Apply __attribute__((const)) where applicable

[albinahlback]

I've seen functions which does something like

for (j = 0; j < n_pow(g, n); j++)

Assuming that g and n does not change, GCC cannot optimize this as it doesn't know if the output will be different each time. Pushing __attribute__((const)) will make it understand it.

This should be pushed onto every function where it applies.

See GCC's documentation for more information.

[math-gout]

Is there a reason for not doing m = n_pow(g, n) and then for (j = 0; j < m; j++) ?

[albinahlback]

It is cheap compared to searching for all such instances as they don't necessarily have to be contained within for-loops. It can also lead to more readable code.

[albinahlback]

Personally, I am a big fan of attributes since that hints the compiler that it can assume even more things. Of course, one should still try to optimize the code as much as possible, as this is only valid for GCC-compatible compilers.

[fredrik-johansson]

I don't think there are too many of these.

One of the drawbacks of this attribute is that the compiler can start to optimize out benchmarked function calls from benchmarking loops :-)

[albinahlback]

That's true.