We are using IPP to speed up image resizing. However, according to our test, AVX2 does not have any performance gains compared to either AVX or SSE4.2
the functions we are relying on is ippiResizeSuper_8u_C1R/ippiResizeSuper_8u_C3R and the other related functions. I did notice that, according to http://software.intel.com/en-us/articles/haswell-support-in-intel-ipp none of the resize functions appear to have haswell specific optimization.
can we expect this to change in future IPP versions? if so, will they be in next version/update of IPP? or AVX2 won't help for these resize functions?
what kind of performance improvement you think we could expect by going from SSE4.2 to AVX2?
none of the resize functions are listed in http://software.intel.com/en-us/articles/haswell-support-in-intel-ipp
is there somewhere else we can check to see whether some functions have AVX/AVX2 specific optimization? thanks
a performance improvement depends on many factors, the best way to check performance is to apply AVX/AVX2 optimized ippiResizeSuper function to your application.