Creating an Amazon FPGA Image (AFI) using AWSEducate

This document guides you through the steps to create an AWS Amazon FPGA Image (AFI), when using AWSEducate instance, which can run on AWS EC2 F1 instance to verify that the design works in hardware. It assumes that a full system (Vitis project) is built which consists of an host application (executable file) and an FPGA binary file (.xclbin).

Create an AFI

  1. Setup CLI and Create S3 bucket

    • Create an empty file called credentials using the following commands

      mkdir .aws
      cd .aws
      gedit credentials
    • Go to the Vocareum window and click on Account Details
    • Click on the Show button
    • Copy the content into the credentials file, save the file, and exit
    • Create a S3 bucket which will be used for registering xclbin (Note: touch command below is necessary to have non-empty folders. The files will be copied into these sub-folders

      aws s3 mb s3://<bucket-name> --region us-east-1 
      aws s3 mb s3://<bucket-name>/<dcp-folder-name>
      touch FILES_GO_HERE.txt
      aws s3 cp FILES_GO_HERE.txt s3://<bucket-name>/<dcp-folder-name>/
      aws s3 mb s3://<bucket-name>/<logs-folder-name>
      touch FILES_GO_HERE.txt
      aws s3 cp FILES_GO_HERE.txt s3://<bucket-name>/<logs-folder-name>/
  2. Configure aws by executing following command and providing credentials

    Note that anytime you want to create AFI, the credentials file content must be updated. You must rerun the command every time before you run the script if this is a new session

    aws configure
    AWS Access Key ID [****************J5NS]:  <hit enter> 
    AWS Secret Access Key [****************N4bG]:  <hit enter>
    Default region name [None]: us-east-1 <make sure us-east-1 is displayed, otherwise change it to it>
    Default output format [None]: json <make sure json is displayed, otherwise change it to it>
  3. Submit the xclbin to generate AFI (with extension awsxclbin) using the following command

    $VITIS_DIR/tools/ -xclbin=<path to an including.xclbin> -s3_bucket=<bucket-name> -s3_dcp_key=<dcp-folder-name> -s3_logs_key=<logs-folder-name>

    The bucket-name, dcp-folder-name, and logs-folder-name should match the one used while creating them

  4. Once submitted successfully, an time_stamp_afi_id.txt file will be created. Open the file and make a note of the afi-id. Execute the following command to see the status of the AFI:

    aws ec2 describe-fpga-images --fpga-image-ids <afi_id>

    Wait for about 30 minutes before the status changes from pending to available. You can log out and login back to check the status

Regenerate .awsxclbin

You can regenerate the .awsxclbin file as long as you have access to *agfi_id.txt and *.xclbin files

  1. Edit these variable with the corresponding names

    export xclbin=<xclbin_filename>
    export agfi_id=<*_agfi_id.txt>
    export awsxclbin=<output_name>
  2. Generate .awsxclbin file

    xclbinutil -i $xclbin --remove-section PARTITION_METADATA --remove-section SYSTEM_METADATA --replace-section BITSTREAM:RAW:${agfi_id} -o ${awsxclbin}.awsxclbin

Copyright© 2021 Xilinx