point! :-)
>
> A further thought...
>
> The calibration can easily be used with the present "Cobblestones"
> scheme. Hence, the server-side development should be relatively easy
> (sorry, bad pun :-) ) and be able to be rapidly put in place, and not
> upset the present credits scheme.
>
> The old CPDN lab machine mentioned earlier that is still running could
> form a starting point as a calibration reference.
>
> Note: That will not solve the problem of trying to credit for what the
> cobblestones do not even measure. I guess that non-FLOPS projects will
> continue to have to add up a few fictitious pseudo-FLOPS!
>
>
> At a later date, once all the discussion for a new credits scheme has
> been discussed, the /same/ calibration scheme should work just the  
> same
> for whatever credits and whatever resources are measured, provided  
> that
> we reference against representative system hardware for calibrating
> whatever mix of resources are being credited.
>
> (Apologies if the long sentence breaks the Google translator :-( )
>
>
> In brief: The calibration scheme should work just the same for any
> credits scheme. You just need your "etalon"(*) computer to be suitably
> representative for all of the mix of resources being referenced.
>
>
> Regards,
> Martin
>
> * I'm guessing that Raistmer's Russian <--> English translation of
> "etalon" means "reference-comparison device", just as for an etalon in
> an interferometer.
>
>
> -- 
> --------------------
> 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.

_______________________________________________
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