On 4 April 2011 15:26, Andrew McNab <[log in to unmask]> wrote:
> On 04/04/2011 15:20, Stuart Purdie wrote:
>>
>> Sam's point about mapping (cpuid, Queue name) to HEPSPEC06 value would
>> remove any such systematic errors from being a problem, of course. In the
>> light of an unknown, surely that's the better way to proceed?
>
> Queue name to HEPSPEC06 was the suggestion in the first email! I don't see a
> problem with adding CPUID in too.
>
The difference is that Queue name -> HEPSPEC is broken (for the
reasons that many people have written too many replies to this thread
concerning ;) ), while in the (site_id, cpu_id) mode, we're actually
just using the site_id to differentiate the cpu_ids in different
clusters better (that is, this is (cpu_id, environment of cpu) really,
rather than (queue_name, specifier_of_additional_features_of_queue). )
That quibbling aside, I don't see how mapping the tuple would break
anything, and if you're happy with it, I'm happy with it....
Sam
> Cheers,
>
> Andrew
>
> --------------------------------------------------------------
> Dr Andrew McNab, High Energy Physics, University of Manchester
>
|