Closed kiniry closed 4 years ago
@podhrmic You can either build this after the bitstream_gen
version with the right licenses lands, or I can, or we roll the two ideas for images together into a single image.
This is already done. The image doesn't contain the HDK, because: 1) it takes only a few minutes to download and build for the first time 2) IMHO setting up the correct version of the HDK should be a part of the CloudGFE bitstream build setup, the docker image only provides all necessary dependencies
Understood. Have you built and pushed the new image to Artifactory yet? I didn't see it yesterday.
Yes, the image is updated with a new license. Have a look at https://github.com/DARPA-SSITH-Demonstrators/BESSPIN-CloudGFE/issues/42 for use instructions. I will close the ticket once I update our host machines
Thanks. I'll try and give it a whirl this afternoon or over the weekend.
@kiniry did you have a chance to try it out?
I have successfully built the AWS hello world. I'll next take it for a spin in AWS before I call this confirmed.
Per my update on Mattermost around 13:00:
AWS FPGA: (19:41:24) - Build complete.
INFO: [Common 17-206] Exiting Vivado at Wed May 13 19:41:24 2020...
real 91m12.341s
user 144m36.260s
sys 8m31.053s
That's not bad, I'd say.
I have iterated on the Docker image and it is now reliable. See gfe:bitstream_gen
.
Such an image's parent would be our Vivado image and it would simply also contain an
aws-fpga
sandbox and we would have runhdk_setup.sh
during image build to cache the DCPs.Effort estimate: 2 hrs