The Help Conquer Cancer (HCC) project researchers have published a paper describing their efforts in re-implementing their algorithms to take advantage of Graphics Processing Unit (GPU) hardware present in many of today's computer's. If the algorithms lend themselves to use a GPU implementation, this can dramatically increase their performance. Their work has led to the testing of a beta release of HCC using GPU hardware on World Community Grid. This paper discusses their experiences.
"High-throughput protein crystallization on the World Community Grid and the GPU"
We have developed CPU and GPU versions of an automated image analysis and classification system for protein crystallization trial images from the Hauptman Woodward Institute's High-Throughput Screening lab. The analysis step computes 12,375 numerical features per image. Using these features, we have trained a classifier that distinguishes 11 different crystallization outcomes, recognizing 80% of all crystals, 94% of clear drops, 94% of precipitates. The computing requirements for this analysis system are large. The complete HWI archive of 120 million images is being processed by the donated CPU cycles on World Community Grid, with a GPU phase launching in early 2012. The main computational burden of the analysis is the measure of textural (GLCM) features within the image at multiple neighbourhoods, distances, and at multiple greyscale intensity resolutions. CPU runtime averages 4,092 seconds (single threaded) on an Intel Xeon, but only 65 seconds on an NVIDIA Tesla C2050. We report on the process of adapting the C++ code to OpenCL, optimized for multiple platforms.
Access to Paper:
To view the paper, please click here.