SYCL* Essentials: Code with Intel® oneAPI DPC++ Library APIs

This two-hour workshop focuses on specific techniques to gain maximum optimization when programming with the Intel® oneAPI DPC++ Library (oneDPL). Based on familiar standards, the APIs for oneDPL components include the C++ Standard Template Library (STL), the Parallel API, and an API for SYCL* kernels. These APIs help maximize productivity and performance across a wide range of accelerators and processors, including CPUs, GPUs, and FPGAs.

This workshop shows you how to:

• Use oneDPL to streamline SYCL programming.
• Take advantage of Parallel API—an implementation of the C++ Standard Template Library algorithms and execution policies used in heterogeneous computing.
• Implement oneDPL algorithms using buffers and Unified Shared Memory (USM).
• Implement iterators in the oneapi::dpl namespace.
• Use oneDPL to implement C++20 Ranges.
• Build an application with oneDPL on the Intel® Developer Cloud to perform gamma correction.

Skill level: Intermediate

0:00 Introduction

1:42 Agenda

3:42 Programming challenges for multiple architectures

4:40 Introducing oneAPI

5:47 oneAPI industry initiative

7:00 Intel oneAPI Base Toolkit

9:18 How the Intel® Developer Cloud for oneAPI works

10:15 Set up an Intel Developer Cloud account

11:23 Walk through a Jupyter* Notebook

13:52 How to get to Module 7 for oneDPL

17:03 Steps to getting started

20:58 Learning objectives

21:09 What is oneDPL?

24:32 Examples of implementing oneDPL

28:10 oneDPL example

30:07 How oneDPL works

30:48 Multiple oneDPL algorithms

32:19 Buffer iterators

34:00 USM pointers

35:58 USM allocators

38:50 Module 7: Hands-on workshop

39:48 Simple oneDPL example

43:35 Example of using multiple oneAPI algorithms

45:20 Example of oneDPL buffer iterators

47:52 Example of oneDPL USM pointers and allocators

51:45 oneDPL gamma correction example

1:01:40 Q&A

1:04:30 The parallel APIs of oneDPL

1:05:58 Reduce by segment

1:08:36 Inclusive scan by segment

1:10:56 Exclusive scan by segment

1:12:28 Binary search

1:14:03 Lower bound

1:16:27 Upper bound

1:17:40 Counting iterator

1:20:28 Transform iterator

1:22:52 Zip iterator

1:24:50 Questions about upper bound

1:27:50 Function object classes

1:28:40 Hands-on: oneDPL parallel notebook

1:29:20 Example of reduce by segment

1:34:50 Example of upper bound

1:37:30 Example of an iterator

1:49:23 Example of a range-based API

1:54:08 Recap of the workshop

1:56:00 Q&A