Cross-architecture Kalman filter benchmarks on modern hardware platforms

D H Cámpora Pérez,N Neufeld,O Bouizi,O Awile

doi:10.1088/1742-6596/1085/3/032046

D H Cámpora Pérez, N Neufeld + Show 2 more

Open Access

https://doi.org/10.1088/1742-6596/1085/3/032046

Copy DOI

Abstract

The 2020 upgrade of the LHCb detector will vastly increase the rate of collisions the online system needs to process in software in order to filter events in real-time. 30 million collisions per second will pass through a selection chain where each step is executed conditional to its prior acceptance.The Kalman filter is a process of the event reconstruction that, due to its time characteristics and early execution in the selection chain, consumes 40% of the whole reconstruction time in the current trigger software. This makes it a time-critical component as the LHCb trigger evolves into a full software trigger in the upgrade.The algorithm Cross Kalman allows performance tests across a variety of architectures, including multi and many-core platforms, and has been successfully integrated and validated in the LHCb codebase. Since its inception, new hardware architectures have become available exposing features that require fine-grained tuning in order to fully utilize their resources.In this paper we present performance benchmarks and explore the Intel® Skylake and Intel® Knights Landing architectures in depth. We determine the performance gain over previous architectures and show that the efficiency of our implementation is close to the maximum attainable given the mathematical formulation of our problem.

Highlights

The projected upgrade of the LHCb detector at CERN in 2020 [1] will increase the estimated collision rate to 30 MHz, requiring data processing at 40 Tbit/s
The processing rate required in the software trigger will increase by a factor 40, due to both the increase in rate of events processed in software, and the influx of larger events
Cross Kalman is a cross-architecture Kalman filter optimized for a variety of SIMD processors [6] [7]. It is optimized for low-rank Kalman filters on multi and many-core processors, and we have successfully integrated it in the LHCb reconstruction software

Summary

Introduction

The projected upgrade of the LHCb detector at CERN in 2020 [1] will increase the estimated collision rate to 30 MHz, requiring data processing at 40 Tbit/s. It is optimized for low-rank Kalman filters on multi and many-core processors, and we have successfully integrated it in the LHCb reconstruction software It serves as a self-contained package to evaluate different architectures in the context of the LHCb upgrade. 2. A vectorized Kalman filter In the LHCb software level trigger, particle trajectory reconstruction (track reconstruction) involves a complex selection chain where the trajectory of each particle is determined in several steps, due to the various subdetectors that compose the LHCb tracker [9]. In order to minimize the impact of data transposition, we generate the static schedule prior to any data generation This way, memory references are populated as the Cross Kalman filter would expect them. We have made our application reentrant in order to support this version of the framework

Results

1.30 GHz 32 MB L2

Conclusions