Re: Multi-Volume Archive + Live Data Playback for PCP Client Tools

[Ken McDonell <kenj@xxxxxxxxxxxxxxxx>]

G'day Frank.

On 04/11/14 08:58, Frank Ch. Eigler wrote:
> ...
> I see what you mena, but unfortunately, the PAPI event code space (at
> least 32 bits) is much larger than the PMID low-order bits space, is
> not densely encoded, and bound to get worse when we address counters
> with extra option flags, "uncore" counters, and per-process/-cgroup.
I was not advocating a fixed mapping ... while there may be more than 
2^22 possible event counters, is it really the case that on one machine 
we could expect that more than 4 million different event codes to be 
enabled and exported by the PMDA?
The pmdaCache* routines would allow a dense allocation of more than 4 
million unique ids and maintain a persistent mapping from the PMNS name 
that the PMDA chooses and those unique ids.
Perhaps someone could send me the output from:

        $ pminfo -dm papi

from a machine where the papi PMDA is running and collecting real data.

Then restart the PMDA, start a different collection profile and repeat 
the command above.
This might give me some concrete context to understand the issues if 
I've got it wrong.