Expand mpi_grp_t usage in io to fix issues in develop

Description

Fix issues in develop introduced recently by two different MRs by removing explicit use of communicators and replacing that by mpi_grp_t and the corresponding MPI wrappers.

changed milestone to %12.0

added Bugfix label

@nicolastd Can you have a look at this? I adapted code that you recently introduced for outputting supercells.

assigned to @nicolastd

@sohlmann The compilation is failing, so I will wait for the fix before reviewing this.

I noticed as well ;) I'll fix it.

added 1 commit

• 8bdbc2b6 - Remove more HAVE_MPI macros and use mpi_grp_t

Compare with previous version

Codecov Report

Merging #1639 (e373d5df) into develop (a83a96f2) will increase coverage by 2.86%. The diff coverage is 47.00%.

@@             Coverage Diff             @@
##           develop    #1639      +/-   ##
===========================================
+ Coverage    63.86%   66.73%   +2.86%     
===========================================
  Files          535      538       +3     
  Lines        94353    96348    +1995     
===========================================
+ Hits         60257    64296    +4039     
+ Misses       34096    32052    -2044     

Impacted Files	Coverage Δ
src/basic/mpi_inc.F90	49.84% <34.44%> (+26.03%)
src/grid/io_function_inc.F90	48.47% <84.61%> (+1.72%)
src/basic/mpi.F90	83.50% <92.85%> (+23.88%)
src/electrons/subspace.F90	54.54% <0.00%> (-14.69%)
src/grid/mesh_init.F90	83.84% <0.00%> (-7.10%)
src/math/sparskit.F90	36.26% <0.00%> (-4.48%)
src/species/upf1.hpp	79.49% <0.00%> (-2.76%)
src/electrons/stress.F90	81.59% <0.00%> (-2.22%)
src/hamiltonian/projector.F90	88.42% <0.00%> (-1.91%)
src/electrons/x_fbe_inc.F90	60.00% <0.00%> (-1.74%)
... and 173 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update a83a96f...e373d5d. Read the comment docs.

resolved all threads

@sohlmann I seems that you broke some output in parallel....

added 1 commit

• d05c40f4 - Remove HAVE_MPI macros and use new MPI wrappers in io

Compare with previous version

I have reimplemented the changes without modifying the logic, just replacing the use of comm by mpi_grp_t objects.

@sohlmann Thanks for the changes. It seems that the PowerPC at least still does not appreciate your changes...

In fact other bots are also failing too.

added 1 commit

• bf85046b - Remove HAVE_MPI macros and use new MPI wrappers in io

Compare with previous version

Yes, I saw this as well - the broadcast is now only done when using more than one process in grp_.

added 1 commit

• b0af8581 - Remove HAVE_MPI macros and use new MPI wrappers in io

Compare with previous version

added 2 commits

• b6b81893 - Add checks for the communicator
• e373d5df - Remove checks on group size, they are now done in the wrappers

Compare with previous version