- Joined
- 12/31/13
- Messages
- 17
- Points
- 13
I just used LU.X = B and solved for X using forward and backward substitution. Found it to be faster than Gaussian Elimination. Implemented using XLW library in c++ to compute in Excel... whereas Excel took 0.28s to compute a the inverse of a 201x201 matrix, single-threaded gaussian elimination took me 14.97s, and LU (using Doolittle - also single-threaded) 8.4s. My overhead against Excel is of course the use of XLFOPERS and MyMatrix class to store the matrices in RAM - these are too slow - I've not yet tried converting to just std::vector <std::vector <double>>.