Hello! My name is Hervé Yviquel. I am a research scientist in parallel computing. This means that I might be able to help if you have a computing problem that you want to solve faster, for example a computer simulation solving a scientific problem. I have been recently contracted as assistant professor at the Computing Institute of UNICAMP and this is an important step in my international career journey. In 2015, I went to Brazil to work with Prof. Guido Araujo as a post-doctoral researcher of the Center for Computing in Engineering and Sciences (CCES), a bit more than a year after defending my PhD at the University of Rennes 1 in France. I spent almost five years as a postdoc of the center allowing me to specialize myself in High-Performance Computing (HPC) and Scientific Computing which are my main research interests today. During my postdoc, I worked on simplifying the programming of computer clusters and supercomputers. We developed a compiler and runtime infrastructure allowing scientists to parallelize their applications by adding simple code directives into their program. Our approach is based on the OpenMP standard which is already largely used by industry and academia, allowing relatively easy adoption. My time at the center has been a truly rich experience thanks to the various collaborations with scientists from other fields like chemistry, physics and mechanics. I always liked teamwork and I really enjoy studying new scientific problems to find the best way to parallelize them. Thanks to the CCES, I also had the opportunity to do a one year postdoc internship at the Barcelona Supercomputing Center, a world class supercomputing research center in Spain. I am still collaborating today with the center as a faculty member of the Computing Institute of UNICAMP: I am currently helping several CCES researchers to parallelize their computing problems and I am frequently using the computing infrastructure of the center. To sum up, participating in a multidisciplinary computing center such as the CCES has been a crucial experience in my academic career.
Task graph of the LU factorization of a complex matrix: the nodes are computational kernels to be parallelized and the edges are data dependencies constraining the parallelization.