Running once before will cause compilation of the called functions for the given types.
julia> A = randn(1000,1000); A = A+A'+eye(1000); x = randn(1000); julia> @time A\x; @time A\x; elapsed time: 0.752317463 seconds (48614156 bytes allocated) julia> @time A\x; @time A\x; elapsed time: 0.031127576 seconds (8016424 bytes allocated) Any subsequent calls to A\x with different random initializations will have performance like the second call shown here. On Friday, 17 January 2014 13:21:21 UTC-6, Andrea Vigliotti wrote: > > Hi All, > > I was comparing the performance of Julia and Matlab in solving systems of > linear equations and I got the following > > Matlab: > >> A = randn(1000); A = A+A'+eye(1000); x = randn(1000,1); > >> tic; A\x; toc > Elapsed time is 0.034763 seconds. > > Julia: > julia> A = randn(1000,1000); A = A+A'+eye(1000); x = randn(1000); > > julia> @time A\x; > elapsed time: 0.192572124 seconds (8012424 bytes allocated) > > I was wondering whether there is anything I can do to improve the > performances of Julia in solving this kind of problems? > > many thanks! > andrea >
