Ulrich Schöbel wrote: > Hi Richard, > > thanks for trying to reproduce my 'problem'. > > I'm using Linux 2.6.12 (Ubuntu Breezy Badger, a debian distro), > Tcl/Tk 8.4.12, sqlite 3.3.4, all pretty recent versions. > > I made the same tests today with the same results. > > Nevertheless, sqlite is by far faster than mySQL, so I'm going > to switch to it, anyway. But I'm still curious, where this effect > is coming from. > >
Could you (or someone) write the test in 'C' and compile it with "-ggprof" flag on gcc and run it in the profiler? I propose that the test program should take one command line argument, the # of iterations. That way you can profile "1" vs "50" vs "100000" independently of each other.
