[14.0.X] instrument fitVertices to output more information when failing assert (issue cms-sw/cmssw#44923)
#45631
Conversation
|
A new Pull Request was created by @mmusich for CMSSW_14_0_X. It involves the following packages:
@cmsbuild, @jfernan2, @mandrenguyen can you please review it and eventually sign? Thanks. cms-bot commands are listed here |
|
cms-bot internal usage |
|
urgent |
|
@cmsbuild, please test |
| ALPAKA_ASSERT_ACC(wv[i] > 0.f); | ||
| bool const wv_cond = (wv[i] > 0.f); | ||
| if (not wv_cond) { | ||
| printf("ERROR: wv[%d] (%f) > 0.f failed\n", i, wv[i]); |
There was a problem hiding this comment.
I've seen that addition of a printf requires a lot of registers.
How much slower is the code after this change, is it significant?
There was a problem hiding this comment.
Is it relevant? The printout should occur every other ten of millions events?
There was a problem hiding this comment.
as I recall, if it is compiled, that's enough
There was a problem hiding this comment.
I don't have any feeling (nor a measurement). If @cms-sw/heterogeneous-l2 have a feeling about it please chime in, otherwise it will have to be measured.
There was a problem hiding this comment.
Yes, it's relevant, because the increase in register use is there even if the branch is never taken at runtime.
In general I would prefer not to add any printf - but since we have been unable to reproduce the assertion offline, I think in this case the extra resource usage is warranted.
Assuming this helps figuring out the source of the problem and fixing it, afterwards we can move the printf inside some if constexpr (debug) { ... } clause.
There was a problem hiding this comment.
How much slower is the code after this change, is it significant?
Hard to guess, it would need to be measured 🤷🏻♂️
There was a problem hiding this comment.
We checked the HLT throughput (details in [*]).
The impact of this PR on the HLT throughput is well below 1%, and somewhat within the uncertainties of these estimates.
[*]
- Input data: run-383631, LSs 476-479, ~40k events (PU ~64).
- HLT menu:
/cdaq/physics/Run2024/2e34/v1.4.3/HLT/V2(current pp online menu). - Release:
CMSSW_14_0_13_patch1_MULTIARCHS(with and without this PR). - Node:
hilton-c2b01-44-01(same hardware as a 2022/23 HLT node) (CPUs: 2 AMD EPYC 7763 64-Core; GPUs: 2 NVIDIA Tesla T4). - 8 jobs, 32 threads and 24 streams per job.
- NVIDIA MPS enabled.
- Each of the two measurements reported above is the average of 4 repetitions.
|
+1 Size: This PR adds an extra 16KB to repository Comparison SummarySummary:
|
|
+reconstruction |
cmsbuild
added
reconstruction-approved
fully-signed
and removed
reconstruction-pending
pending-signatures
labels
|
This pull request is fully signed and it will be integrated in one of the next CMSSW_14_0_X IBs (tests are also fine) and once validation in the development release cycle CMSSW_14_1_X is complete. This pull request will now be reviewed by the release team before it's merged. @sextonkennedy, @antoniovilela, @rappoccio, @mandrenguyen (and backports should be raised in the release meeting by the corresponding L2) |
|
+1 |
backport of #45630
PR description:
Title says it all, as per #44923 (comment)
PR validation:
cmsswcompilesIf this PR is a backport please specify the original PR and why you need to backport that PR. If this PR will be backported please specify to which release cycle the backport is meant for:
Verbatim backport of #45630 for 2024 data-taking operations.