Davide Rossetti

Background

About Davide Rossetti

Network interconnect optimized for GPUs: SW/HW design and development of a high-performance 3D Torus cluster interconnect with GPU optimizations: direct access to GPU memory using P2P/GPUDirect over the PCIe bus, integration with CUDA APIs, etc- Operating System development: device driver development mainly under Linux kernel on Intel architecture, deep knowledge of issues and bottlenecks as of the HW interaction between CPUs and PCI/PCI-X/PCIe busses- High Performance Computing: writing scientific/numeric programs to be run on large number of processors with high performance networks- Networking interconnects HW/SW design: design of high performance PCI cards, at the level of the HW DMA engine design principles and kernel level integration- Development of application level libraries: communications libraries (MPI), numerical libraries (BLAS), graphics toolkits- Processor architectures: design of custom processor architectures optimized on specific numerical problems, deep knowledge of commercial architectures (Intel, AMD, Atmel DSPs) with their issues- Computer Graphics: OpenGL, X11, Gnome/Gtk, KDE/Qt, MS Win32, MFC- Web technologies: small scale experiences with PHP, Perl, ASP- Multi-threaded programming: OpenMP, MS Win32 threads, Posix threads - Compiler Technologies: contributed to the INFN OpenZz dynamic parser, added support for native complex type to GCC (to support apeNEXT architecture)- Debugging of complex HW/SW systems. No fear of hard work on exciting, novel, mixed software and hardware fields. publications: http://scholar.google.com/citations?hl=en & Specialties: SW/HW platforms based on PCIe and x86 architecture. Strong C/C++ language skills. CUDA. Many-Core/GPU computing. Computing Architecture Design. Software Design and Development. Multi-treaded, Multi-core, Multi-processor, parallel programming.