The problem with optimising ARM assembly code, which we do a lot of for critical routines, is it varies every time you do so. There are so many other influences, especially with multicore processors and memory access, that you really can spend huge amounts of time for little or no gain. For commercial applications this can make sense - processors are never quite fast enough and we don't overclock - but for home use I'd just stick to C/C++.
Statistics: Posted by MikeDB — Sat Dec 28, 2024 5:43 pm