Color code: (critical, news during the meeting: green, news from this week: blue, news from last week: purple, no news: black)
High priority Framework issues:
- Start / Stop / Start: 2 problems on O2 side left:
-
- All processes are crashing randomly (usually ~2 out of >10k) when restarting. Stack trace hints to FMQ. https://its.cern.ch/jira/browse/O2-4639
- TPC ITS matching QC crashing accessing CCDB objects. Not clear if same problem as above, or a problem in the task itself:
- Stabilize calibration / fix EoS: New scheme: https://its.cern.ch/jira/browse/O2-4308: Reported 2 issues to GIulio, waiting for a fix. Status?
- Fix problem with ccdb-populator: no idea yet - since Ole left, someone else will have to take care.
- TF-status message (from https://github.com/AliceO2Group/AliceO2/pull/13495) sent by readout-proxy. Status?
Sync reconstruction
- Waiting for RC to test COSMIC replay data set.
- Waiting for RC to test STOP timeout impact.
Async reconstruction
- Remaining oscilation problem: GPUs get sometimes stalled for a long time up to 2 minutes. Checking 2 things:
- does the situation get better without GPU monitoring? --> Inconclusive
- We can use increased GPU processes priority as a mitigation, but doesn't fully fix the issue.
- ḾI100 GPU stuck problem will only be addressed after AMD has fixed the operation with the latest official ROCm stack.
- Limiting factor for pp workflow is now the TPC time series, which is to slow and creates backpressure (costs ~20% performance on EPNs). Enabled multi-threading as recommended by Matthias - need to check if it works.
EPN major topics:
- ALMA9 / ROCm 6.x deployment on hold, until AMD manages to fix the random server reboots or we find a workaround.
- Fast movement of nodes between async / online without EPN expert intervention.
- 2 goals I would like to set for the final solution:
- It should not be needed to stop the SLURM schedulers when moving nodes, there should be no limitation for ongoing runs at P2 and ongoing async jobs.
- We must not lose which nodes are marked as bad while moving.
- Interface to change SHM memory sizes when no run is ongoing. Otherwise we cannot tune the workflow for both Pb-Pb and pp: https://alice.its.cern.ch/jira/browse/EPN-250
- Lubos to provide interface to querry current EPN SHM settings - ETA July 2023, Status?
- Improve DataDistribution file replay performance, currently cannot do faster than 0.8 Hz, cannot test MI100 EPN in Pb-Pb at nominal rate, and cannot test pp workflow for 100 EPNs in FST since DD injects TFs too slowly. https://alice.its.cern.ch/jira/browse/EPN-244 NO ETA
- DataDistribution distributes data round-robin in absense of backpressure, but it would be better to do it based on buffer utilization, and give more data to MI100 nodes. Now, we are driving the MI50 nodes at 100% capacity with backpressure, and then only backpressured TFs go on MI100 nodes. This increases the memory pressure on the MI50 nodes, which is anyway a critical point. https://alice.its.cern.ch/jira/browse/EPN-397
- TfBuilders should stop in ERROR when they lose connection.
- Allow epn user and grid user to set nice level of processes: https://its.cern.ch/jira/browse/EPN-349
Other EPN topics:
AliECS related topics:
- Extra env var field still not multi-line by default.
GPU ROCm / compiler topics:
- AMD Status: Random server reboots are more rare with ROCm 6.3.1, but still happening. No improvement on the other issues. Send again a detailed list of issues to AMD, and will have a meeting on Thursday to discuss how to progress. From the 2 AMD engineers working for us, 1 is in parental leave and 1 retired, so currently no one available to debug. List of issues ordered by severity:
- Random server reboots on MI100: Tried several workarounds, but no solution found so far. Giada spotted some weird FairMQ problems in the large scale test, which could probably be due to some memory corruption happening.
- Random crashes on MI100 due to memory error, can be worked around by serializing all kernel and DMA transfers, which has 20% performance degradation.
- Miscompilation leading to crashes, worked around by changing our code, but compiler bug still there.
- Provide an RPM ROCm version with all fixes, so that we don't need to compile clang manually with custom patches.
- Proper way to enable amdgpu-function-calls instead of hacking AMD scripts and binaries.
- hipHostRegister has become very slow when more than 1 GPU visible (via ROCR_VISIBLE_DEVICES).
- Try to find a better solution for the problem with __device__ inline functions leaking symbols in the host code.
TPC GPU Processing
- WIP: Use alignas() or find a better solution to fix alignment of monte carlo labels: https://its.cern.ch/jira/browse/O2-5314
- Waiting for TPC to fix bogus TPC transformations for good, then we can revert the workaround.
- Waiting for TPC to check PR which uses full cluster including average charge and occupancy map errors during seeding.
- Added protection that system headers cannot be included in GPU device code, which was the cause for the RTC compilation failures in November not seen in the CI.
- Added option requested by TPC to reject clusters for dEdx based on cluster flag mask.
- With POCL trunk and Clang 19, for the first time managed to run our OpenCL 2 code. After some fixes, full slice tracking with OpenCL2 was working with exact same results as OpenCL1 / CPU, thus OpenCL 2 now superseds OpenCL 1 code.
- With OpenCL 1 no longer needed, and I don't do any more tests with AliRoot in the last year, started a larger cleanup campaign:
- Removed all OpenCL 1 code, and merged OpenCL common and OpenCL2 into one OpenCL library.
- Removed all code related to AliRoot.
- Still ongoing to add a feature to the standalone benchmark to create new default calib objects if the format changes. This will allow to keep using Run 2 data with the Standalone benchmark if they are without distortions.
- Removed all workarounds we had to run with ROOT 5, for compilers that do not support C++11, and for missing C++17 support.
- Unfortunately, OpenCL for C++ 2021 is still at C++17, so we cannot yet have full C++20 support.
- In the process of removing several code paths that are obsolete now, since they were used only in legacy code (e.g. OpenCL1 could not run all kernels on CPU, AliRoot needed slice data output between sector tracing and merging).
- Pending OpenCL2 issues:
- printf not working due to confirmed bug in clang, fix is being prepared. Prevents further debugging for now.
- GPU MemClean not working in TPC clusterization, need to debug.
- Crash in merger, which can be worked around by disabling clang SPIRV optimization. Probably bug in clang, but need to fix printf first to debug.
- Also with optimization disabled, crashing later in TPC merging, need printf to debug.
- Next high priority topic: Improvements for cluster sharing and cluster attachment at lower TPC pad rows.
TPC processing performance regression:
- Final solution: merging transformation maps on the fly into a single flat object: Still WIP
General GPU Processing