1 Tested on Windows XP SP2 (X86)
2 SMP - tested (AMD X2 - i.e. 2 CPUs)
3 Kernel accounting _IS_ inaccurate
5 winhog (gcc -o winhog.exe winhog.c -lwinmm) can be used to demonstrate
8 http://blog.kalmbachnet.de/?postid=28
10 Apparently RMClock (http://cpu.rightmark.org/products/rmclock.shtml)
11 either somehow sets CR4.PCE or hooks to the #GP(0) handler so that
12 RDPMC works from user-space.
14 Therefore if RMClock is running, APC can invoked with `-M' switch and
15 by sampling TSC/PMC0 one can estimate the idleness of the system.
17 No idea how this behaves in presence of frequent (no pun intended)
18 CPU frequency changes.
20 As it is now ml_apc.c must be compiled with the compiler supporting
21 Microsoft style inline assembly, in other words compilation will most
22 likely fail if trying to build with Mingw or Cygwin based OCaml
26 CLOCKS IN UNHALTED STATE
27 Idle factor = 1 - ------------------------
30 To get rid of console (and by extension ability to read error/help
31 messages) one can unREM the last line of build.bat before building.