GPGPU

PDSL performs a lot of curreting-edge research projects on graphics processing units (GPUs). GPUs are typically orders-of-magnitude faster than multi-core CPUs, and they are best embracing a concept of many-core technology today. In particular, we are interested in general-purpose computing on GPUs, a.k.a., GPGPU. Do you know the idea? It uses thousands of cores of  the GPU for general-purpose computing beyond graphics!  Image processing, search algorithms, scientific simulations, whatever parallel applications obtain significant performance improvements. GPUs however need a lot of better understandings to deploy in the real world. Topics of particular interest at PDSL include:

• First-class resource management
• Power and performance analysis
• Dynamic voltage and frequency scaling
• Low-latency zero-copy I/O
• On-device microkernel
• Real-time preemptive execution models

ATTENTION: We are developing Gdev, an open-source runtime driver for NVIDIA GPUs. This facilicates better understandings of GPU resource management. Why? Because you have full access to the source code!