lu-factorization-writeup

A blas level 3 implementation has been done for the LU factorization . I have tested it for various sizes as inputs to the matmat_ function. The value for the resid parameter will increase as the matrix size increases when we use the checkLU method to reconstruct the entire matrix. 

The performance of lu factorization does depend on the block size and may vary on multiple machines and depends on multiple factors.

Before starting coding of the assignment, i discussed the matlab scripts for the assignment, fortran code, pivoting logic and mkl reference documentation with Shreya and Shambhavi. 

i have tested my code on silo for matrix sizes higher than 1000 and got the performance results. Performance timings do vary based on the block size and the matrix size. I have included lumatmat.c, lumatvec.c and make.inc as they are the files I have changed. to run it on silo, i did a module load intel and gcc.