Hi, > Just fixed it - thanks! > > > On Jun 10, 2015, at 10:31 AM, Nathan Hjelm <hje...@lanl.gov> wrote: > > > > Looks like a bug in the CM optimization that went into 1.10 yesterday. I > > will let the author know to fix this one.
Today I could build openmpi-v1.8.5-50-g793344d, openmpi-v1.10-dev-79-g841ec20, and openmpi-dev-1868-g12d3c9c. A simple matrix multiplication in my heterogeneous environment worked fine with all three packages. Thank you very much to everybody who helped to fix the problems in openmpi-v1.10-* and to identify the error in my script to build the upcoming openmpi-1.8.6 version. Kind regards Siegmar