Hi, since getting my hands on a 8-core Mac Pro, I’ve had problems with running simulations with openMP enabled.
I just discovered that all simulations run (and bake) very slowly with an openMP enabled build. Without openMP the simulations use exactly one core at 100% and run many times faster.
The same happens if using an openMP enabled build and feeding it the environment variable OMP_NUM_THREADS with a value of “1”. If the value is increased, even only just to “2”, all simulations slow down terribly.
This was discussed with JesterKing, who found some threads about gcc 4.2 not working well with openMP. I tried gcc4.4 and gcc4.5, but couldn’t get Blender to compile. The thing I find strange is that openMP seems to work on my dualcore MacBook Pro, with all the same software.
Thanks a lot for any advice or clues!