Introduction to oneAPI and the Intel® Developer Cloud

oneAPI is a productive, smart path to freedom for accelerated computing that delivers a unified programming model for development across diverse architectures. It includes:

• Unified and simplified languages
• Libraries for expressing parallelism delivering uncompromising built-in high-level language performance across a range of hardware including CPUs, GPUs, and FPGAs

This workshop is for developers with a foundational knowledge of C or C++, FPGA, and GPU programming.

In this workshop, Praveen Kundurthy (a developer evangelist from Intel for AI and oneAPI) covers:

• How oneAPI can solve the challenges of programming in a heterogeneous world
• The SYCL* language for high-performance computing (HPC) applications and oneAPI programming model
• How to use SYCL buffers and accessors for data and memory management between host and device
• The basics of graphs and dependences in SYCL
• Test-drive Intel® tools and libraries in the Intel® Developer Cloud
• How to get familiar with using a Jupyter* Notebook on Intel Developer Cloud for hands-on training through simple SYCL code examples

0:00 Introductions

2:16 Agenda

3:05 Learning objectives

3:47 Programming challenges for multiple architectures

4:45 Introducing oneAPI

5:53 Intel® toolkits

7:15 Intel® oneAPI Base Toolkit

9:20 Intel® DPC++ Compatibility Tool

10:28 Intel® VTune™ Profiler

11:19 Intel® Advisor

14:10 Set up Intel Developer Cloud and Jupyter* environments

19:20 SYCL essentials course

20:04 What is the oneAPI implementation of SYCL?

21:10 A complete SYCL program example

24:13 Buffer memory model

28:17 Submit to a device

29:35 Important classes in SYCL

31:25 Accessor modes

32:10 Parallel kernels

33:35 Basic parallel kernels

34:32 ND-Range kernels

37:09 SYCL code anatomy

41:18 Buffer: sub_buffers

43:40 sub_buffers example continued

45:39 Asynchronous execution

49:20 Synchronization: Host accessors

50:40 Synchronization: Buffer destruction

51:45 Custom device selector

53:35 Hands-on lab

1:05:08 Module 2

1:15:15 Execution graph scheduling

1:16:10 Read after write (RAW)

1:17:45 Write after read and write after write

1:19:18 Linear dependency chain graphs and Y pattern graphs

1:20:20 Linear dependence using in-order queue example

1:21:45 Y pattern using an example of in-order queues

1:23:03 Recap of workshop