[PWGHF] TrackIndexSkimCreator: hoist collision-invariant slices out of the 2/3-prong loops#16666
[PWGHF] TrackIndexSkimCreator: hoist collision-invariant slices out of the 2/3-prong loops#16666ktf wants to merge 1 commit into
Conversation
|
O2 linter results: ❌ 0 errors, |
vkucera
left a comment
There was a problem hiding this comment.
Please add the file name in the PR title to specify the scope of the changes.
|
@vkucera improved. Actual validation of the code done at https://alimonitor.cern.ch/hyperloop/view-wagon/57243/test-statistics |
I think you confused it with another PR. |
|
yes, indeed. |
|
The description was meant for #16663 |
|
Error while checking build/O2Physics/o2 for 0c38cad at 2026-06-21 07:28: Full log here. |
|
What is this missing to be merged? |
|
@ktf Has this been validated, i.e. before-vs-after comparison of output and performance? |
|
I was actually hoping for you (PWGHF) to do it... |
In run2And3Prongs the negative-track index slice was rebuilt inside the positive-track loop, and the D* soft-pion slices inside the (pos,neg) pair body, though all three depend only on the collision. Hoist the neg-track slice to collision scope, and lazily materialize the soft-pion slices once per collision via std::optional. Pure loop-invariant code motion.