https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #18 from Eric Gallager ---
(In reply to linzj from comment #16)
> Created attachment 43783 [details]
> This patch trying simulate a local escaped node for function call.
>
> This patch trying simulate a local escaped node for
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #17 from linzj ---
After this patch the code still have redundant calls for the inline pass has
finished. But the indirect function calls has been removed.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #16 from linzj ---
Created attachment 43783
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=43783=edit
This patch trying simulate a local escaped node for function call.
This patch trying simulate a local escaped node for
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #15 from linzj ---
The structalias solve_graph is not resolving the nodes reachable from a source
node, which make my try fail.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #14 from linzj ---
I mean call_may_clobber_ref_p_1 function in tree-ssa-alias.c. It does use the
clobber set of the call statement. So making the clobber set right should have
positive effect on the code, right?
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #13 from linzj ---
But the code looks like only using the clobber set of the call statment. I will
look into that.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #12 from rguenther at suse dot de ---
On Fri, 16 Mar 2018, manjian2006 at gmail dot com wrote:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
>
> --- Comment #11 from linzj ---
> This patch makes the printf clobber/use set
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #11 from linzj ---
This patch makes the printf clobber/use set correct. But it doesn't seem like
much help.
Maybe the escape set only comes from the function wise one.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #10 from Richard Biener ---
(In reply to linzj from comment #8)
> Created attachment 43675 [details]
> patch to discuss
>
> This patch aims to make the escape result for call statement accurate.
>
> But I find the test case still
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #9 from linzj ---
sra may not relate to this problem. I make sra get involved, because fre fails
to recognise a mem ref and a component ref are the same reference. Maybe we
should work on fre
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
linzj changed:
What|Removed |Added
CC||manjian2006 at gmail dot com
--- Comment #8
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
--- Comment #7 from linzj ---
The core of this problem is escaped result is function wise, not block wise, or
instruction wise. Any place in the function the local variable escapes, will
count that variable as escaped.
Actually the printf does
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84824
Richard Biener changed:
What|Removed |Added
Keywords||alias, missed-optimization
13 matches
Mail list logo