If you want it to look nice and are running on 0.4, just switching to slice(A, 1:n, k) ↔ slice(A, 1:n, λ)
should also get you a performance boost (especially for large matrices). Obviously you could do even better by devectorizing, but it wouldn't be as pretty. Off-topic, but your use of unicode for this is very elegant, and eye-opening for me. Best, --Tim On Wednesday, March 25, 2015 09:24:09 AM Matt Bauman wrote: > The swap could be done without temporaries, but I assume you're also trying > to match the look of the pseudocode? > > On Wednesday, March 25, 2015 at 11:22:41 AM UTC-4, Jiahao Chen wrote: > > Here is some code I wrote for completely pivoted LU factorizations. > > Can you make it even faster? > > > > Anyone who can demonstrate verifiable speedups (or find bugs relative > > to the textbook description) while sticking to pure Julia code wins an > > acknowledgment in an upcoming paper I'm writing about Julia, and a > > small token of my appreciation with no cash value. :) > > > > Reference: G. H. Golub and C. F. Van Loan, Matrix Computations 4/e, > > Algorithm 3.4.3, p. 132. > > > > Thanks, > > > > Jiahao Chen > > Staff Research Scientist > > MIT Computer Science and Artificial Intelligence Laboratory