Comments by "Vitaly L" (@vitalyl1327) on "CPU Faster Than You Think | Prime Reacts" video.

  1. Transposing is just a simplest thing you can do to improve cache access pattern. You can do a lot more complex scrambling-descrambling to optimise caches, or GPU local memory. E.g., for a lot of image processing algorithms, representing the image planes as Z-curves (or other space-filling curves) result in a much better locality.
    1