homme SYCL changes #6594
homme SYCL changes #6594
Conversation
|
components/homme/src/theta-l_kokkos/cxx/HyperviscosityFunctorImpl.cpp
Outdated
Show resolved
Hide resolved
|
As I understand it, the int/bool code is just to work around a compiler bug affecting the F90-C++ interface; essentially, the compiler's treatment of c_bool is broken, or something like that. In particular, using bool in a kernel is not an issue; we do that in several spots. Oksana, is all of that correct? If so, I can add a commit to this branch next week that isolates the workaround to just the F90-C++ interface code. In particular, I'd like to remove the two theta_hydrostatic_mode variables from control_mod.F90. Is that OK? |
Yes, correct about int-bool issue.
yes, thanks! |
|
@bartgol thank for reducing the number of commits but you could do this one more time using --author on the commit command to give author credit to Oksana? |
|
Of course. I'll do it as soon as I get back to my computer |
bartgol
force-pushed
the
bartgol/homme/sycl
branch
from
df31c59 to
6c33665
Compare
* Disable timing in prim_main for the first two teps * Add some timers in EulerStepFunctor
bartgol
force-pushed
the
bartgol/homme/sycl
branch
from
6c33665 to
132b3a5
Compare
|
@rljacob @oksanaguba I edited the commits, to fix the author. @ambrad FYI, in case you pulled and was planning to do the mods you mentioned... |
|
I think I'll have a commit to push to this branch by tonight. I've made the changes and am running homme_integration on Chrysalis. |
Also fix preqx's use of use_moisture. Remove MOIST-DRY enum to avoid confusion, since it's no longer used.
|
Done. I also fixed some moisture-related code in preqx so that homme_integration passes on Chrysalis. |
oksanaguba
force-pushed
the
bartgol/homme/sycl
branch
from
79ebefa to
f30afb2
Compare
|
homme suite pass on chrysalis |
Changes needed for spot/aurora: sycl kokkos spaces, temporary TP changes (TP init will be sorted out later), switch booleans to ints in F<-->xx interfaces. see #6569 for the original branch/history. [bfb]
|
in next |
|
hommexx ers_d test failed on cdash with an mpi error, but i ran it by hand with no issues. |
These are due to PR E3SM-Project#6594 and break the E3SM-repo EAMxx build. Once the SCREAM and E3SM repos are unified, we can back out these workarounds. The workarounds are isolated to components/homme, keeping the commit separation of components/eamxx (SCREAM repo) and components/homme (E3SM repo) clean.
…6645) Hommexx: Temporarily work around some EAMxx-Hommexx incompatibilities. These are due to PR #6594 and break the E3SM-repo EAMxx build. Once the SCREAM and E3SM repos are unified, we can back out these workarounds. The workarounds are isolated to components/homme, keeping the commit separation of components/eamxx (SCREAM repo) and components/homme (E3SM repo) clean. Fixes #6635. [BFB]
Hommexx: Temporarily work around some EAMxx-Hommexx incompatibilities. These are due to PR #6594 and break the E3SM-repo EAMxx build. Once the SCREAM and E3SM repos are unified, we can back out these workarounds. The workarounds are isolated to components/homme, keeping the commit separation of components/eamxx (SCREAM repo) and components/homme (E3SM repo) clean. Fixes #6635. [BFB]
These are due to PR E3SM-Project#6594 and break the E3SM-repo EAMxx build. Once the SCREAM and E3SM repos are unified, we can back out these workarounds. The workarounds are isolated to components/homme, keeping the commit separation of components/eamxx (SCREAM repo) and components/homme (E3SM repo) clean.
Changes needed for spot/aurora: sycl kokkos spaces, temporary TP changes (TP init will be sorted out later), switch booleans to ints in F<-->xx interfaces.
see #6569 for the original branch/history.
[bfb]