On Mon, 2006-08-28 at 18:07 -0400, Peter Tanski wrote: > I do know that the Mach kernel implementation of pthreads does not > have shared barriers.
There is a problem with the PPC, it has only weak memory ordering and no workaround other than a heavy cache flush. I think we observed this, running a multithreading test on a dual core G5, when using 2 CPU's performance dropped by a factor of 18. Yup. 18 times slower. x86 family have stronger ordering. So perhaps the issue is not so much that Mach kernel doesn't have the barriers .. but simply that they cannot be provided on the PPC? [I'm no expert in any of this.. :] > Do you think you could use Scons? You best ask Erick that .. but I very much doubt it. Why bother anyhow? Our system is more considerably more advanced and it works (more or less). The big thing we're missing is network building. We can do cross-compilations, but only on shared file systems. EG: Cygwin/win32 or Win32/Win64 is OK, but if you built with MingW-gcc under Linux on a dual boot machine without a shared partition it would be hard :) -- John Skaller <skaller at users dot sf dot net> Felix, successor to C++: http://felix.sf.net ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Felix-language mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/felix-language
