Talks & Seminars
|Challenges in Compiling for Graphics Processing Units|
|Jayant Kolhe and Dibypran Sanyal, nVidia|
|Date & Time: November 6, 2006 11:45|
|Venue: SIC-301, KReSIT|
In this talk, we will present rising need, importance and set of new challenges we face in building compiler related technology to emerging field of GPUs (Graphics Processing Units). GPUs are becoming highly programmable and to use full potential of these GPUs, need for programming these GPUs at higher level and generating optimal code for higher level programs is increasing. Since GPUs have significantly different characteristics than CPUs in certain aspects and programming models are different, there are several interesting problems we face in building compilers for GPUs. In this talk, we will briefly discuss such problems, efforts done to solve these problems and open problems that might be interesting to research community.
|Speaker Profile: |
Jayant Kolhe is sr. manager of compiler technologies at NVIDIA Corp., Santa Clara. Jayant has been instrumental designing and building programmability in GPUs at NVIDIA and has worked on designing Cg programming language, programming model for GPUs and lately, programming model for GPGPU (General purpose programming on GPUs). Before working on compilers for GPUs, Jayant worked on building Graphics libraries such as OpenGL and real time rendering pipeline at NVIDIA and Evans & Sutherland. Jayant has done his bachelors from IIT, Bombay and Masters from UNC Chapel Hill and his interests include compilers, graphics, operating systems and parallel programming.
Dibyapran Sanyal is manager of compiler team at NVIDIA, Pune. Dibyapran is currently involved in building up a world class compiler team at NVIDIA's Pune Design Center. Before working for NVIDIA, Dibyapran has worked on dynamic binary compilation at Hewlett-Packard and on C to RTL compilation at Poseidon Design Systems. Dibyapran's areas of interest are compiler optimizations, in particular machine dependant and parallelizing optimizations.