I began using IPP for image processing recently. When I use the function:
"ippiSub_16s_C1RSfs" instead of pixel-by pixel subtraction it works very slow - more than 2ms for a 352X272 pixels image.
The strange thing is that if I decrease the number of threads from 8 to 1 the software runs faster - less than 1ms. I like it to run as fast as possible...
CPU: i7 CPU Q 720 @1.60GHz 1.60GHz
Memory: 4.00 GB (2.94 GB usable)
OS: 32-bit, windows 7
Thanks in advance,
amount of data to process is quite small in your particular case, that might be a reason for slowdown (threading overhead is bigger than actual processing time).
You may also need to chech you call ippInit function (to initialize IPP dispatcher) in case you link with static libraries.