Hello,<br><br>I'm building a test app via makefile on osx that uses mpich2 (which I compiled from source and whose binaries end up in /usr/local/bin).<br><br>It is a trivial mpi app and builds via this trivial makefile:<br>
<br>test : main.cpp<br> mpicxx -g main.cpp -o testmpi<br><br>From a terminal window, I can make and execute it with the expected results, with one or multiple processes.<br><br>--<br><br>If I create an "External Build System" project in Xcode and point to this makefile and try to build the target, XCode tries to use the mpicxx that is located in /Developer/usr/bin despite the fact that from a commad prompt, "which mpicxx" reports /usr/local/bin/mpicxx. <br>
<br>So one can go to /Developer/usr/bin and mv all the mpi* binaries into mpi_hidden or some such, at which point Xcode will complain that it can't find mpicxx. So one can create a link from /Developer/usr/bin that references the mpich2 binary /usr/local/bin/mpicxx. Now things will build and work as desired.<br>
<br>My question, then... What is the recommended practice to get Xcode to use the mpicxx I want without all this problematic moving around/hiding of binaries? One option would be to cleanly remove the openmpi that comes with osx; a better option would be a way to spec the path that Xcode will search for binaries, but I don't know how to do either.<br>
<br>thanks,<br>thomas blom,<br>ut austin<br>