Looking at the CPM code, I see that the process times are aggregated on a per component basis. If you have say 10 threads, meaning you scale out a particular component, process times for all of them are aggregated (added together). So it seems that what you are seeing is working as designed. You may need to divide the process time reported by the CPE by the number of threads to get component process time.
Jerry C On Tue, Jan 18, 2011 at 11:02 AM, Eric Riebling <[email protected]> wrote: > PERFORMANCE REPORT BUG > > There appears to be a bug when generating the information after > a CPE completes, that depicts the total processing time. If there > are more than 1 thread in the CPE, the processing times per thread > are added together, so a 10 threaded setup will report a total time > that is actually 10 times longer than elapsed. > > IMPORT BY NAME ISSUES WITH COMPONENT DESCRIPTOR EDITOR > > A suggestion to improve by example, relating to "import by name" > - in order for "import by name" to work properly in the Component > Descriptor Editor, imported projects > have to list their descriptors as "Source folders on build path" > such that they get copied (automatically) to the Default output > folder (usually project-name/bin). This is not the case in the > uimaj-examples project, so anyone using that as a template is > possibly going to be confused in two ways: > > * 'import by name' not working > * cognitive dissonance: XML descriptors aren't binaries, yet > they get copied there as such whenprojects are set up > such that they are listed as Source on build path > (in order to make 'import by name' work properly) > -- > Eric Riebling GHC 6713, LTI, SCS, CMU > 412.268.9872 http://www.cs.cmu.edu/~er1k >
