Huxelerate Performance Estimator¶
Huxelerate Performance Estimator allows to:
Quickly estimate attainable performance on a variety of computing architectures
Visually compare multiple architectures to identify to most suitable for your computational workload
Guide you through the optimization process to achieve desired performance
The tool is based on the Berkeley Roofline Model adapted to account for reconfigurable architectures and extended to provide analysis that are code and architecture dependent.
You can run Huxelerate Performance Estimator directly on your local machine through a Docker container. The Estimator will produce a profiling report file, that needs to be uploaded online to obtain the final report. All your reports will be accessible from your Performance Estimation page online. Registration is required only to upload the profiling report!
Note
Huxelerate Performance Estimator is always updated with new computing architectures to provide you the most comprehensive comparison and guide your architectural choices.
Quick Start¶
To Quick Start using Huxelerate Performance Estimator, you can download the Docker container containing all the necessary software:
1 | docker pull huxelerate/performance_estimator:latest
|
Note
If you do not have Docker installed on your local machine, you can easily setup Docker following the official Docker guide.
Once the Docker container is ready, there are three steps to your performance estimate:
Profile your application using the Performance Estimator
Register to the online platform and upload the profiling report
Analyze the profiling report