Sturla Molden <sturla.mol...@gmail.com> wrote: > Just a tiny comment on this: It might be easier for the LAPACK library to > use SIMD extensions (SSE2, SSE3, AVX) if we use DTRTRS. So we need to > benchmark with relevant matrix sizes to see which strategy is the better. > One is better for SIMD, the other is better for cache.
Actually, I can think of two other methods as well. One would be to loop over DGEMV, the other would be to use DGEMM. And of course, having the data transposed (struct of arrays instead of array of structs) would give a different set of methods. I have no idea which will be the faster. Sturla ------------------------------------------------------------------------------ Flow-based real-time traffic analytics software. Cisco certified tool. Monitor traffic, SLAs, QoS, Medianet, WAAS etc. with NetFlow Analyzer Customize your own dashboards, set traffic alerts and generate reports. Network behavioral analysis & security monitoring. All-in-one tool. http://pubads.g.doubleclick.net/gampad/clk?id=126839071&iu=/4140/ostg.clktrk _______________________________________________ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general