Revision history [back]

I did some testing, and the flips are significantly faster than the affine for my computer, and they use less CPU (only one thread) while running.

Here are the codes, not as a function, which I'll put together and submit later. Clockwise is positive.

Rotate 90 (2.66 seconds/1000 iterations for flip vs 5.59/1000 for affine)

 flip(input.t(), output, 1);

Rotate 180 (If the matrix is continuous, it may be better to do std::reverse, but I have yet to test this)

(1.18/1000 for flip, 2.53/1000 for affine)

flip(input, output, 1);
flip(output, output, 0);

Rotate 270 (2.19/1000 for flip, 5.90/1000 for affine)

flip(input.t(), output, 0);