%0 Conference Paper %1 10.1145/3585341.3585369 %A Breyer, Marcel %A Van Craen, Alexander %A Pflüger, Dirk %B Proceedings of the 2023 International Workshop on OpenCL %C New York, NY, USA %D 2023 %I Association for Computing Machinery %K myown from:vancraen exc2075 %R 10.1145/3585341.3585369 %T Performance Evolution of Different SYCL Implementations Based on the Parallel Least Squares Support Vector Machine Library %U https://doi.org/10.1145/3585341.3585369 %X In 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 Vector Machine (SVM) implementation for huge dense data sets, but we have also created a representative benchmark for a frequently encountered task in scientific computing, a (implicit) matrix-vector multiplication. PLSSVM supports multiple backends—OpenMP, CUDA, HIP, OpenCL, and SYCL—to be able to target the most widely used hardware platforms in machine learning and scientific computing. In this paper, we use PLSSVM to compare different DPC++ and Open SYCL (formerly known as hipSYCL) versions over the period of one year. Furthermore, we compared two versions (one from February and the other from November 2022) with each other and report their respective performance evolution in depth. We also put these results in relation to our other implemented backends and report their performance portability on three different hardware platforms, an NVIDIA and AMD GPU and an Intel CPU. Our results show that installing new DPC++ and Open SYCL versions can have surprisingly vast impacts in both directions. In our case, the nd_range kernel runtimes were up to faster on an NVIDIA GPU when using a newer DPC++ compiler. Also for Open SYCL, using the new omp.accelerated compilation flow improves the nd_range performance on CPUs by over . When compared to OpenCL, in our results, SYCL also offers a better performance portability while being easier to use, indicated by drastically fewer lines of code needed in our PLSSVM library. While OpenCL only has a performance portability of , DPC++ achieved the highest value with within the performance metric provided by Pennycook et al. 23. The code, utility scripts, and documentation are all publicly available on GitHub: https://github.com/SC-SGS/PLSSVM. %@ 9798400707452

@inproceedings{10.1145/3585341.3585369, abstract = {In 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 Vector Machine (SVM) implementation for huge dense data sets, but we have also created a representative benchmark for a frequently encountered task in scientific computing, a (implicit) matrix-vector multiplication. PLSSVM supports multiple backends—OpenMP, CUDA, HIP, OpenCL, and SYCL—to be able to target the most widely used hardware platforms in machine learning and scientific computing. In this paper, we use PLSSVM to compare different DPC++ and Open SYCL (formerly known as hipSYCL) versions over the period of one year. Furthermore, we compared two versions (one from February and the other from November 2022) with each other and report their respective performance evolution in depth. We also put these results in relation to our other implemented backends and report their performance portability on three different hardware platforms, an NVIDIA and AMD GPU and an Intel CPU. Our results show that installing new DPC++ and Open SYCL versions can have surprisingly vast impacts in both directions. In our case, the nd_range kernel runtimes were up to faster on an NVIDIA GPU when using a newer DPC++ compiler. Also for Open SYCL, using the new omp.accelerated compilation flow improves the nd_range performance on CPUs by over . When compared to OpenCL, in our results, SYCL also offers a better performance portability while being easier to use, indicated by drastically fewer lines of code needed in our PLSSVM library. While OpenCL only has a performance portability of , DPC++ achieved the highest value with within the performance metric provided by Pennycook et al. [23]. The code, utility scripts, and documentation are all publicly available on GitHub: https://github.com/SC-SGS/PLSSVM.}, added-at = {2023-04-11T06:36:32.000+0200}, address = {New York, NY, USA}, articleno = {24}, author = {Breyer, Marcel and Van Craen, Alexander and Pfl\"{u}ger, Dirk}, biburl = {https://puma.ub.uni-stuttgart.de/bibtex/2f3b151af423c0fde516afa04cce85ffc/ipvs-sc}, booktitle = {Proceedings of the 2023 International Workshop on OpenCL}, doi = {10.1145/3585341.3585369}, interhash = {494c6a4cbe6b39183a5bee7e04920128}, intrahash = {f3b151af423c0fde516afa04cce85ffc}, isbn = {9798400707452}, keywords = {myown from:vancraen exc2075}, location = {Cambridge, United Kingdom}, numpages = {12}, publisher = {Association for Computing Machinery}, series = {IWOCL '23}, timestamp = {2023-11-16T00:54:15.000+0100}, title = {Performance Evolution of Different SYCL Implementations Based on the Parallel Least Squares Support Vector Machine Library}, url = {https://doi.org/10.1145/3585341.3585369}, year = 2023 }