Hi, i was using some c++11 concurrency classes on ARM (with gcc 4.9.2) that have an underlying use of libgcc atomic functions, for example __gnu_cxx::__atomic_add_dispatch. In its implementation, (in the file gcc-sources/libstdc++-v3/include/ext/atomicity.h) I can see that the macro _GLIBCXX_ATOMIC_BUILTINS is not defined, and therefore the builtin implementation of __atomic_fetch_add is not used. Instead, the mutex implementation is used which makes it significantly slower. As far as I could see, if I explicitly call one of these atomic built-in functions they work ok and the correct atomic assembly is emitted (using ARM's ldrex and strex instructions). So, the question is why this _GLIBCXX_ATOMIC_BUILTINS macro is disabled. is there a reason why we aren't using the atomic builtins, or I need to enable it. How can we do the enabling? Is the implementation incomplete? Thanks
-- ______________________________ <http://www.tallertechnologies.com> Marcos Díaz Software Engineer San Lorenzo 47, 3rd Floor, Office 5 Córdoba, Argentina Phone: +54 351 4217888 / +54 351 4218211/ +54 351 7617452 Skype: markdiaz22
_______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel
