This is a nice paper. I think that for matrix multiplication it ends up saying pretty much the same as we've been discussing.
The OP said "serial" code. Vector code isn't serial. However it's easy to try vectorization these days. The latest versions of gcc will do a very reasonable job with the right optimization flags. However be aware that if you create a binary with vector code for one machine it can easily fail on a different one. Unlike the basic x86 instruction set, there are several levels and branches of "standards." On Mar 6, 1:47 am, Sairam Ravu <ravu...@gmail.com> wrote: > Here is the nice link for writing fast matrix-matrix > multiplication.http://spiral.ece.cmu.edu:8080/pub-spiral/abstract.jsp?id=100 > > Apart from this we can vectorize the code and also we can do > unrolling to get very good performance. > > -- > With love and regards, > Sairam Ravu > I M.Tech(CS) > Sri Sathya Sai Institute of Higher Learning > "To live life, you must think it, measure it, experiment with it, > dance it, paint it, draw it, and calculate it" -- You received this message because you are subscribed to the Google Groups "Algorithm Geeks" group. To post to this group, send email to algogeeks@googlegroups.com. To unsubscribe from this group, send email to algogeeks+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/algogeeks?hl=en.
