-
Notifications
You must be signed in to change notification settings - Fork 2
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
#163: Belos: Provide Tpetra
version of LSQR examples
#168
Conversation
Tpetra
version of LSQR examples
* #128: set up GPU CI pipelines * #128: specify cuda dir * #128: temporarily run GPU on every push; stop MPI builds * #128: specify correct dockerfile * #128: provide correct build/source dirs * #128: rework MPI in build script * #128: rework MPI in other gpu build script * #128: CI build script try another configuration and fix invalid path * #128: fix missing letter in path for a gpu build * #128: add newlines to end of files * #128: add spack find -p to find cuda root * #128: only run one pipeline; add cuda paths * #128: add kokkos variables * #128: add Tpetra_INST_SERIAL:BOOL=ON * #128: add CUDA root flag * #128: use correct kokkos architecture * #128: enable cusolver and cusparse * #128: emulate local build * #128: try with different docker image * #128: update cuda path * Try adding debug flag for Buildx * Tpetra: Disable cudaMemcpyAsync for Intercept.cpp * #128: lower -j * #128: use defaul kokkos architecture * #128: use fewer processes for GPU testing * 128: re-enable Kokkos_ARCH_AMPERE86 * #128: add cuda sample build to CI to validate CUDA * #128: run cuda test on NGA host * #128: update jobs dependency on CI for cuda * #128: add CUD sample run * #128: remove command not existing in CI * #128: change cuda path * #128: try to display information about driver * #128: change bad command in CI * #128: fix command * #128: try install nvidia util in Docker container * #128: remove commands * #128: fix cuda path * #128: fix dockerfile * #128: add different cuda test images * Run container in separate step * Remove not needed code * Apply changes to Epetra=OFF * Try to build and run docker within same step * #128: remove unused old CI files * #128: check both gpu pipelines * #128: Tpetra_INST_SERIAL=ON * #128: fix workflow name * #128: rework with cuda 11.4 dockerfile * #168: try to simplify CI sheel script * #128: try simplify shell script * #128: remove librairies path for blas and lapack to check if resolved * #128: try remove Lapack and blas lib paths from cmake call * #128: try again changing path dynamically * #128: fix another path * #128: fix blas path * #128: apply working conffiguration to other build scripts * #128: restore triggering workflows on PR * #128: disable GPU build job for PR having `EpetraMPI T1` label * #128: enable GPU build only with EpetraMPI T2 and EpetraMPI T3 labels * #128: upload test log * #128: fix typo * #128: fix artifacts * #128: add junit report for tests * #128: add junit reporting in CI and set * #128: fix artifact name * #128: fix artifacts missing * #128 fix extra slach char in path * #128: fix artifacts path * #128: fix path in gitbub action * #128: try mounting artifacts folder into the host runner * #128: use same logic for gpu or non-gpu pipelines * 128: Finalize pipelines (GPU on push, MPI cancellations) * 128: remove label requirements * Revert "Tpetra: Disable cudaMemcpyAsync for Intercept.cpp" This reverts commit 5db2d5d. * #128: test intercept reversion * Revert "Revert "Tpetra: Disable cudaMemcpyAsync for Intercept.cpp"" This reverts commit de87a22. * #128: fix underscore * #128: run GPU pipeline on merge to fy23 develop --------- Co-authored-by: Thomas Dutheillet-Lamonthézie <[email protected]> Co-authored-by: Jacob Domagala <[email protected]>
* #128: specify cuda dir * #128: temporarily run GPU on every push; stop MPI builds * #128: specify correct dockerfile * #128: provide correct build/source dirs * #128: rework MPI in build script * #128: rework MPI in other gpu build script * #128: CI build script try another configuration and fix invalid path * #128: fix missing letter in path for a gpu build * #128: add newlines to end of files * #128: add spack find -p to find cuda root * #128: only run one pipeline; add cuda paths * #128: add kokkos variables * #128: add Tpetra_INST_SERIAL:BOOL=ON * #128: add CUDA root flag * #128: use correct kokkos architecture * #128: enable cusolver and cusparse * #128: emulate local build * #128: try with different docker image * #128: update cuda path * Try adding debug flag for Buildx * Tpetra: Disable cudaMemcpyAsync for Intercept.cpp * #128: lower -j * #128: use defaul kokkos architecture * #128: use fewer processes for GPU testing * 128: re-enable Kokkos_ARCH_AMPERE86 * #128: add cuda sample build to CI to validate CUDA * #128: run cuda test on NGA host * #128: update jobs dependency on CI for cuda * #128: add CUD sample run * #128: remove command not existing in CI * #128: change cuda path * #128: try to display information about driver * #128: change bad command in CI * #128: fix command * #128: try install nvidia util in Docker container * #128: remove commands * #128: fix cuda path * #128: fix dockerfile * #128: add different cuda test images * Run container in separate step * Remove not needed code * Apply changes to Epetra=OFF * Try to build and run docker within same step * #128: remove unused old CI files * #128: check both gpu pipelines * #128: Tpetra_INST_SERIAL=ON * #128: fix workflow name * #128: rework with cuda 11.4 dockerfile * #168: try to simplify CI sheel script * #128: try simplify shell script * #128: remove librairies path for blas and lapack to check if resolved * #128: try remove Lapack and blas lib paths from cmake call * #128: try again changing path dynamically * #128: fix another path * #128: fix blas path * #128: apply working conffiguration to other build scripts * #128: restore triggering workflows on PR * #128: disable GPU build job for PR having `EpetraMPI T1` label * #128: enable GPU build only with EpetraMPI T2 and EpetraMPI T3 labels * #128: upload test log * #128: fix typo * #128: fix artifacts * #128: add junit report for tests * #128: add junit reporting in CI and set * #128: fix artifact name * #128: fix artifacts missing * #128 fix extra slach char in path * #128: fix artifacts path * #128: fix path in gitbub action * #128: try mounting artifacts folder into the host runner * #128: use same logic for gpu or non-gpu pipelines * 128: Finalize pipelines (GPU on push, MPI cancellations) * 128: remove label requirements * Revert "Tpetra: Disable cudaMemcpyAsync for Intercept.cpp" This reverts commit 5db2d5d. * #128: test intercept reversion * Revert "Revert "Tpetra: Disable cudaMemcpyAsync for Intercept.cpp"" This reverts commit de87a22. * #128: fix underscore * #128: run GPU pipeline on merge to fy23 develop --------- Co-authored-by: Thomas Dutheillet-Lamonthézie <[email protected]> Co-authored-by: Jacob Domagala <[email protected]>
Hello @stmcgovern The new I think there are 2 possibilities if we want to have a new Tpetra version of this example:
I think the option A is easier to apply but the option B might be more logic even if we have to modify 2 packages at once. As I am writing this comment I am working on making the example working by ignoring the dependency problem but we will have to know how to make it work at build stage to not have dependency issue. |
Currently the example does not complete before test timeout of 1500 seconds in the CI if we include the example as a test. |
9605912
to
a1b19bc
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good to me, after rebase should be good to go
56b7a2c
to
504841c
Compare
504841c
to
0b158b3
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The Preconditioned version "PrecLQSRTpetraExFile.cpp" is provided but is not built (commented out in CMakeLists.txt) due to the circular dependency issue under consideration by the package developers.
Fixes #163
PR to Trilinos: trilinos#12300