SDx Version used for AFI Development | Compatible SDAccel Runtime |
---|---|
2017.4 | Runtime installed by sourcing "sdaccel_setup.sh" while using HDK Ver 1.4.X when environment variable RELEASE_VER=2017.4 |
2018.2 | AWS FPGA Developer AMI 1.5.0 (XRT is pre-installed) or Runtime installed with XRT Version 2.1.0 |
2018.3 | AWS FPGA Developer AMI 1.6.0 (XRT is pre-installed) or Runtime installed with XRT Version 2.1.0 |
2019.1 | AWS FPGA Developer AMI 1.7.0 (XRT is pre-installed) or Runtime installed with XRT Version 2.1.0 |
-
Please note Amazon Linux 2 or Amazon Linux are not supported by Xilinx XRT at this time. Please use Centos/RHEL or Ubuntu when using Xilinx XRT Runtimes for the AFIs generated using Xilinx SDx 2018.2 and 2018.3 toolsets.
-
Launch an F1 instance using Centos 7 or Amazon Linux AMI's
-
Update to get the latest packages.
$ sudo yum update
- Reboot your Runtime Instance to ensure all updates are running.
- Install the required packages
$ sudo yum install git
$ sudo yum install gcc
$ sudo yum install gcc-c++
$ sudo yum install kernel-headers
$ sudo yum install kernel-devel
$ sudo yum --enablerepo=epel install ocl-icd ocl-icd-devel opencl-headers
$ sudo mkdir -p /etc/OpenCL/vendors/
- Launch an F1 instance using an Ubuntu 16.04 LTS
- Install the required updates
$ sudo apt-get update
- Reboot your Runtime Instance to ensure all updates are running.
- Install the required packages
$ sudo apt-get install gcc
$ sudo apt-get install g++
$ sudo apt-get install make
$ sudo apt-get install linux-headers-`uname -r`
$ sudo apt-get install linux-libc-dev
$ sudo apt-get install ocl-icd-dev ocl-icd-libopencl1 opencl-headers ocl-icd-opencl-dev
$ sudo mkdir -p /etc/OpenCL/vendors/
-
Using an instance running FPGA Developer AMI or an on-premises machine with access to a Xilinx SDAccel Tools Installation, first source $AWS_FPGA_REPO_DIR/sdaccel_setup.sh and then run following commands:
-
If using Ubuntu or Debian distributions set GLIBPATH env variable to Ubuntu. If using any other OS distribution set GLIBPATH to default.
-
Set env variable 'XLNXRTE' to intended runtime install directory path.
$ export GLIBPATH= <Ubuntu or default -- see note above>
$ export XLNXRTE=<your runtime install directory path>
$ mkdir -p $XLNXRTE/runtime/platforms/$(DSA)/driver
$ mkdir -p $XLNXRTE/lib/lnx64.o
$ mkdir -p $XLNXRTE/runtime/bin
$ mkdir -p $XLNXRTE/runtime/lib/x86_64
$ cp $SDACCEL_DIR/userspace/src/libxrt-aws.so $XLNXRTE/runtime/platforms/xilinx_aws-vu9p-f1-04261818_dynamic_5_0/driver/
$ cp $SDACCEL_DIR/tools/awssak/xbsak $XLNXRTE/runtime/bin/
$ cp $XIILNX_SDX/lib/lnx64.o/$GLIBPATH/libstdc++.so* xlnxrte/lib/x86_64/
$ cp $XIILNX_SDX/runtime/bin/xclbinsplit xlnxrte/runtime/bin/
$ cp $XIILNX_SDX/runtime/bin/xclbincat xlnxrte/runtime/bin/
$ cp $SDACCEL_DIR/aws_platform/xilinx_aws-vu9p-f1-04261818_dynamic_5_0/sw/lib/x86_64/libxilinxopencl.so $XLNXRTE/runtime/lib/x86_64/
$ cp /opt/Xilinx/SDx/2017.4.rte.dyn/setup.sh $XLNXRTE/
$ cp /opt/Xilinx/SDx/2017.4.rte.dyn/setup.csh $XLNXRTE/
- You may need to update path in $XLNXRTE/setup.sh and $XLNXRTE/setup.csh script to match your runtime instance.
- Copy $XLNXRTE directory created to $HOME on your Runtime Instance.
Xilinx SDx 2018.2: Install 2018.2 XRT.
Xilinx SDx 2018.3: Install 2018.3 XRT.
Xilinx SDx 2019.1: Install 2019.1 XRT.
- Log back on to the Runtime Instance:
$ export XILINX_SDX=$HOME/$XLNXRTE
- You should be able to run your FPGA accelerated application as described here, without needing to launch a new F1 instance
- Once you have your application running you should be able to create a Runtime AMI based your Runtime Instance as specified here.
- Please see Section 5 of the AWS Marketplace Seller's Guide for more details.