What you can see is that (at least for me) Julia takes about twice as long to solve eigenproblems than Matlab. was conducted using well-known general purpose software, Matlab 2018b. Input and output are through plain-text files written to disk.
Its unusably slow when adjusting options in the plot window. Matlab Code function timer = timer_eigen(n, matrix_size) I do not recommend the use of Matlab 2018b for visualization on OSX. opengl software is slower compared to opengl hardware (as now the Graphics card are much more powerful) and hence forcing MATLAB to use opengl hardware may.
To better understand this discrepancy between Julia and Matlab I timed the eigenproblem solver for several different matrix sizes.įunction timer_eigen(n::T,matrix_size::UnitRange Answers (1) By default MATLAB tries opengl hardware on the machine, and if some reason it detects some incompatibility with graphics drivers and it might have used the opengl software. perform79transform - biorthogonal 7/9 1D wavelet transform - performwavelettransformisotropic - multidimensional isotropic (i.e. performliftingtransformbyname - string based interface.
This is expected, however Julia runs much slower than Matlab. performliftingtransform / performliftingtransformslow - 1D wavelet transform via lifting (general interface). I tried to apply the tips in the performance section.Īnalyzing the profiler, I found that most time is spend on the eigenvalue problem (and also on LU decomposition for solving systems of equations). Setting the figure property Interruptible to on and the BusyAction to queue does not solve the problem - this is the default already. I noticed that my code in Matlab is running much (much) faster than in Julia. But in Matlab 6.5 to 2018b (most likely newer versions also) the motion of the mouse calls the callback already. SBFEM requires solving a lot of eigenvalue problems. I am working on implementing SBFEM in Julia.