10 May
2005
10 May
'05
5:05 p.m.
On May 10, 2005, at 10:41 AM, Cory Nelson wrote:
It might be a good idea to launch the benchmark in a seperate thread, then call something like:
double GetThreadTime(HANDLE t) { __int64 c, e, k, u; GetThreadTimes(t, (FILETIME*)&c, (FILETIME*)&e, (FILETIME*)&k, (FILETIME*)&u);
return u/10000000.0; // user-mode time in seconds }
to get a more precise time.
At 60x overhead, I don't think more precise timing is the problem :) Signals likely needs some kind of overhaul w.r.t. performance. The major problem now is the slot groups. Doug