> "Ivica Bukvic" <[EMAIL PROTECTED]> writes: > > Anyone care to comment or (please) dissuade me from potentially > making a costly mistake? ;-)
I don't think it needs to be either/or -- I run sfront under Linux at work doing real-time things, and I run sfront at home using OS X doing real-time things, and both more or less work OK. Personally, I think the biggest issues are the different approaches OS X and Linux takes on real-time scheduling (constraint scheduling for OS X, SCHED_FIFO and friends for Linux), and on the need and methods for keeping the application in RAM to avoid swapping (the sfront OS X port took at lot of work to get this right, and in some ways sfront has a simple memory usage pattern ...). I think Linux may have something to learn from OS X on the former issue, and OS X may have something to learn from Linux on the latter issue -- at the moment at least, the "lazy swap-in" in OS X, even when the machine has hundreds of MBytes sitting unused, forces the application to do its own implicit memory management to avoid artifacts during start up. ------------------------------------------------------------------------- John Lazzaro -- Research Specialist -- CS Division -- EECS -- UC Berkeley lazzaro [at] cs [dot] berkeley [dot] edu www.cs.berkeley.edu/~lazzaro -------------------------------------------------------------------------