Adaptyst sprint #2: summary
The second Adaptyst sprint at CERN took place between 27 May 2026 and 17 June 2026. Originally planned for 2 weeks, it was extended by 1 week due to unforeseen medical problems of the Adaptyst main R&D person as mentioned in the previous post. Additionally, these problems unfortunately made the overall productivity during the sprint lower than expected. Despite this, there was still progress made which is explained in the remainder of the article.
During the period above, the scope of some issues turned out to have to be changed for variety of reasons such as encompassing too many big things within a single task. Therefore, it should be noted that all work items discussed in this article are in their end-of-sprint versions.
The following public issues were scheduled to be performed (the order here may not reflect the actual one):
- Add Adaptyst Analyser API:
- Add code regionisation support to linuxperf
- Facilitate Adaptyst and modules setup with pre-compiled binaries
Furthermore, 3 CERN-internal issues were scheduled, not available to be viewed publicly. These will be discussed separately using private channels if needed.
Adapt the existing Python API (subtask of “Add Adaptyst Analyser API”)
This task was completed, resulting in the first preliminary version of the Adaptyst Analyser API allowing users to query performance analysis results programmatically and export links to their specific elements (e.g. a flame graph of a selected thread/process) in a given Adaptyst Analyser web server instance.
As the API has specifically been requested by the LHCb experiment at CERN, a meeting will be set up with them to demonstrate this functionality and determine the next steps.
Document the existing REST API and document the Python API (subtasks of “Add Adaptyst Analyser API”)
Originally planned to be done during the second sprint, these two tasks were moved to the “Waiting for third party” state because a demo meeting has to be held with the LHCb experiment before any substantial documentation can be produced and published.
Add code regionisation support to linuxperf
Some concrete thought was given to how to address this closer to the end of the sprint. Hopefully, this will make good foundation for the proper implementation work of the code regionisation support in linuxperf, scheduled to be done in the already-ongoing third Adaptyst CERN sprint.
Facilitate Adaptyst and modules setup with pre-compiled binaries
When considering how to approach implementing this task, we realised that the patched “perf” dependency of the linuxperf module complicated significantly handling streamlined installation of Adaptyst-team-made modules. Therefore, we decided to tackle (long overdue) moving linuxperf from the patched-“perf” arrangement to an in-house equivalent first, e.g. via perf_event_open() syscall and/or eBPF. This linuxperf change is already part of the third sprint, but given the complexity of the task, it may still be unfinished by the end of the sprint.
Next sprint
Adaptyst sprint #3 already started yesterday on 18 June 2026: it can be followed in the issue tracker and in the forum.