Speeding up PROCESS

[jonmaddock]

In GitLab by @mkovari on Jan 14, 2021, 13:37

A profiling tool was run, generating the diagram below that shows the number of times each procedure is run, and the fraction of the total time used by that procedure. Some procedures are run tens of millions of times, for reasons that are not clear.

For example, subroutine impradprofile is run 4,472,799 times, and uses 65.55% of the total execution time. Almost all of this is due to the two functions:

function pimpden(imp_element, ne, te)
    !! Total impurity radiation density (W/m3)

and

function pbremden(imp_element, ne, te)
    !! Bremsstrahlung radiation density (W/m3)

Explanation:

This shows that 63% of the total time is spent in these two routines alone (excluding the calls within them).

It would be very interesting for a keen individual to look into this! @ajpearcey @asbrown @kzarebsk @cjwgriesel @stuartmuldrew @jonmaddock @jmorris-uk ??

[jonmaddock]

In GitLab by @kzarebsk on Mar 12, 2021, 13:13

I am performing an investigation into moving from linear to binary search in the interpolation.
Currently in the PROCESS code a match between input temperature and the temperature array is found by linearly going through all the elements. I tried out the pimpden function using a binary search and investigated the results.

[jonmaddock]

In GitLab by @mkovari on Mar 18, 2021, 13:12

Please don't forget issue #1230.

[jonmaddock]

In GitLab by @wg7580 on Jul 20, 2022, 08:56

closed