Acceleration of High Dimensional Matrix Multiplication Processes With Cache Usage Optimization and Parallel Programming Techniques

Optimization of cache on the matrix operations according to calculation times, which are commonly used in many field such as image processing, robotics, econometrics, civil engineering and quantum physics were investigated. The calculation performance is made 14 times faster than using the different algorithms instead of the known matrix multiplication, by using the temporal and spatial locality principles in the study. Effect of the cache which is very important for computer performance, has been tried to detect. Processing times have significantly shortened thanks to cache utilization optimization methods, also it has been aimed that extending the lifetime of the system by preventing use the cache and other processing units more. After cache optimization, it has been aimed to shorten the computation times of intensive matrix operations using parallel programming techniques. Thus, both the cache was used more efficiently and the most effective computations were tried to be performed using the appropriate data parallelism. Thanks to parallel programming techniques applied with 5 computers after cache optimization, the calculation process is made 59 times faster than the commonly used technique. Speedup values are calculated according to the different number of computers in the calculations made by parallel programming. In addition to computer-based fast computation methods for matrix operations, also we have focused on algorithms that have a negative effect on computation