Skip to content

Conversation

@jjokella
Copy link
Collaborator

No description provided.

lnerger-pdaf and others added 30 commits May 8, 2024 11:41
…This circumvents a bug in the LAPACK implementation of OpenBLAS <0.3.27)
…oes now suppose MPI-paralleliza

tion. There is also one variant in which the MPI-specification is given as arguments in the call. Th
e old routine is still available as PDAF_diag_CRPS_nompi. The new routines use SANGOMA_quicksort for more efficient sorting.
… dim_obs_p=0 when OMI is used. Now the number of calls is consistent with domains having dim_obs_p>0. This allows for global MPI operations in the observation operator
…clude file (necessary as on the Cray always 'ftn' is used, but might refer to ifort, cray-ftn, or gfortran)
…ices. In this case, the routine profRinvA_l is specified by the user. Thus, the multiplicaation with any form of R and be implemeented.
For inputs longer than 100 characters, only the first 100 characters
are used as input, which may lead to unwanted behavior (f.e. when the
input string is a path).

This check aborts the execution, whenever the input string was longer
than 100 characters.
For paths, the `READ` function supplied with the default format
specifier `*` reads the input only up to the first appearance of the
character `/`.
Error  message:
```
221 | WRITE (*,'(2x, a, x, a)') "called handle=", TRIM(string)
| 1
Error: GNU Extension: X descriptor requires leading space count at (1)
```
…ore parser_mpi because the new check for the string lengths uses abort_parallel
…I observation localization weights into a new subroutine
lnerger-pdaf and others added 27 commits August 28, 2024 13:38
…f OMI-internal variables in user-supplied variants of PDAFomi_init_dim_obs_l
…s_l.F90. This replaces the previsou PDAFomi_init_dim_obs_l, which is still present as PDAFomi_init_dim_obs_l_old. The optimization yields better performance - for isotropic between 6-10x.
…. Now all allocations can be done in OMI-provided routines. Also storing the local index, distance and radii can be done in an OMI-provided routine. However, using this routine slowsdown the executation, in particular on the NEC.
…ep an example of using g2l_state_pdaf and l2g_state_pdaf
needs to change for Intel/Gnu compilers
@jjokella jjokella force-pushed the tsmp-pdaf-patched-v2.3 branch from 2471de7 to b81a289 Compare December 19, 2025 13:35
@jjokella jjokella merged commit a9fdb46 into tsmp-pdaf-patched Dec 19, 2025
1 check passed
@jjokella jjokella deleted the tsmp-pdaf-patched-v2.3 branch December 19, 2025 13:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants