When executing code with sse intrinsics from a posix thread the program will crash. I believe there is a byte alignment issue with glibc or gcc where it's only aligning 8 bytes instead of 16, which is what sse wants. I've compiled the sse multi-threaded code with the intel compiler and it works fine. My sse code not using pthreads also works fine under gcc.
This report also has detailed data on the problem: http://lists.debian.org/debian-gcc/2003/debian-gcc-200307/msg00200.html
This is a dup of bug 10395 but really it needs something like what is in bug 6737 which is request for stack realignment attribute. If I would have it also fixed in glibc. The reason why the intel compiler works as it realigns and does not think the stack is align ever. The reason why redhat works is that it has a "fixed" glibc. *** This bug has been marked as a duplicate of 10395 ***