This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Add more conditions to fold_build_cleanup_point_expr


Andrew Pinski <pinskia@physics.uc.edu> writes:

> I found that the C++ front-end would create a cleanup point expr on a
> return expr even if the return expression is known not to throw.
> This adds to fold_build_cleanup_point_expr information on when
> return expressions really need a cleanup point expression.
> This reduces amount of statements allocated for ir.ii (PR 13776)
> by more than 5x (412,936 down to 81,816).  I did not measure a speed
> increase for this testcase but that could be because I think the
> collector only ran once on my machine which has 1G of memory.
>
> OK? Bootstrapped and tested on ppc-darwin with no regressions.

OK provided that you add commentary explaining what the two nodes are
that you are checking TREE_SIDE_EFFECTS on.

zw


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]