Marius Tolzmann wrote: > There seems to be another problem with gcc 4.7.1.. We had some serious > issues with programs crashing predictable on one machine and never on > others with identical software/kernel. > > It turned out that programs compiled with gcc 4.7.1 that are run on a > processor with FMA4 extension segfault, if they entered fma4-optimized > sections of the glibc. After recompiling glibc (same version) with 4.7.2 > everything was fine.. so I actually never had a second look at what the > problem really was nor did I do any more research in any bug databases. > So I am not sure where this bug was fixed and what the actual problem > was. But it actually broke emacs and some bio-software on cpus with the > FMA4 extension. > > (After debugging and disassembling the error was not really in the FMA4 > instruction itself but in the code gcc created after calling the fma4 > instructions..) > > gcc 4.7.1 seems to be fishy 8)
Can you check the patch at http://gcc.gnu.org/ml/gcc-patches/2012-08/msg00421.html or possibly build with -mno-fma4 Another patch to try: http://gcc.gnu.org/ml/gcc-patches/2012-08/txt00060.txt Or http://gcc.gnu.org/ml/gcc-patches/2012-08/txt00062.txt The discussion starts at http://gcc.gnu.org/ml/gcc-patches/2012-08/msg00421.html Checking my systems, none have fma* in /proc/cpuinfo so I can't test it. -- Bruce -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
