[pypy-issue] [issue900] Using cProfile on Windows generates negative tottime values

[Fijal]

Fijal <fijall at gmail.com> added the comment:

QueryPerformanceCounter is definitely not the best. Right now we directly embed 
RTLD instruction in the assembler, which is much faster than have a call there. 
Inlined function calls can be very cheap on pypy - down to few assembler 
instructions. In this case having the overhead of two calls (for start and end) 
is not acceptable. It seems that affinity setting does not work, the x86 docs 
clearly state what RTLD will return and it has to be a monotonic counter for 
that processor.

Cheers,
fijal

________________________________________
PyPy bug tracker <tracker at bugs.pypy.org>
<https://bugs.pypy.org/issue900>
________________________________________