This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/86072] Poor codegen with atomics
- From: "rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Thu, 07 Jun 2018 08:27:04 +0000
- Subject: [Bug tree-optimization/86072] Poor codegen with atomics
- Auto-submitted: auto-generated
- References: <bug-86072-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86072
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2018-06-07
Ever confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Andrew is right. Still confirmed.
Somebody has to decide if it's worth optimizing them and has to sit down and
exactly specify what kind of optimizations are valid.
I guess it's worth optimizing them if these cases appear in real-world code
(and then we'd like to see examples).
As for validity not optimizing them leads to (wanted/required) side-effects
like being a barrier for a lot of compiler optimizations. Which may mor
may not be designed that way. I'm sitting on patches teaching points-to
about atomics, specifically making them not escape points or uses/clobbers
for all memory. Esp. the latter I'm not sure is a good idea to improve,
because there's nothing in the compiler making atomics "special", they
are just modeled as function calls.