Robert Mueller-Albrecht
ABSTRACT
Since widening its scope and becoming its independent working group 5 years ago, SYCL has evolved into the alternative to CUDA by championing an open-source and standards-based approach to accelerator offload, including GPUs. SYCL became the cornerstone of the oneAPI industry initiative, which combines it with a range of library and API specifications to create a multiarchitecture, multi-vendor programming model with the components needed by a developer relying on an ecosystem of ready-to-use device interfaces for common optimized and specialized routines. In September last year, this industry initiative became the starting point of the Unified Acceleration Foundation (UXL), governed by the Linux Foundation and hosted by the Linux Foundation's Joint Development Foundation (JDF). Its steering members include Arm, Fujitsu, Google Cloud, Imagination Technologies, Intel, Qualcomm Technologies Inc., Samsung, and VMware. SYCL and the drive to wide SYCL 2020 adoption and compliance are still at the heart of this effort. To help enable ecosystem-wide usage across workloads and accelerator hardware, it must be possible to conveniently migrate the CUDA-based legacy codebase of popular applications to SYCL. SYCLomatic was developed to provide a tool to ingest a project targeting C++ with CUDA-based GPU offload and translate it into ready-to-build C++ with SYCL code. This poster and accompanying paper go into advanced considerations for successful, and performant migrated code.
- Intel. 2023. SYCLomatic. https://github.com/oneapi-src/SYCLomaticGoogle Scholar
- Intel. 2023. Velocity Bench. https://github.com/oneapi-src/Velocity-BenchGoogle Scholar
Recommendations
A Comparison of SYCL, OpenCL, CUDA, and OpenMP for Massively Parallel Support Vector Machine Classification on Multi-Vendor Hardware
IWOCL '22: Proceedings of the 10th International Workshop on OpenCLIn scientific computing and Artificial Intelligence (AI), which both rely on massively parallel tasks, frameworks like the Compute Unified Device Architecture (CUDA) and the Open Computing Language (OpenCL) are widely used to harvest the computational ...
Performance Evolution of Different SYCL Implementations based on the Parallel Least Squares Support Vector Machine Library
IWOCL '23: Proceedings of the 2023 International Workshop on OpenCLIn machine learning and scientific computing, some of the biggest challenges are efficient and performant portable computing. With our Parallel Least Squares Support Vector Machine (PLSSVM) library, we have not only developed an unrivaled Support ...
Exploring the performance and portability of the k-means algorithm on SYCL across CPU and GPU architectures
AbstractThe aim of SYCL is to reduce the gap between the performance and code portability of the main accelerators used in HPC, such as multi-vendor CPUs, GPUs, and FPGAs. To evaluate SYCL’s performance portability, this paper uses the k-means algorithm ...
Comments