Re: Why is FLUKA slower than MCNPX?

Someone mentioned:
|--------------------------------------------------------------------------|
|"I have recorded the electron current on the outer surface of the |
|shielding. For the tally of the first channel (0 - 11.5 keV), FLUKA's |
|result is much higher than MCNPX's. The FLUKA spectrum show a obvious peak|
|at the first channel. [..]" |
|--------------------------------------------------------------------------|

As the results are not equal, this is not a comparable example of
FLUKA being slower than MCNPX. However, I shall respond as if it
were...

|--------------------------------------------------------------------------|
|"I am now making some intercomparisions of FLUKA and MCNPX. [..] |
|[..] |
| |
|It turned out that FLUKA runs much slower than MCNPX. [..] |
|[..] |
| |
| |
|[..]" |
|--------------------------------------------------------------------------|

FLUKA is not compiled by an optimizing compiler. Furthermore, compiler
developers probably do not have a copy of FLUKA source so they are not
as likely to provide compilers which optimize for FLUKA. One compiler
company has listed MCNP as one of the codes which can be compiled:
WWW.PGroup.com/resources/tips.htm
Another compiler company had a similar list (but lacking MCNP) at
WWW.PathScale.com/documentation

Compiler companies tend to concentrate on producing compilers which
produce efficient code for SPEC benckmarks. Many of the programs
listed on the two webpages mentioned above are in a SPEC benchmark:
WWW.SPEC.org/cpu2006/results/res2011q3/cpu2006-20110621-17253.html

The executables distributed for MCNP are actually compiled by a
different compiler.

Yours faithfully,
Nicholas Collin Paul de Glouceſter
Received on Fri Jul 29 2011 - 16:49:34 CEST