Not even the blustery winter weather that struck the Brazilian town of Curitiba last week could slow down more than 1,500 students, researchers, developers, scientists and industry professionals as they lined up for NVIDIA’s first GPU Computing Developer Forum, held at the local Federal Parana University.
The two-day event’s opening keynote featured our own NVIDIA Fellow David Kirk, talking about scientific research using GPUs. Our Solutions Architects Steve Harpster and Ty McKercher also spoke, covering Kepler architecture, new features on CUDA 5.0, and VGX technology.
Among the most popular sessions was one by Pedro Lopes, from University of Sao Paulo, covering OpenACC – a new standard launched by NVIDIA, Cray, PGI and CAPS – which is becoming increasingly important in the region. Research centers interested in porting their legacy codes to NVIDIA GPUs are using it to simplify and accelerate the porting process.
Paulo Souza, a top CUDA programmer at Petrobras, the world’s fourth-largest oil company, talked about how the company is obtaining enormous advantages using GPUs — 20X in performance/node, 12X in performance/W, and 11X in performance/$.
And one of Brazil’s brightest mind, Jairo Panetta, head researcher for the National Institute for Space Research, talked explained how parallel computing is tackling the obstacles that big Brazilian companies have in the use of supercomputing.
“Supercomputing should be at society’s service,” he said. “When it’s not applied in companies that serve the population, it fails to fulfill its function.”
Among other highlights of the two-day event:
- A crowd of more than 200 students flocked to four CUDA training sessions presented by Federal Fluminense University professors. The university was just recently were announced as the first CUDA Center of Excellence in Brazil and Latin America.
- A total of 20 different universities and research centers from around the country displayed 24 posters on research projects using GPU computing.