Re: Slow OS calls on one chip type

Art Kagel <art.kagel@gmail.com> wrote: > The BIG problem are the 6.7MILLION failed calls to the OS close() > system call! they propably come from a post fork/clone loop that is clearing the file descriptors before a new shell is being called using execve (something like for (i=3; i<65536; i++) close(i)) each '!echo "something"' line produces extra 65536-3 failed (EBADF) 'close' calls 'close' on non-valid file desciptors can be easily benchmarked on both systems just to be sure that it is not a kernel issue also the strace can show what happens between execve and write(1,..) call (from "echo") on both systems (like scanning NFS directories that are on the PATH or LD_LIBRARY_PATH or something like that) -- Best Regards butthead