mkdir build
cd build
For CentOS 7.x, enable devtoolset-7 for C++14 support with:
scl enable devtoolset-7 bash
and continue with:
cmake ..
make DeepPointV2FPGA
In order to automate the building modes, an script for PasteBin has been developed to automatically upload the log files generated by XOCC during the compilation and the linking processes to PasteBin.com. Just make sure that it is enabled in the main CMakeLists.txt and username, password, and API key of your PasteBin account are set.
The linking process requires a large amount of free memory(~30GB of ram for 8 parallel jobs) and close to 5GBs of disk space which takes almost 14 hours to complete with a i7-6700HQ
machine.
Considering that step one is already done and current directory is build
. This step generates *.xo
files needed for the linking process.
For SW-Emulation:
make compile_swemu
For HW-Emulation:
make compile_hwemu
For HW(system build for real FPGA):
make compile_hw
Considering that steps one and two are already done and current directory is build
. This step generates requested *.xclbin
file needed for the host program.
For SW-Emulation:
make link_swemu
For HW-Emulation:
make link_hwemu
For HW(system build for real FPGA):
make link_hw
The autobuild scripts are intended to make building process on an AWS instance easier. They compile and link the project consecutively and take log of each step in a text file. Finally, after finishing up, the instance would be powered off with the sudo poweroff
command. (considering that the root user has no password)
For SW-Emulation:
bash autobuild_swemu
For HW-Emulation:
bash autobuild_hwemu
For HW(system build for real FPGA):
bash autobuild_hw
Considering that steps one, two and three are already done, current directory is build
and the default shell is bash
. This command is the unified solution to launch the host program in sw-emu
, hw-emu
or hw
modes.
sh LaunchDeepPointV2FPGA.sh
The launcher script forwards its arguments to the host program.