dsl/compiler: Extend solver_parameters handling in PETScSolve and update logging #2718
Conversation
…nctionality examples: Add example using argparse in conjunction with PetscInitialize
…to set petsc options
ZoeLeibowitz
changed the title
| 'ksp_type': 'gmres', | ||
| 'pc_type': 'none', | ||
| 'ksp_rtol': 1.e-5, | ||
| 'ksp_atol': 1.e-50, |
There was a problem hiding this comment.
That's an interesting "numerical" number
There was a problem hiding this comment.
It is the default one used in PETSc
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## petsc #2718 +/- ##
==========================================
- Coverage 84.89% 84.46% -0.44%
==========================================
Files 268 269 +1
Lines 52464 52832 +368
Branches 4457 4472 +15
==========================================
+ Hits 44540 44622 +82
- Misses 7155 7440 +285
- Partials 769 770 +1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
devito/petsc/iet/logging.py
Outdated
| Example: | ||
| -------- | ||
| >>> self.function_list |
There was a problem hiding this comment.
Should self.function_list be allowed to be mutable? Ditto with self.petsc_option_mapper
JDBetteridge
left a comment
JDBetteridge
left a comment
There was a problem hiding this comment.
Not all comments need addressing before merge
|
|
||
|
|
||
| petsc_variables = get_petsc_variables() | ||
| # TODO: Use petsctools get_petscvariables() instead? |
| def get_petsc_type_mappings(): | ||
| try: | ||
| petsc_precision = petsc_variables['PETSC_PRECISION'] | ||
| except KeyError: |
There was a problem hiding this comment.
Shouldn't this be a hard error if the precision cannot be determined?
There was a problem hiding this comment.
This is called inside devito/passes/iet/languages/C.py so it;s not a hard error otherwise CI will break. However it's also called inside PETScSolve, so maybe this should be extended
| if petsc_precision == 'single': | ||
| mapper[ctypes.c_float] = petsc_scalar |
There was a problem hiding this comment.
I think a lot of this shouldn't be indented, but I can't tell because github has messed up the formatting.
…et_info, clean up and more tests
ZoeLeibowitz
force-pushed
the
solver_params
branch
from
2d1fd0c to
09a1d20
Compare
solver_parametersdictionary inPETScSolve.solver_parametersdict.get_infoargument toPETScSolveto collect specific statistics from the solver.