(oops, I meant to send that to the matplotlib list) Hi,
I was looking at the transform code recently.. On Thu, July 19, 2007 7:31 pm, John Hunter wrote: > The potential cost is not in the 3x3 vs 4x4, but in the extra row of > junk data you would store in the data matrix, which is N extra values for > plotting N points . The matrix multiplication would be 3x3 * 3xN vs 4x4 * > 4xN , so there would be a cost in memory and performance. I'm not so clear about what you are planning for the transforms and matrices in mpl1, especially in relation to the 4x4 to 3x3 matrices. Couldn't you just pass around 4x4 matrices, but then truncate them to 3x3 right before you apply them? If you are passing any affine transforms to the backend, you are going to be breaking apart your matrix anyway. (agg accepts the affine transform like tuple a,b,c,d,tx,ty) Also, my impression is that the matrix multiplication strategy in numpy is going to be slow if it happens a lot. I am guessing what you are going to do is do a matrix mult once just for the nonlinear transform when show() is called, but it will not happen for redraws (due to panning etc). When panning, only the affine part is changed, and the backend takes care of that efficiently (in C, for agg). Therefore the matrix mult is very rare. Is that correct? Allan ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Matplotlib-devel mailing list Matplotlib-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
