Changes

; Optimize uses of __cosf and __sinf. : Since using the device functions made such a big change and since there were two identical calls to each, I stored the value of their result in a register and used the register twice. This didn't make such a big change but I am sure that if I had done it before using __cosf and __sinf, the difference would have been much bigger. On the smallest image, the tiem went from '''93μs to 91μs''' and on the largest image, the time went from '''1232μs to 1196μs'''.

; Experimental division optimization. : This last optimization made almost no difference but I more wanted to try something new. I changed "'''int col = index / rows;'''" to "'''int col = (index - row) / rows;'''". The idea is that if the computer gets a whole number from division, it would be a bit faster. I don't fully understand why it worked but there was a consistent speed up of around 0.5% to 1%. On the smallest image, the time went from '''91μs to 90μs''' and on the largest image, the time went from '''1196μs to 1191μs'''.

==== Unoptimized ====