| |
|
|
|
|
Code Mechanics |
Code Mechanics -- Can They Help Your Code?
The NCSA Performance
Analysis team is here to help you!
Whether you are experienced with our systems or a new user,
it's possible that the Performance Engineering and Computational Methods
(PECM) team -- in conjunction with experts from
SGI/Cray and HP/Convex -- can study your code and offer you some
tips for tuning your code for a specific architecture at NCSA,
making it more efficient using the system and your allocation.
Alliance-affiliated high-performance computational researchers may
send email to Mark Straka
with the following information about
your code. Mark will contact you with any questions or with some tips.
Before you send mail, though, take a look at the
quick tips PECM has assembled to help you make
big gains in your code.
- Contact information
- name, email address
- Allocation information
- PSN if you are an Alliance-allocated user, site where you compute if not at
the Alliance, system on which you compute
- Code information
- language(s), size [lines, files, subroutines], system(s) the code runs
on, disk and memory requirements, any special libraries, compile instructions,
approximate CPU time needed, degree of
parallelism [not possible, not yet implemented, implemented but not optimized]
cache effects -
multiplatform comparisons -
top performers -
scalability - i
tips -
benchmarks -
code mechanic -
highlights home -
PECM
|
|
|
|
|
