Cerebras CSL SDK for HPC Training Workshop

Zoom information will be shared with attendees
Murali Emani, Venkatram Vishwanath (LCF)

The Cerebras computation model uses a dataflow architecture. The CSL Language provides the ability to program the Cerebras Wafer Scale Engine (WSE) and its dataflow architecture. This workshop will provide an overview of CSL and the SDK. We will walk through porting several computational kernels using CSL to run on the WSE. The target audience is science application developers who write simulation and analysis code and would like to port these to the Cerebras system.

This will be held over two days - July 25 and August 1. On July 25, we will provide an overview of the Cerebras architecture, CSL language, porting common commutation kernels, among others.
On August 1, we will have a hands-on day to port computational kernels driven by application teams. The attendance for Day 2 might be restricted to teams willing to work on the system to port code over.

This training is for Argonne only.  We will record both days and make this available to attendees


    • 12:30 PM 1:30 PM
      Cerebras Architecture Overview 1h
    • 1:30 PM 3:00 PM
      CSL SDK Overview 1h 30m

      CS Language overview and using the SDK

    • 3:00 PM 4:30 PM
      Porting and running Computational Kernels 1h 30m
    • 12:30 PM 4:30 PM
      Hands-On Porting 4h