The Hyperspectral Computing Laboratory (HyperComp) has been doing research on remotely sensed hyperspectral image analysis for several years now. The most important topics covered include spectral unmixing of hyperspectral data, classification and high performance computing. Unmixing is important to understand hyperspectral data at a sub-pixel level for low spatial resolution data, by identifying a set of pure spectral components (called endmembers) and their corresponding fractional abundances in each (possibly mixed) pixel of the hyperspectral scene. For high spatial resolution data, classification assigns a unique label to each pixel in the scene. In both cases, the very large dimensionality and complexity of hyperspectral scenes add challenges that can only be tackled by exploiting high performance computing architectures such as commodity graphics processing units (GPUs) or field programmable gate arrays (FPGAs) for onboard processing. This talk will present the most important advances developed at HyperComp in the aforementioned research fields, which comprise many important contributions to the area of remotely sensed hyperspectral presented in the talk.