	Contact Us \| Intranet \| Search

Performance FAQ

User Information Home

Compute Resources

Software

Math Libraries
Chemistry
Biology
Structural Mechanics
Fluid Dynamics
Performance
Visualization

Data

Advanced Support for TeraGrid Applications

NCSA's Help Desk is available 24 hours a day, seven days a week, 365 days a year:
help.ncsa.illinois.edu
217-244-0710
help@ncsa.illinois.edu

Frequently Asked Performance Related Questions

Job Timing

What can be done to get an accurate report of the time used for an MPI job?
How do I run and time some parallel C programs on the SGI machines?
To get really accurate timings, do I have to run my programs in a "dedicated" queue ?
How do I time a Fortran subroutine?
How do I run and time parallel Fortran programs on the SGI machines?
Measuring Speedup and Scaling
Is the best way to measure speedup to run the same problem on increasing number of processors or to increase the problem size in proportion to the number of processors?
Is a good way to measure how much a code improves when parallelized to compare CPU time to wall clock time?
Using Performance Tools and Interpreting Timing Reports
How do I measure the amount of time spent in a particular loop in a fortran program?
How should I use the calipers option to measure the number of cycles, cache misses, etc.?
How can I determine if cache misses are contributing to the performance of an application.?
When running the ideal experiment on a multi-processed job parallelized with pfa, I see routines listed from the runtime system ( I assume) like __mp_evc_get and __mp_wait_for_completion. What exactly do these routines do?
What tools can I use to profile my mpi code to see where the time goes?
I tried using prof, but I can not seem to get what I need.
Counting MFLOPS - Using the R10000 Hardware Counters
How do I count floating point operations?
How can I obtain some performance measurements (CPU times MFLOPS, etc) for some FORTRAN programs?
MPI
What can be done to get an accurate report of the time used for an MPI job?
Is there a way to get the equivelent timing info for an mpi job as the following command gives for a regular job?
How can I quantify the performance of my MPI code?
Is upshot the only parallel performance analyzer available for MPI?
Miscellaneous System Issues (Memory, I/O, etc.)
Can you tell me what types of operations take up system time ?
Is there an option that I can give to get the total amount of memory that my code used?
Is there an explanation for such big CPU differences between Cray C90 and SGI-PCA performance?
Is there a way to write things out to a file in the middle of a batch job's execution?

Tuning Highlights | PECM | SCD | FAQ