Say i have a 1920 x 1080 image the size of a pixel could be a custom element. e.g. 8 * sizeof(float). I need to do some scanline operations on this pitched memory. My question is how do I do vertical scanning on a pitched memory more efficiently. Obviously horizontal scan is easy.
Ways that i can think of are:
- bind pitched memory to a texture memory.
- write a rotate kernel for pitched memory and do the horizontal scan
So what is the textbook way of doing this.