Re: Unexpected optimizer behavior

[Ian Lance Taylor <iant at google dot com>]

Frank Neuhaus <dtag00@gmail.com> writes:

> Yes I agree that since resize is not inlined, the way pixels changes
> during the resize call is unknown to the optimizer operating on main.
> But _after_ this call to resize, all changes to pixels _have_ to be
> caused by main, and thus are known to the optimizer of the main
> function. That means it should be able to tell that pixels cannot
> possibly change after the execution of resize(*). And more
> specificially that it cannot possibly change during the execution of
> the loop.

You are looking for flow-sensitive alias analysis.  gcc gets better at
that with each release, but there is still a long way to go.  At least
the last time I looked closely at alias analysis, around gcc 4.3, if a
pointer escaped, it escaped.  There was no ability to track that a
pointer only escaped during part of a function.

This is the point where I say that gcc is free software, and
contributions are always welcome.

Ian