Hi Song shuhua,
The ybatchCalc workspace method is the best way to make a calculation for
multiple profiles. ARTS will then parallelize over the batch cases which is
most efficient. This of course assumes that you have more batch cases than you
have cores, otherwise some of the cores cannot be utilized. Here is a
controlfile example for ARTS 2.4 which demonstrates the use of ybatchCalc:
https://github.com/atmtools/arts/blob/v2.4.x/controlfiles/artscomponents/clearsky/TestBatch.arts
Hope this helps.
Cheers,
Oliver
> On 4 Nov 2022, at 10:16, 宋淑华 wrote:
>
> Dear all, Hello, I am running a batch scattering radiation simulation program
> on a multi-core processor. It involves the simulation of multiple profiles on
> multiple frequency grids. And the scattering data I use is the ARTS single
> scattering database. However, it is found that the role of multi-core
> processors cannot be fully realized. Even though I choose multiple cores by
> running the command "-n", the actual operation only use a few cores. I'm
> sure that the ARTS I compiled supported OpenMP. I guess maybe my program is
> not efficient enough. I would like to ask how to make calculations more
> efficient through reasonable settings when writing programs. For example,
> loop multiple profiles for a single frequency point, or calculate multiple
> frequency grids for a single profile, or other more reasonable ways. Looking
> forward to your replies and suggestions.
>
> Best Regards, Song shuhua
smime.p7s
Description: S/MIME cryptographic signature
