p4haskell

A GPU-accelerated back end for the PR programming language.

Usage

1. The first step is to compile a p4 program with the GPU back end present in my fork of the p4c compiler¹, and dump the AST, for example: ./build/backends/gpu/p4c-gpu port-forwarding.p4 --toJSON port-forward.json
2. The next step is to use this compiler to generate a CUDA file, for example: stack run -- -i firewall.json > kernel.cu
3. The final step is to place this CUDA file into the runtime² as src/kernel.cu, and compile the runtime using cargo build --release. The runtime can then be executed using sudo env LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/lib" target/release/dpdk-gpu-test -s payload_1024.pcap

Footnotes

1. https://github.com/simmsb/p4c ↩

2. https://github.com/simmsb/dpdk-testing-stuff ↩