November 29, 2021 to December 3, 2021
Virtual and IBS Science Culture Center, Daejeon, South Korea
Asia/Seoul timezone

Streamlining LHCb's GPU High Level Trigger for online and offline environments

contribution ID 755
Not scheduled
20m
Windmill (Gather.Town)

Windmill

Gather.Town

Poster Track 1: Computing Technology for Physics Research Posters: Windmill

Speaker

Niklas Nolte (Massachusetts Institute of Technology (US))

Description

The triggerless readout of data corresponding to a 30 MHz event rate at the upgraded LHCb experiment together with a software-only High Level Trigger will enable the highest possible flexibility for trigger selections. During the first stage (HLT1), track reconstruction and vertex fitting for charged particles enable a broad and efficient selection process to reduce the event rate to 1 MHz. Tracking and vertexing at 30 MHz represents a significant computing challenge. LHCb utilizes the inherent parallelism of the triggering process to meet throughput requirements with GPUs. This implies a major change in software that has to be integrated with data aquisition and the second HLT stage. This talk focusses on efficient task scheduling to process event batches and seamless integration and translation into existing trigger software to enable smooth operation in online and offline environments, with or without GPUs available.

Significance

We introduce scalable control flow into the GPU project Allen and streamline its operation such that it seamlessly integrates with the Gaudi project. It can now be executed as a Gaudi job, even without the existence of GPUs, reaping all the benefits from the existing infrastructure like the WLCG.

Speaker time zone Compatible with America

Primary author

Niklas Nolte (Massachusetts Institute of Technology (US))

Presentation materials