New Release v0.17.1
Increased bandwidth and improved performance when using the SMI protocol
As a follow up to our last release, where we introduced our new SMI protocol for interacting with off-chip shared memory, this release brings the option to increase the bandwidth of these memory transactions. We have also included an SMI-ready template to help you get started with applying the new protocol to your own applications.
A new version of our tooling has been released this week with improved error messages and a bug fix so we now allow graph generation for FPGA code with no accompanying CPU command code.
To update your tooling you have two options. First run
reco version to see which version you currently have installed. Then:
- If you have v0.5.0 you can now simply run
reco updateand you will be guided through the process.
- If you have a version older than
v0.5.0please see our update instructions for guidance.
SMI increased bandwidth
A new field has been added to the program settings held in
reco.yml allowing you to specify using 512 bit memory transactions. This was previously set at 64 bit.
reco.yml now has the following format:
memory_interface: smi memory_width: 512 ports: 2
Have a go with our new SMI-ready template, you’ll find it in our tutorial materials repo. Guidance on forking this repo can be found in our second tutorial.
For a full description of SMI, please see our blog post on SMI here.
And for a for a full rundown of this release, see our release notes.
Currently we’re working on improving our web app, and creating some new tutorials to help you get the most out of our service. If there’s anything you’d like to see us cover, please let us know here.