Michael Hudson wrote: > Could I persuade you to try again --with-pymalloc? I did. It was only slightly faster (1-2 percent maybe). AFAIK, glibc uses Doug Lea's malloc which is pretty fast. Neil