Optimizations in code

[marcin-wadolkowski]

I saw that in many places there is for example:

• n % 8 which can be replaced by (i & 7)
• n / 8 which can be replaced by (i >> 3)
• in for loops n++ can be replaced by ++n
  These are small changes but I assume that the code is written in C because of performance

[marco6]

Hi Marcin, thanks for the time you spend on reading our code and starting this conversation!

I think things like the one you are proposing are already easily taken care of by any modern C compiler when possible. See here for example. Even if it wasn't the case I would probably still vote for keeping the operation as-is as & 7 is IMHO much more cryptic than an explicit % 8.

Last, we would be introducing troubles with signed integers. Those tricks work only for unsigned numbers. Take for example the following code:

int main() {
    printf("%d %d\n", 2 >> 3, 2 / 8);
    printf("%d %d\n", -2 >> 3, -2 / 8);
    return 0;
}

This prints:

0 0
-1 0

Which is a bit weird, isn't it? :-)

Regarding the ++n vs n++ it think it is more something from C++? Where iterators need to return a copy during post-increment? For things like "normal" register integer there shouldn't be any difference I think.