Speaker
Description
In ALICE, LHC Run 3 marks a major step toward GPU-centric data processing. During the synchronous (online) phase, GPUs are fully dedicated to Time Projection Chamber reconstruction and compression. During the asynchronous (offline) phase, additional reconstruction tasks can be offloaded to GPUs to improve overall computing efficiency and throughput. We describe the porting of the full reconstruction chain to generic GPU platforms and its integration with the mainline ALICE GPU reconstruction framework. The algorithms are currently being commissioned for the upgraded Inner Tracking System (ITS2). The presentation covers integration, performance characterization, and validation against the CPU baseline. Results are shown for representative datasets used in physics data production — an important step toward commissioning the production pipeline. Finally, we present a time-aware extension to the reconstruction algorithm that compensates for time-dependent detector effects (...). We describe the method and demonstrate how it improves tracking performance and thus extends ALICE’s physics reach in Run 3.