From Lisandro (author of mpi4py). Sounds like at the very least we should be shipping a newer mpi4py.
Begin forwarded message: > From: Lisandro Dalcin <dalc...@gmail.com> > Date: December 9, 2009 8:36:03 PM PST > To: Robert Bradshaw <rober...@math.washington.edu> > Subject: Re: [sage-support] Re: Running Sage on a 64-bit Scientific > Linux machine > > Gus seems to be a bit confused about thread-based parallelism (and GIL > issues we all know) versus process-based parallelism and mpi4py (with > no issues at all, but it is (much?) harder programming model because > there is no shared data but explicit messaging between processes). > > Moreover, the failures he comment are likely related to a completely > outdated mpi4py version (way back the time mpi4py switched from > hand-written C to Cython) currently available on Sage. mpi4py uses a > carefully monkeypatched (subclass-based) standard distutils for > building, so perhaps updating mpi4py in Sage would be a rather easy > task? > > Additionally, having 8 cores does not mean you are going to be able to > run 8 times faster, even in the case of trivial parallelism. In my > particular apps (solving PDE's with Krylov-based iterative methods > involving sparse matrix-vector products) I can barely get a 2.5X > speedup using 8 cores. So in this case mpi4py could be seen as a > nonsense... Unless you are a parallel distributed application > developer using mpi4py on you desktop box for dev&test purposes, and > finally move your scripts to a big cluter to run a 1000 times bigger > problem in 2000 or more processors. In this case, having mpi4py > installed even in a old, single-core machine do makes a lot of sense > (this is what I do, I develop on a old Pentium 4 32bits single core > with just 1GB RAM... but use my cluster for production runs) > > > -- > Lisandro Dalcín > --------------- > Centro Internacional de Métodos Computacionales en Ingeniería (CIMEC) > Instituto de Desarrollo Tecnológico para la Industria Química (INTEC) > Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET) > PTLC - Güemes 3450, (3000) Santa Fe, Argentina > Tel/Fax: +54-(0)342-451.1594 -- To post to this group, send email to sage-support@googlegroups.com To unsubscribe from this group, send email to sage-support+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-support URL: http://www.sagemath.org
