Migrating CUDA to SYCL: A HEP Case Study with ROOT RDataFrame

Jolly Chen; Monica Dessole; Ana-Lucia Varbanescu

doi:10.1145/3648115.3648122

Migrating CUDA to SYCL: A HEP Case Study with ROOT RDataFrame

Jolly Chen, Monica Dessole, Ana-Lucia Varbanescu

Computer Architecture Design and Test for Embedded Systems

Research output: Contribution to conference › Poster › Academic

Abstract

The world’s largest particle accelerator, located at CERN, produces petabytes of data that need to be analysed efficiently, to study the fundamental structures of our universe. ROOT RDataFrame is an open-source C++ data analysis interface, developed for this purpose. With the increasing heterogeneity in computing facilities, it becomes crucial to efficiently support GPGPUs. In this work, we describe our prototyped CUDA implementation and its migration to SYCL, focusing on a core high-energy physics (HEP) operation in RDataFrame – histogramming. In addition, we provide a comparative performance analysis of two SYCL compilers (AdaptiveCpp and DPC++) against the reference CUDA implementation.

Original language	English
Number of pages	2
DOIs	https://doi.org/10.1145/3648115.3648122
Publication status	Published - 8 Apr 2024
Event	12th International Workshop on OpenCL and SYCL, IWOCL 2024 - Chicago, United States Duration: 8 Apr 2024 → 11 Apr 2024 Conference number: 12

Workshop

Workshop	12th International Workshop on OpenCL and SYCL, IWOCL 2024
Abbreviated title	IWOCL 2024
Country/Territory	United States
City	Chicago
Period	8/04/24 → 11/04/24

Keywords

n/a OA procedure

Access to Document

10.1145/3648115.3648122

Cite this

@conference{55474296e3b84f2bb96f39739687ddd7,

title = "Migrating CUDA to SYCL: A HEP Case Study with ROOT RDataFrame",

abstract = "The world{\textquoteright}s largest particle accelerator, located at CERN, produces petabytes of data that need to be analysed efficiently, to study the fundamental structures of our universe. ROOT RDataFrame is an open-source C++ data analysis interface, developed for this purpose. With the increasing heterogeneity in computing facilities, it becomes crucial to efficiently support GPGPUs. In this work, we describe our prototyped CUDA implementation and its migration to SYCL, focusing on a core high-energy physics (HEP) operation in RDataFrame – histogramming. In addition, we provide a comparative performance analysis of two SYCL compilers (AdaptiveCpp and DPC++) against the reference CUDA implementation.",

keywords = "n/a OA procedure",

author = "Jolly Chen and Monica Dessole and Ana-Lucia Varbanescu",

year = "2024",

month = apr,

day = "8",

doi = "10.1145/3648115.3648122",

language = "English",

note = "12th International Workshop on OpenCL and SYCL, IWOCL 2024, IWOCL 2024 ; Conference date: 08-04-2024 Through 11-04-2024",

}

TY - CONF

T1 - Migrating CUDA to SYCL

T2 - 12th International Workshop on OpenCL and SYCL, IWOCL 2024

AU - Chen, Jolly

AU - Dessole, Monica

AU - Varbanescu, Ana-Lucia

N1 - Conference code: 12

PY - 2024/4/8

Y1 - 2024/4/8

N2 - The world’s largest particle accelerator, located at CERN, produces petabytes of data that need to be analysed efficiently, to study the fundamental structures of our universe. ROOT RDataFrame is an open-source C++ data analysis interface, developed for this purpose. With the increasing heterogeneity in computing facilities, it becomes crucial to efficiently support GPGPUs. In this work, we describe our prototyped CUDA implementation and its migration to SYCL, focusing on a core high-energy physics (HEP) operation in RDataFrame – histogramming. In addition, we provide a comparative performance analysis of two SYCL compilers (AdaptiveCpp and DPC++) against the reference CUDA implementation.

AB - The world’s largest particle accelerator, located at CERN, produces petabytes of data that need to be analysed efficiently, to study the fundamental structures of our universe. ROOT RDataFrame is an open-source C++ data analysis interface, developed for this purpose. With the increasing heterogeneity in computing facilities, it becomes crucial to efficiently support GPGPUs. In this work, we describe our prototyped CUDA implementation and its migration to SYCL, focusing on a core high-energy physics (HEP) operation in RDataFrame – histogramming. In addition, we provide a comparative performance analysis of two SYCL compilers (AdaptiveCpp and DPC++) against the reference CUDA implementation.

KW - n/a OA procedure

U2 - 10.1145/3648115.3648122

DO - 10.1145/3648115.3648122

M3 - Poster

Y2 - 8 April 2024 through 11 April 2024

ER -

Migrating CUDA to SYCL: A HEP Case Study with ROOT RDataFrame

Abstract

Workshop

Keywords

Access to Document

Fingerprint

Cite this