Conversation
|
/run-hpsf-gitlab-ci |
|
Took a manual review and it missed a lot of lifetime dependencies. Oh well. Let's see if the examples added now in comments help it to convert. I'll feed it via |
|
GitLab CI 1436197 finished with status: failed. See details at https://gitlab.spack.io/amrex/amrex/-/pipelines/1436197. |
|
/run-hpsf-gitlab-ci |
|
@WeiqunZhang when the async arena runs out of memory and starts to drop oldest handled memory, does it first run a GPU stream sync to ensure correctness? |
|
GitLab CI 1436835 finished with status: success. See details at https://gitlab.spack.io/amrex/amrex/-/pipelines/1436835. |
|
Yes the async arena does a stream sync if there are too many freed allocations in it. |
|
/run-hpsf-gitlab-ci |
|
GitLab CI 1436879 finished with status: success. See details at https://gitlab.spack.io/amrex/amrex/-/pipelines/1436879. |
|
GitHub runners have some unrelated issues rn in fetching apt dependencies |
|
besides flaky (CPU) CI runners, all tests pass now. I finished my self-review. I am pretty sure our GPU test coverage is not very strong, so a good review is welcome 🙏 |
3 participants
Summary
Follow-up to #4955 as discussed
cc @bathmatt @roelof-groenewald
Additional background
Checklist
The proposed changes: