On 05/04/2011 12:51, Stephen Jones wrote:
> Ewan MacMahon wrote:
>
>> My understanding of the currently available information is that
>> the benchmark value published in GlueHostProcessorOtherDescription
>> should be the weighted average for the CPUs behind a CE, and should
>> not be scaled. If my understanding is correct then I think that's the
>> value that you're looking for.
>
> You're understanding is correct. The
> GlueHostProcessorOtherDescription:benchmark is the total hepspec06 of
> the cluster, divided by the number of logical cpus in that cluster. It
> therefore represents the average strength of one logical cpu.
Logical CPUs means hyperthreading changes it. If it's per-core,
then it should be sufficient.
Cheers,
Andrew
--------------------------------------------------------------
Dr Andrew McNab, High Energy Physics, University of Manchester
|