https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #15 from rguenther at suse dot de ---
On Tue, 1 Apr 2025, ak at gcc dot gnu.org wrote:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
>
> ak at gcc dot gnu.org changed:
>
>What|Removed |A
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #14 from ak at gcc dot gnu.org ---
>to do this for entry_size < G.pagesize * GGC_QUIRE_SIZE, this should
>avoid fragmenting the virtual address space. Possibly do this only
>for USING_MADVISE, not sure.
Okay let me test that.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
ak at gcc dot gnu.org changed:
What|Removed |Added
CC||ak at gcc dot gnu.org
--- Commen
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #12 from Richard Biener ---
(In reply to andi from comment #8)
> > > Needs a workload where it matters
> >
> > PR119387 had
> >
> > 85.81% 1500713 cc1plus cc1plus [.]
> > ggc_internal_alloc(un
> >
> > for m
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #11 from Richard Biener ---
(In reply to Andi Kleen from comment #10)
> It doesn't really help for the PR119387 test case, perhaps not surprising
> because it optimizes freeing not allocation:
>
> Summary
> ./gcc/cc1plus-opt -w -s
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #10 from Andi Kleen ---
It doesn't really help for the PR119387 test case, perhaps not surprising
because it optimizes freeing not allocation:
Summary
./gcc/cc1plus-opt -w -std=c++20 ~/gcc/git/tsrc/119387-formatted.ii -quiet
ran
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
Andi Kleen changed:
What|Removed |Added
Attachment #60907|0 |1
is obsolete|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #8 from andi at firstfloor dot org ---
> > Needs a workload where it matters
>
> PR119387 had
>
> 85.81% 1500713 cc1plus cc1plus [.]
> ggc_internal_alloc(un
>
> for me. Can we keep an index to freelist from
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #7 from Richard Biener ---
(In reply to Andi Kleen from comment #6)
> Created attachment 60907 [details]
> patch for multiple free lists in ggc-page
>
> I saw it in some profile, but later trying didn't help anymore.
>
> Needs a wo
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #6 from Andi Kleen ---
Created attachment 60907
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=60907&action=edit
patch for multiple free lists in ggc-page
I saw it in some profile, but later trying didn't help anymore.
Needs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #5 from Richard Biener ---
Btw, I'd say for the sake of avoiding virtual memory fragmentation free_unit
should be equal to GGC_QUIRE_SIZE. But we should possibly merge adjacent
entries we don't free to power-of-two chunks and possib
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #4 from Richard Biener ---
Created attachment 57858
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57858&action=edit
better release_pages
Ah, and it's not so much fragmentation but large free_unit (1MB) that's hard
to get to.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
--- Comment #3 from Richard Biener ---
Created attachment 57856
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57856&action=edit
quick skip-list patch
Before:
> /usr/bin/time ./cc1plus -quiet -o /dev/null /tmp/a-test-poly.ii -O
173.29use
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
Richard Biener changed:
What|Removed |Added
CC||jakub at gcc dot gnu.org
--- Comment #
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114563
Richard Biener changed:
What|Removed |Added
Keywords||compile-time-hog
See Also|
15 matches
Mail list logo
