# The ACTS project¶

The A Common Tracking Software (Acts) project is an attempt to preserve and evolve the track reconstruction software of the LHC era towards HL-LHC and beyond. It has been initiated in 2016 starting from the ATLAS Common Tracking Software. Given the changing computing landscape, dedicated care of parallel code execution is taken, and is written in C++17.

A coherent write-up of the project has been published in 2022 in Springer’s CSBS.

Note

Acts is designed as a library that contains components for assembling a track reconstruction suite for High Energy Physics and Nuclear Physics. Acts does not strive to provide a complete experiment framework, but rather modules and connections to be used within an experiment context. These connections contain e.g. binding mechanisms to different geometry libraries, a cost-free yet flexible mechanism to use experiment specific contextual data (calibrations, detector alignment, experiment conditions), or simply the possibility to integrate an external screen logging facility.

The library is structured as follows:

• The Core library contains considered production ready components that can be interfaced to experiment code

• The Plugin folder contains additional extensions that can be optionally switched on to use increase the functionality of the software suite, but also in general increase dependencies to other/thirdparty libraries

• The Fatras library contains a fast track simulation module, that is based on the same concepts that are used for the ATLAS Fatras fast track simulation

• An Examples folder that contains a minimal test framework used for showcasing and integration testing,

• A Tests folder that contains unit tests, benchmark tests and other integration tests

Tip

Developers and R&D meetings can be found in the Acts indico category.

## Philosophy¶

In order to minimize virtual function calls and complex inheritance schemes, Acts does - in general - not define module interfaces, but rather relies on a data-centric pattern. Wherever possible, composition is favoured over inheritance/code extension.

Code execution performance is a big focus in Acts in order to serve the needs of the HL-LHC experiments; however, this is attempted to be reached mainly by technical means rather then compromising the physics performance.

## R&D Testbed¶

Acts should also provide a testbed for fruitful algorithm R&D, and hence is closely coupled to the development of the Open Data Detector which can be built as a thirdparty dependency as part of the Acts project, and builds the backbone of the chain demonstrators.

In addition, two dedicated R&D lines are part of the acts-project, one for machine learning based/inspired modules acts-machine-learning, and one for massively parallel code execution acts-parallelization, which mainly focuses on GPU accelerators and portability.

Code spill-over from the R&D lines into the main Acts repository are performed on demand and depending on the maturity of the R&D projects.