Python-Based Matrix Operation Accelerator


Product Description

Xilinx’s GEMX (General Matrix Operation) library provides a set of high performed engines for accelerating applications that heavily depend on matrix operations.  This library comes with a set of Python APIs to enable software, especially Python developers to easily leverage the performance advantage of those engines.

Key Features and Benefits

  • High performance dense and sparse matrix operation accelerator
  • Reduce data movement overhead via instruction-controlled engines
  • Support row-major format dense matrix operations
  • Support COO format sparse matrix operations
  • Support fully connected network operations
  • Easy-to-use and highly efficient Python APIs

Accelerator Card Support

Getting Started on Nimbix Cloud

Access Example Usage Code

Test drive in the Nimbix cloud

1. Download GEMX Python interface package

2. Access Xilinx Alveo card in the Nimbix cloud

  • Follow the steps to login to your Nimbix account.
  • Launch application “Xilinx SDAccel Development & Alveo FPGA 2018.3”  Select “Desktop Mode with FPGA”.
  • Choose machine type “16 core, 128 GB RAM, Xilinx Alveo U200 FPGA (nx5u_xdma_201830_1)”.

3. Run examples

  • Copy the xilinx-alveo-gemx-python-interfaces-v1.1.tar file to your work space on the Nimbix node
  • Unzip the downloaded xilinx-alveo-gemx-python-interfaces-v1.1.tar file via command
    tar -xvzf xilinx-alveo-gemx-python-interfaces-v1.1.tar
  • Navigate to Alveo_Python folder
  • Follow the steps described in the document below to set up your environment
  • More instructions about the Python examples and APIs can be found in
  • GEMX engine specification can be found here .