FPGA Configuration

System Requirements

Using DAPHNE's build script with FPGA kernels support requires the additional QUARTUSDIR environment variable to be defined. An example command is presented in fpga-build-env.sh or in the following command:

export QUARTUSDIR=/opt/intel/intelFPGA_pro/21.4

To build DAPHNE with FPGA kernels support, the flag --fpgaopencl has to be used:

./build.sh --fpgaopencl

To run the pre-compiled FPGA OpenCL kernels included in the DAPHNE repository, an installed and configured FPGA device is required. Our example kernels have been tested using an Intel(R) PAC D5005 card.

DAPHNE contains some example linear algebra kernels developed using the T2SP framework. The example pre-compiled FPGA kernels can be used at DaphneDSL level. To prepare the system for the pre-compiled FPGA kernels, some FPGA and OpenCL environment variables are required. The easiest way to set up the required variables is to use the init_opencl.sh script from installed Intel(R) Quartus software or from the Intel(R) OpenCL RTE or Intel(R) OpenCL SDK packages.

Example script usage:

source /opt/intel/intelFPGA_pro/21.4/hld/init_opencl.sh

For additional details, see the Intel guide or the SDK for OpenCL.

Pre-compiled FPGA Kernels

To use a pre-compiled FPGA kernel, an FPGA image is required (*.aocx). The FPGA device has to be programmed with a particular image which contains the required kernel implementation. An example FPGA programming command using example FPGA image:

aocl program acl0 src/runtime/local/kernels/FPGAOPENCL/bitstreams/sgemm.aocx

Additionally, the BITSTREAM environment variable has to be defind in the system as follows:

export BITSTREAM=src/runtime/local/kernels/FPGAOPENCL/bitstreams/sgemm.aocx

When another FPGA image contains an implementation for another required computational kernel, then the FPGA device has to be reprogrammed and the BITSTREAM environment variable has to be changed.