Bisqwit, can you run your filter against this ROM:viewtopic.php?f=2&t=15080
I just noticed that when I talked in the other thread (linked above) about the QuickBASIC NTSC codec, which is faster in performance than my previous C++ NTSC codec, I had totally forgotten that I wrote this C++ version of the faster algorithm, too.
Yes, the one in this thread should be the fastest I have. If there are still performance complaints, … well, too bad. First, make sure your compiler inlines the lambda functions though. It is pretty important.