http://bugs.winehq.com/show_bug.cgi?id=1667
------- Additional Comments From Andrew.Talbot@talbotville.com 2003-07-12 06:03 ------- Mike,
After some research, I would offer the following assertions, which may need verification:-
1. On systems that do not have the RDTSC instruction, the High Performance Counter is fed from a chip at 1.193182 MHz. So, maybe the fall-through value for frequency->s.LowPart of 1000000 might more accurately be set to 1193182;
2. On systems that do have the RDTSC instruction, the values that QueryPerformanceCounter() and QueryPerformanceFrequency() return should be individual clock-cycles and Hz, respectively; i.e. the divisions by 1000 would seem to be wrong. My supporting evidences for this assertion are:-
a) see http://www.devx.com/SummitDays/Article/16293/1763; b) RP8 works well on my system if I remove the scaling-down of counter->QuadPart and frequency->QuadPart from /dlls/kernel/cpu.c.
-- Andy.