You don't need the system time in microseconds. What you need is the O() function that describes what is really going on. Does the function vary with N, NlogN, N^2, or maybe N^3? That is the question you need to be asking most. Sure some function calls are more expensive than others, but the same rule applies to them also. I don't have the code in front of me at the moment (it would require rebooting into LINUX, since I'm in Winblows right now), but I can tell you without even looking that something in there must be operating with O() >= N^2 in places. Those are the places which need to be examined the most. That is the point behind true RDBMS theory, as well as basic code optimization in C/C++ and SQL query optimization.
Drew Northup, N1XIM > -----Original Message----- > From: Mikhail Ramendik [mailto:[EMAIL PROTECTED] > Sent: Tuesday, October 26, 2004 2:15 PM > To: DBMAIL Developers Mailinglist > Subject: [Dbmail-dev] Help me, I don't speak C... > > > Hello, > > Well, this is really not a dbmail question, but as I need it to try and > speed up dbmail, I hope you'll bear with me. > > I am trying to understand why _ic_fetch() is slow. For this I need log > entries with millisecond, not second, precision. Apparently syslogd > can't do this, so I tried to add the following to the trace function in > debug.c : > > struct timeval _tv; > struct timezone _tz; > > gettimeofday(&_tv,&_tz); > vsyslog(LOG_NOTICE,"Microseconds: %d", _tv.tv_usec); > > (Of course I also added #include<sys/time.h> in the beginning). > > This causes a segfault. Why? And what is the right way of getting any > system time value in milli-or microseconds? (I don't care if it wraps > every second or not). > > Yours, Mikhail Ramendik > > > > > >
