Port a custom-made TCP/IP stack (called IxStack) from running on a CPU to run on a GPU. The GPU will receive the configuration parameters of the test and start generating TCP/IP packets and send them to the network card (TX path). When the network card receives some packets, the packets will be parsed by the IxStack running on GPU and update the internal state machine of IxStack.

You’ll Work On creating an application that will run on the GPU:

🔹Copy the config to the GPU 🔹Design the TX path to run multiple packet tasks in parallel 🔹Design the RX path to parse the packets and group them for better parallel processing 🔹Copy the packets to/from the GPU directly from/to the network card

What you will gain:

• Compile a GPU application using NVidia CUDA or OpenCL
• Design a GPU application by thinking of the things that can be done in parallel
• Run CUDA/OpenCL kernels and to performance profiling
• Interact with the network card driver to pass the packets directly to/from the network card to the GPU.