Aaron Bannert wrote: > On Thu, Feb 14, 2002 at 03:19:54PM -0800, Ian Holsman wrote: > [...] > >>same configuration/same source code/same docroot. >> >>by my calcs v31 now does 14 mutex locks per read >>as opposed to v29 5 locks per read >> >> >>I'm thinking that the shared memory changes have made my machine >>into a furnace. >> >>BTW when I ran the same kind of test on 20-dec with 30-dev I was showing >>a 10% speed increase.. now it looks like a 30% slowdown. >> > > Shared memory changes? There are no explicit locks in apr_shm_t, and > only during "allocation" for apr_rmm_t, AFAIK. Are you referring to > something else? > > -aaron > > found a config difference. v29 had the following: export LD_PRELOAD=/usr/lib/libmtmalloc.so
with this line v31 now outperforms v29. thanks for the help guys..
