Basic file parsing

• Read file into map of <string, vector<float>> holding 4098 floats
• find bounding min and max values for each column (after file load)
• Record times and report

Matrix multiplication compared to Boost library solution, single processor

• Take two matrices, multiply them together
• Transpose right hand matrix to reduce cache misses by making columns contiguous rows
• Improve on Boost library times significantly

Multiprocessing: Find Manhattan metric (L1 Norm) for each vector, compared to specified vector of floats (from homework 0 file). Find K-nearest neighbors of comparison vector based on distance.

• Read in homework 0 file as before: into map<string, vector>
• Block-partition input
• Send input blocks to processes
• Write process' k-nearest neighbors to shared memory
• After convergence, filter k-nearest neighbors from unified shared memory solution
• Return solution

Multithreading: Use boost threads to accomplish homework 2

Use OpenMPI to accomplish Manhattan distance problem

GPU programming: use CUDA to perform median filters of varying sizes (3x3, 7x7, 11x11, 15x15) on lena.pgm

• Load .pgm into buffer on CPU
• Load buffer into GPU
• Perform filter kernel function
• Load back onto cpu
• Compare to CPU solution for accuracy
• Compare to CPU time

GPU programming: use CUDA to perform Sobel Filter (edge detection) on lena.pgm