GCC appears to be paranoid about cache misses, which of course was an issue when we had 4k caches. But now they are huge it really could be a little less worried.I'm not really a programmer, I consider myself a lamegrammerI am including loop unrolling here.but somewhere on Stackoverflow, they said at least GCC is bad at loops unrolling and it doesn't do it even if specifically told to do so. For example, manual unrolling gave me more performance in some repetitive data processing. Of course at the cost of higher size. But for example full unroll, while it produced huge binary, but it wasn't the fastest one.
but somewhere on Stackoverflow, they said at least GCC is bad at loops unrolling and it doesn't do it even if specifically told to do so. For example, manual unrolling gave me more performance in some repetitive data processing. Of course at the cost of higher size. But for example full unroll, while it produced huge binary, but it wasn't the fastest one.Statistics: Posted by MikeDB — Sun Oct 13, 2024 9:50 am