OFFLINE

 

The second half of 2013 has been devoted to the validation of the 6_2_X cycle, while developing the 7_0_X one. In order to cope with the needs of the Physics group, its deadline has been extended till the beginning of next year.

The Core Software team has completed the migration of the CMSSW software repository from CVS to Git, hosted by GitHub (https://github.com/cms-sw/cmssw). To date almost 400 developers have forked the official repository and we have over 80 single contributors per month. For a quick overview of what's happening in CMSSW you can look at the "Repository Pulse" pages at https://github.com/cms-sw/cmssw/pulse/monthly. It is highly recommend, although not required, to move usercode areas under GitHub. We provide support for centrally supported "cms-analysis" repositories (https://github.com/cms-analysis).

Work progresses on allowing CMSSW jobs to utilise more than one CPU core, i.e. the multi-threaded framework. A major milestone was reached in September when the CMSSW framework became fully thread-safe and was able to run simple jobs utilising many CPU cores. The ongoing challenge is to update the simulation, reconstruction and analysis code so it can make efficient use of the new multi-core facilities offered by the multi-threaded framework.

The Generator group is facing a twofold challenge: support for the long-term legacy Monte Carlo production at 7 and 8 TeV based on the 5_3_X release, and the new 13 TeV production based on the 6_2_X release. An important step achieved for legacy production is the possibility to use different Pythia8 versions in 5_3_X, which allows having available the latest version of the program without removing the version used so far in the production cycle. Similarly we are working to backport Tauola++ in 5_3_X without breaking backward compatibility with the older fortran version of Tauola. The main change for 13 TeV production is the move of the baseline parton-shower code from Pythia6 to Pythia8. This change requires a novel validation effort, but it has also become an opportunity to review all our validation workflows. Attention is given to new NLO codes, as well as to a renewed tuning effort. In order to address all the challenges, three subgroups have been created last July: integration/validation, matrix element generators, and tuning.

In Simulation, there have been recent advances on several different fronts. For Fast Simulation, work continues on the integration of the Full Sim digitisation and pile-up mechanisms into the FastSim framework. Excellent MC-Data agreement is found for calorimeters, the capability to simulate out-of-time pile-up with FastSim is also now available, the capability to mix reconstructed tracks from the minbias pile-up events has been introduced for this purpose. Finally, several new flexible structures have been introduced to the FastSim code to help with simulations for a variety of upgrade geometries in the Tracker and calorimeters. To help with group management, we are looking to add two Level-3 positions, to be responsible for Tracker/Muon infrastructure and calorimetry.

The Full Simulation team has been working to integrate the interfaces required by the new version of Geant 4, version 10, which will be capable of multi-threaded processing. We expect to begin the adoption of Geant 4.10 early in the new year. Large gains in processing time have been made recently with the optimisation of the Russian Roulette configuration used to parametrise the impact of low-energy photon and neutron hits, with some smaller gains from an optimised treatment of mathematics libraries and other code changes. We see an almost 30–40% improvement in the time required to simulate interactions with Geant 4 using these methods. Going forward, we expect to deploy a new workflow for pile-up simulation called "Pre-Mixing" in the early part of the new year. This will substantially reduce the number of files needed for Monte Carlo production by creating single events with large amounts of pile-up, then using these for production.

The L1 Trigger offline software group is preparing the code necessary to emulate the L1 Trigger for the 2015 run, using an interim trigger, and in 2016 and beyond, using the full upgrade trigger. The design effort has focused on unifying the code across subsystems, by developing common solutions to shared problems such as an evolving hardware configuration.

Since our summary in the previous report in June, several developments took place in CMS reconstruction software. The legacy production release 53X remained generally unchanged until September (5_3_12_patch1), when new algorithms used in b-tagging and in tau reconstruction were added to aid future or ongoing analyses. A similar work was added on top of the 6_2_X cycle. Among the most interesting features coming with the new 7_0_X release, there are the reorganization of the electron/photon (eγ) reconstruction, which will deliver eγ reconstruction fully integrated in the particle-flow algorithmic flow and provide a more complete Global Event Description. Many major technical infrastructure changes were also made in the software to comply with the multi-threading environment where we plan to be able to run next year. Work to converge on a tracking configuration robust under a high-pileup environment with 25ns bunch spacing is ongoing as well.

The Physics Analysis Tools software is very stable and routinely used in physics analyses by the collaboration. Beside detailed web documentation, users are also regularly trained to use PAT during five-days tutorials. A multi-variate electron-identification analysis package has been added and the software has been adapted to the new defaults in the tau and b-tagging reconstructions. In development releases, final polishing of the so-called "unscheduled" processing mode takes place as well as updates required for a multi-threaded CMSSW framework version, where the latter task affects almost all software modules.

The basis for upgrade software development has moved to the CMSSW_6_2_X release cycle, which will be the basis for the 2014 simulations in support of the Phase 2 Technical Proposal. Currently CMSSW_6_2_0_SLHC4 is available, supporting the Phase 1 Pixel and HCAL upgrades as well as the BE5D Phase 2 outer tracker, including critical developments in the Phase 1 HCAL local reconstruction necessary for good performance at high pile-up. Other recent developments include the integration of tracking using the BE5D tracker. The next step is to refine the tracking configuration used to be optimised for the very high pile-up expected in Phase 2. Other geometry components targeted for the Technical Proposal release are in progress, and targeting either fast simulation or full simulation use cases.


by Fabio Cossutti, Peter Elmer, and SEXTON-KENNEDY Elizabeth