Closed GNiendorf closed 2 months ago
/run all
The PR was built and ran successfully in standalone mode. Here are some of the comparison plots.
The full set of validation and comparison plots can be found here.
Here is a timing comparison:
Evt Hits MD LS T3 T5 pLS pT5 pT3 TC Reset Event Short Rate
avg 50.6 328.3 127.7 58.1 105.6 506.5 124.4 167.5 104.5 2.4 1575.5 1018.4+/- 265.5 434.2 explicit_cache[s=4] (target branch)
avg 50.6 330.6 128.0 56.3 103.8 496.9 121.1 165.2 104.6 2.9 1560.0 1012.5+/- 260.3 434.4 explicit_cache[s=4] (this PR)
The PR was built and ran successfully with CMSSW. Here are some plots.
The full set of validation and comparison plots can be found here.
/run all
The PR was built and ran successfully in standalone mode. Here are some of the comparison plots.
The full set of validation and comparison plots can be found here.
Here is a timing comparison:
Evt Hits MD LS T3 T5 pLS pT5 pT3 TC Reset Event Short Rate
avg 47.8 322.3 119.2 47.1 96.1 506.8 118.9 158.7 103.8 2.4 1523.1 968.6+/- 254.1 415.1 explicit_cache[s=4] (target branch)
avg 47.4 322.3 119.7 48.1 94.4 549.0 119.5 160.1 112.1 2.7 1575.3 978.9+/- 255.4 431.5 explicit_cache[s=4] (this PR)
The PR was built and ran successfully with CMSSW. Here are some plots.
The full set of validation and comparison plots can be found here.
@ariostas Is there any way you could add an optional parameter to the CI to test with the low pT cut? The only thing that would have to change is adding a --ptCut 0.6 for the standalone code, although I'm not sure how you would update the CMSSW run.
Is there any way you could add an optional parameter to the CI to test with the low pT cut?
Yeah, I'll figure out how one is supposed to adjust parameters in CMSSW and I'll add the option to the CI.
/run all lowpt
It should work now unless I messed something up. You'll just have to add "lowpt" somewhere in the comment, and if you want to test both setups then you'll just have to make two comments. For now, I didn't make the pt cut configurable (so only 0.8 and 0.6), but if there's a need I can do that.
/run all lowpt
(I did mess something up. Trying again.)
There was a problem while building and running with CMSSW. The logs can be found here.
The PR was built and ran successfully in standalone mode (low pT setup). Here are some of the comparison plots.
The full set of validation and comparison plots can be found here.
Here is a timing comparison:
Evt Hits MD LS T3 T5 pLS pT5 pT3 TC Reset Event Short Rate
avg 54.6 327.5 361.1 142.7 442.5 1128.1 269.4 772.5 246.0 2.0 3746.3 2563.6+/- 814.1 982.2 explicit_cache[s=4] (target branch)
avg 55.9 339.6 383.3 190.3 577.4 1239.4 291.2 836.1 268.6 15.6 4197.3 2902.0+/- 997.9 1144.7 explicit_cache[s=4] (this PR)
The PR was built and ran successfully with CMSSW (low pT setup). Here are some plots.
The full set of validation and comparison plots can be found here.
@ariostas Thank you for updating the CI! @slava77 I think this PR is ready for review when you get a chance.
/run all
The PR was built and ran successfully in standalone mode. Here are some of the comparison plots.
The full set of validation and comparison plots can be found here.
Here is a timing comparison:
Evt Hits MD LS T3 T5 pLS pT5 pT3 TC Reset Event Short Rate
avg 47.3 324.1 120.2 47.1 96.8 505.3 119.3 160.1 104.4 1.9 1526.4 973.8+/- 256.0 416.4 explicit_cache[s=4] (target branch)
avg 48.2 328.1 120.3 47.0 97.4 550.5 120.2 162.8 112.8 2.2 1589.4 990.7+/- 254.7 430.3 explicit_cache[s=4] (this PR)
The PR was built and ran successfully with CMSSW. Here are some plots.
The full set of validation and comparison plots can be found here.
Here is a timing comparison:
pLS time went up from 505 to 551. Is this real? (I can't quite match with the changes in the code)
@GNiendorf please run the timing locally to check. Thank you.
@slava77 I don't see any timing difference locally
Current CPU timing (4555, 7f3fe16b8e3):
This PR CPU timing:
Current GPU timing (4555, 7f3fe16b8e3):
This PR GPU timing:
This PR adds the occupancies for the 0.6 GeV pT threshold. It changes the current if statements for the occupancies into a 2d matrix of category and eta number for cleaner code. There are 2 matrices for each object, the current 0.8 GeV occupancies matrix and an additional matrix for the 0.6 GeV occupancies. This PR also cleans up some code copy-paste related to the occupancy kernels for selecting the relevant eta and category bins. Lastly, I also added the script that was used to select the 0.6 GeV occupancies.