On Mon, Aug 11, 2008 at 8:44 AM, Matthieu Brucher <[EMAIL PROTECTED]> wrote: > > Hi David, > > I recall you propsoed a plugin framework for numpy. Where are you on > that matter ? Still waiting for numscons ?
The problem is not so much the build part, but the clear separation I was talking about. My experience with ATLAS convinced me the only way to make sse work reliably is to detect the CPU arch at runtime; compiling binaries incompatible on different arch is just not scalable and confuse users. Technically, to make this work, you need a clear different in the codebase between the code which calls the C functions changed at runtime, and the ones which are called by python. Hopefully, I will have more to say within the end of the week cheers, David _______________________________________________ Numpy-discussion mailing list Numpy-discussion@scipy.org http://projects.scipy.org/mailman/listinfo/numpy-discussion
