SYCL* & SYCLomatic Add Multiplatform Parallelism to Workloads

SYCL* & SYCLomatic Add Multiplatform Parallelism to Workloads

@IntelDevTools

Subscribe Now

Stay in the know on all things CODE. Updates are delivered to your inbox.

Sign Up

Overview

Enable your C++ applications with SYCL*-based parallelism and GPU offload using SYCL, an open parallel-programming framework based on C++ that allows developers to target multivendor hardware accelerators.

Topics covered:

  • SYCL kernel templates used to add code and data parallelism
  • The steps for using the Intel® DPC++ Compatibility Tool and SYCLomatic to migrate CUDA* code to SYCL (SYCLomatic is the open source version of the Compatibility Tool.)
  • Diagnostic message handling and resolution during migration, including resolving code-migration warnings and error messages
  • Library call migration from the CUDA Basic Linear Algebra Subroutine library (cuBLAS), NVIDIA* CUDA Fast Fourier Transform (cuFFT), and NVIDIA Random Number Generation library (cuRAND) to the Intel® oneAPI Math Kernel Library (oneMKL)
  • An example walk-through of an N-Body motion simulation workload migration to SYCL

Skill level: Intermediate

 

Jump to:


You May Also Like
 

 Highlights

01:43

Options for adding platform parallelism: adding simple parallel circle kernels or migrating GPU offload code.
02:13

Benefits of adopting the SYCL programming model.
04:00

Introduction to oneAPI and SYCL.
07:11

How to add simple parallel kernels to C++.

how to use a basic parallel kernel

how to use an N D range kernel

how to use a hierarchical parallel kernel

00:23.40

How to migrate CUDA code to SYCL.

a pictorial workflow for migrating CUDA to SYCL

00:21.28

How to use an optional intercept build step before running the migration tool.

a pictorial workflow for an intercept build

00:23.58

Basic options within the migration tool.

a list of basic code options with explanations

00:27.22

How the code looks before and after migration.

a side by side comparison of how cuda code looks after its migration to sycl

another side by side comparison of how cuda code looks after its migration to sycl

00:34.20

How the Intel DPC++ Compatibility Tool and SYCLomatic migrate library calls.

00:36.54

An overview of features in oneMKL that support migrating library calls.

a side by side comparison of how cuda calls look after their migration to sycl using one m k l

00:42.39

Diagnostic warning messages that appear during migration.

00:44.39

A demonstration of the migration process.

00:51.36

Ways to contribute to SYCL.

00:54.56

Q&A.

Featured Software

Download the following as stand-alone products:

Or get them as part of the Intel® oneAPI Base Toolkit.

 

Download Code Samples