Hi, On Wednesday 24 June 2009 09:53:13 James Turner wrote: > An observation: this change has stopped SGAtomic being (by default) a > proxy for osg/OpenThreads Atomic. There's a downside to this - > OpenThreads Atomic has a specialisation (not of the template kind, but > the #ifdef kind) for OS-X atomics - > _OPENTHREADS_ATOMIC_USE_BSD_ATOMIC. On Mac we're still on GCC version > 4.0 by default [1], so we don't get the GCC built-in case for SGAtomic > and friends - and will fall back to a pthread-mutex implementation. > > I know there's an argument for having Simgear not depend on OSG, which > I agree with, and I know it's awkward to depend on OpenThreads when > it's typically only distributed as part of the OSG tree, it just seems > like a step backwards - I'd prefer to delegate all thread support to > OpenThreads and maintain less code ourselves. > > [1] - yes, it's ridiculous that the stock compiler is still 4.0. A > newer version (4.2) is available, but I'm hesitant to switch to it > when the system compiler is still 4.0.
Yes, for such a compiler we just fall back to an anway safe mutex implementation. I did not port the changes from osg referenced back into simgear. Note that original code for simgears atomic as well as osg's atomic stems from some simulations code I wrote. Using osg stuff here will also pull at least openthreads into that parts of simgear that have nothing todo with scenery and osg and such. But if you need that BSD stuff please port that to simgear! Sorry for the inconvenience. But anyway, is it notably slower then for you? Just for curiosity? And yes we have currently gcc 4.4 with vast of improvements. Greetings Mathias ------------------------------------------------------------------------------ _______________________________________________ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
