C and Fortran API for Interpolator #178
Conversation
the test is still not passing. pushed it to save it at current state
the values are not being compared yet
passing with 10% error margin. [future] find best degree and min required support to minimize the epsilon of the test
with 10% error margin and [temporary] it doesn't work when a source and a target becomes coincident: discussed here: SCOREC#148 (comment)
compiling, yet to be properly merged to the other apis and testing
- compiling - yet to be completed and tested
keeps the mesh and lib alive
- test is going through and finding the supports but goes infinitely - needs number of supports and max_radius etc. as parameters - still needs more structs, enums to represent interpolation types and other functionalities - still needs evaluate function
now the test is passing and finding supports - the mesh is used temporarily
test is passing for an uniform field. It just checks if the functionalities are at the place
- tested but failing for some specific target points
when xgc's corresponding omega_h mesh is not available,it needs mixed interpolation. all these methods are based on point cloud they can be improved search can be unstable (find better ways) and changed the radius adaptation so that it stops after 100 iterations even if it hasn't found enough supports
still needs to solve the problem with number of neighbors in mesh based search by merging the both radius adaptation
remains todos
separate helper function header passes interpolation class test fix issue SCOREC#207
except for prints inside gpu function since fmt is not supported for GPU
jacobmerson
force-pushed
the
mergecfortapi
branch
from
80db493 to
b59c425
Compare
|
Looks like there's an issue reading the mesh in the test. Looking at the The reading issue is described in SCOREC/omega_h#183 and was fixed in SCOREC/omega_h#184. It worked before because you cleared the cache before, and Omega_h was using the latest commit with the fix. For instance, the last passing test job above was using SCOREC/omega_h@d1f09bb, which includes the abovementioned pull request. My guess is that #231 specifies Omega_h version to be scorec-v11.0.0. It does not include the mesh reader fix. |
|
Thanks. The test is passing. The installation is failing here. But locally, for me, it is working. I am trying to sort it out. |
|
Looks like that new test case for installations is paying off! :) My guess is it has something to do with the fact that you have |
The goal of this PR is to separate the mls_interpolation.hpp and mls_interpolation_impl.hpp so that the implementation is not included in the header file.
this commit modifies the targets for the capi so that pcms::capi is an interface that includes the subcomponents. This matches more closely to the way that the main pcms::pcms library has been put together.
jacobmerson
force-pushed
the
mergecfortapi
branch
from
5c6fbe4 to
86ce66f
Compare
this commit modifies the targets for the fortranapi so that pcms::capi is an interface that includes the subcomponents. This matches more closely to the way that the main pcms::pcms library has been put together.
jacobmerson
force-pushed
the
mergecfortapi
branch
from
ac79a06 to
76b88bd
Compare
|
Maybe we can try to put |
That's a solid theory. Let me give it a try. @Sichao25 it worked! thanks |
4 participants
test_normalization_routine#177test_normalization_routine