Ihar Hubchyk
Ihar Hubchyk
Hi, @MeKot , did you have free time to have a look into this? :)
@0x72D0 you might have an interest in this ;) There are 3 or 4 places only to cover.
My first assumption is regarding to the code for loading final data into variable. We cannot assume that std::vector is aligned correctly but we could makes checks in the function...
You could see that every function uses unaligned data load and save which is a big drawback in performance.
I mean that end user (somebody else) would not care about this thing while calling the function.
No-no. I'm talking about general solution. We use std::vector only for one function while we have many of them :) We should concentrate our focus of image data load and...
No, it's a raw allocated array. Additionally, we handle ROI (region of interest) which could be smaller than original image size. Even if you allocate an image with correct alignment...
Yes, it is. Another idea which comes into my mind is that we can create a template function with 2 typenames. One typename parameter is load function for input data,...
Let's assume we create a template function: ```cpp template simd loadSSE(simd * data); ``` Then we could have a specification of this template as: ```cpp template simd loadSSE(simd * data)...
Yes, we need to check. Memory allocator is an optional feature.