The Geometry Seminar will have a computationally-oriented talk from John Humphrey and Dan Price of EM Photonics, the developers of the widely-used CULA Tools implementation of LAPACK for NVIDIA‘s CUDA GPU programming platform.
Title: Using GPUs to Improve Numerical Calculations
Abstract: GPUs have been a topic of intense research for accelerating numerical processing, due to their high FLOPS/dollar and FLOPS/watt ratios. In particular, the field of numerical linear algebra has been a field of high payoff due to the applicability of the GPU and the widely useful nature of calculations such a system solutions and eigenproblem analysis. We will discuss our experience in this area in light of our CULA package for GPU accelerated linear algebra operations, which Temple University has leveraged in the creation of their PyCULA package.