Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]] On Behalf Of Andrew McNab said:
> "The second value of this attribute MUST be published only in
> the case
> the CPU power of the SubCluster is computed using the Benchmark
> HEP-SPEC06."
That just means that you shouldn't estimate it, it should only be published if you have run the benchmark program (which I assume everyone in GridPP has by now).
> ie cores vs logical CPUs? ie including hyperthreads or not?
At the time the benchmark was defined hyperthreads seemed to have gone away as an issue, so it wasn't explicitly considered, but I don't think it really changes anything, hyperthreaded CPUs and cores are effectively the same thing for these purposes. (In fact the original reason the LogicalCPUs attribute was defined was to deal with hyperthreading, since multicore CPUs weren't around then. In computing it seems that anything comes around again if you wait long enough :)
However it may be that it would be worth the GDB defining this more explicitly given that it's becoming more important.
Stephen
|