1. The code now works without CUDA_LAUNCH_BLOCKING set by using explicit
synchronizations where required.
2. The code has also been modified to use thread specific memory spaces,
which for Kokkos' Cuda backend means per thread streams.
This limitation is caused by the fact that VTK-m caches compiler
location during the build stage for these tests, and this fails
on the test image as the compiler might not exist or is in a
different path
2d1b609b3 Use Ubuntu instead of rhel8 for cuda+kokkos
769248583 Make sure we use c++14 when using CUDA 11+
64efa6401 Kokkos: make sure we don't pass multiple rdc flags
b2f4c8e5e Switch -O3 to -O2 on Linux with Cuda 10
db57ed26a Fix warnings
452f61e29 Add Kokkos backend
Acked-by: Kitware Robot <kwrobot@kitware.com>
Merge-request: !2164
The `ContourTreeUniformAugmented` tests for Cuda consistently
times out while running on the `centos7_gcc48` and `rhel8_test_centos7`
configurations. Temporarily exclude it to improve the CI turn-around time
while this issue is being resolved.
934f085e0 Build diy as a library
f0a37ac6a Merge branch 'upstream-diy' into diy-mpi-nompi
7687aabf8 diy 2020-06-05 (b62915aa)
6ca2b9f87 Point to new version of Diy
Acked-by: Kitware Robot <kwrobot@kitware.com>
Acked-by: Robert Maynard <robert.maynard@kitware.com>
Merge-request: !2123
The centos7 on rhel8 was not running the install tests
as the path to cmake/ctest wasn't valid. This causes
the causes the following issues:
1. NOT_RUN tests with REPEAT don't show up in CMake failed
output fixed by:
https://gitlab.kitware.com/cmake/cmake/-/merge_requests/4801
2. failed tests cause ctest_test to return an error code,
which propagates to the script, which propagates to
having ctest running the script return an error.
When VTK compiles VTK-m for inclusion as an accelerator library, it sets
an option to compile filters and other components for a different (and
expanded) set of common types. We should test this configuration.
A single gitlab-runner can run multiple test stages
concurrently on the same hardware. When we have 3
jobs asking for 100% of the cpu's we get a 300%
reduction in performance caused by task switching
Problem was that CMake 3.12 was the CMake version used in the
build image, and test image was using 3.13. This was a problem
as the install test invocation aren't backwards convertible and
therefore failed.
To simplify reproducing docker based CI workers locally, VTK-m has python program that handles all the
work automatically for you.
The program is located in `[Utilities/CI/reproduce_ci_env.py ]` and requires python3 and pyyaml.
To use the program is really easy! The following two commands will create the `build:rhel8` gitlab-ci
worker as a docker image and setup a container just as how gitlab-ci would be before the actual
compilation of VTK-m. Instead of doing the compilation, instead you will be given an interactive shell.
```
./reproduce_ci_env.py create rhel8
./reproduce_ci_env.py run rhel8
```
To compile VTK-m from the the interactive shell you would do the following:
```
> src]# cd build/
> build]# cmake --build .
```