The first thing you should do is to narrow down the area in your application where the largest performance difference is
.

If it is related to I/O then the root cause is either IPC overhead or driver performance.

If it is purely the number of CPU cycles burnt for computing the exact same thing then the question is more about the 
compiler than than the OS as such.
Were you compiling the windoze version yourself too or you were grabbing some binaries built by someone else?
It is possible that your windoze binaries are using hardware optimizations that are not enabled during the QNX build 
(SSE) or more efficient libraries (e.g., Intel Integrated Performance Primitives).

Whatever the specific reason, your first job is to narrow it down as much as possible.

--
Mate


On Nov 13, 2013, at 4:21 PM, robort smith <community-noreply@qnx.com>
 wrote:

> I have compiled opencv on QNX successfully,and it can work normally.  The problem is that it isn't efficient.  It cost
 more time for QNX to run the same function than windows.
> 
> Is there anything I can do about this problem? 
> 
> 
> 
> 
> _______________________________________________
> 
> General
> http://community.qnx.com/sf/go/post106729
> To cancel your subscription to this discussion, please e-mail general-multimedia-unsubscribe@community.qnx.com