high performance computing on graphics processing units: hgpu.org

Label-free optical bionsensors are known to be accurate and reliable tools for measuring and monitoring certain biomolecular interactions. In recent years, new techniques and technologies have emerged that enable high-throughput biosensing at lower system size, cost, and complexity. In particular, the LED-based Interferometric Reflectance Imaging Sensor (IRIS) has been demonstrated as a viable alternative to previously established high-end biosensors and its effectiveness can be augmented by the calibrated fluorescence enhancement (CaFE) technique. While IRIS enables high-throughput detection, its movement towards low-cost and field portability has introduced new processing challenges. Since the nature of the processing lends itself well to data level parallelism, it is natural to explore single instruction, multiple data (SIMD) architectures to improve processing performance. NVIDIA’s CUDA technology enables parallel computing on compatible graphics processing units (GPUs) and is an ideal platform for exploiting data level parallelization. A CUDA implementation of the CaFE/IRIS model has been developed, yielding significant performance improvements over the current MATLAB implementation.