PROFILING OPTIONS

PROFILE -- Enable Profiling (default time profiling)

PROFILECOUNTERS -- Enable SGI Hardware Counters Profiling (instead of time)

PROFILECALLS -- Profile each invocation of every profiled function

PROFILESTATS -- Calculate standard deviation of exclusive time/counter

Use either PROFILE or PROFILECOUNTERS. PROFILECALLS and/or PROFILESTATS can be added to either. For e.g.,
% conf O2K64 MPI 72BETA PROFILE
or
% conf O2K64 MPI KAI_NEW PROFILECOUNTERS
or
% conf O2K64 MPI KAI_NEW PROFILE PROFILECALLS
or
% conf O2K64 MPI 72BETA PROFILECOUNTERS PROFILECALLS PROFILESTATS

SGI R10000 provides hardware registers for counters of low-level performance data. It answers questions like :"How many instructions were executed by each templated function? How many Cycles? Floating Point instructions? Loads? Stores? Cache misses?" So profiling can count wallclock microseconds using fast SGI hardware timer or any of 32 other quantities.

Running the Profiled Application

While running the application the user specifies in environment variable T5_EVENT0 the quantity to be measured :
% setenv T5_EVENT0 21 % mpirun -np 4 app --commlib mpi