If you really want to open that can of worms, how about the fact that a 
floating point add (which is a floating point operation) is dramatically 
easier than a floating point cos().

... and the fact that an AMD processor might do adds dramatically faster 
than an Intel, but do cos() slower.

A hypothetical "add...@home" could give dramatically higher credit just 
because it does particularly fast floating point operations, especially 
compared to "trigonome...@home."

I don't see a way out, short of doing exactly what Eric Korpela's script 
tries to do -- normalize FLOPS credit to that predicted by the 
(imperfect) benchmarks.

Nicolás Alvarez wrote:
> El Martes 21 Jul 2009 16:54:01 Martin escribió:
>> My thought is that we must have a semantic shift so that what is
>> usefully utilised is rewarded, and not just *time spent* (perhaps busyly
>> uselessly spinning wheels) on whatever hardware.
> 
> The GPU and CPU apps don't necessarily make the same amount of floating point 
> operations. If someone optimizes one of the two apps so that it can do the 
> same with (slightly?) less calculations, and you grant credits per flops, 
> then GPU and CPU get different credits for doing the exact same task (meaning 
> same input, same output). If that happens, credits aren't really 
> reflecting "work done", in my opinion.
> 
> By the way, credits are already defined proportional to flops: "1/100th day 
> of 
> CPU time on a computer that does both 1000 double-precision MIPS and 1000 
> integer MIPS." In other words, "a 1 GigaFLOP machine, running full time, 
> produces 100 units of credit in 1 day."
> _______________________________________________
> 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