This topic describes the best practices for heterogeneous computing services. Select from the following topics based on your business scenario to learn about the associated best practices.
Elastic GPU Service
- Deploy an NGC environment on a GPU-accelerated instance
Describes how to deploy a NVIDIA GPU Cloud (NGC) environment on a GPU-accelerated instance. In the example, the TensorFlow deep learning framework is used.
- Use RAPIDS to accelerate machine learning tasks on a GPU-accelerated instance
Describes how to use the NGC-based Real-time Acceleration Platform for Integrated Data Science (RAPIDS) libraries that are installed on a GPU-accelerated instance to accelerate data science tasks and machine learning tasks and improve the efficiency of using computing resources.
FPGA as a Service (FaaS)
- Best practices for the Register Transfer Level (RTL) design on FPGA-accelerated instances
- Project modes and directories used by RTL
Describes the project modes and directories used by the RTL compiler and provides a sample framework to help you understand how to use RTL.
- Use the RTL design on an f3 instance
Describes how to implement the RTL design on an f3 instance.
- Project modes and directories used by RTL
- Best practices for using Open Computing Language (OpenCL) on FPGA-accelerated instances
- Overview of the FaaS f3 SDAccel development environment
Describes the FaaS f3 SDAccel development environment. The FaaS f3 SDAccel development environment is based on Xilinx SDAccel dynamic 5.0. You can develop and apply the FaaS f3 SDAccel development environment based on OpenCL.
- Use OpenCL on an f3 instance
Describes how to use OpenCL on an f3 instance to create an image and burn the image to an FPGA.
- Overview of the FaaS f3 SDAccel development environment