Thanks for the clarification, but the state doesn't flip on the next minor GC, it has to wait for the next major GC.
On a minor GC the state is saved in previous_ and after the minor GC it is restored, so nothing that happens in the minor GC can change the state from SWEEPING to NOT_RUNNING. It probably doesn't matter. Presumably the distinction between SWEEPING and NOT_RUNNING doesn't affect performance? On Fri, Nov 15, 2024 at 10:10 AM Omer Katz (chromium.org) < omerk...@chromium.org> wrote: > Hi Erik, > > Yeah, I see how that can be confusing. > In my previous answer I assumed you meant there are pages that are only > swept when we start the next GC, which is uncommon but possible. > What you're describing now, where all pages are swept but the GC remains > in a sweeping state, is very likely and intentional. > > Sweeping the pages can happen either concurrently or during allocations. > Switching state happens after all pages are swept but can be delayed. > Finalizing sweeping, and the state change that goes with it, requires some > accounting work which, for performance reasons, we don't want to do during > an allocation (it also has to happen on the main thread so can be > concurrent). > Therefore the state remains as is until we can "safely" (i.e. without > incurring an unnecessary performance cost) finalize sweeping, which is > usually when we start the next GC (minor or major). > > I hope that clarifies things. > > Omer > > On Thu, Nov 14, 2024 at 9:30 PM Erik Corry <erikco...@chromium.org> wrote: > >> I see. I was just confused because the GC state stays in SWEEPING until >> the next big GC, even when all pages have been swept. >> >> On Wed, Nov 13, 2024 at 4:56 PM Omer Katz (chromium.org) < >> omerk...@chromium.org> wrote: >> >>> Sweeping happens either concurrently (disabled by --single-threaded-gc) >>> or on allocation if we will otherwise run out of memory. >>> It's possible that you don't hit the cases that would require sweeping >>> on allocation. >>> >>> On Wed, Nov 13, 2024 at 4:49 PM Erik Corry <erikco...@chromium.org> >>> wrote: >>> >>>> We run with --single-threaded-gc. It looks like no sweeping of old >>>> space takes place until the next old-space GC starts up. Does this sound >>>> likely? I'm putting a breakpoint in Sweeper::RawSweep, and it's not hit >>>> between old-space-GCs, even though the GC state is SWEEPING. Is there >>>> a different place that does the sweeping work? >>>> >>>> -- >>>> -- >>>> v8-dev mailing list >>>> v8-dev@googlegroups.com >>>> http://groups.google.com/group/v8-dev >>>> --- >>>> You received this message because you are subscribed to the Google >>>> Groups "v8-dev" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to v8-dev+unsubscr...@googlegroups.com. >>>> To view this discussion visit >>>> https://groups.google.com/d/msgid/v8-dev/CAHZxHpiAZ0EQNJFmubnwvVxtVAu1vJTFg-1AXNEXeX8ojMqazg%40mail.gmail.com >>>> <https://groups.google.com/d/msgid/v8-dev/CAHZxHpiAZ0EQNJFmubnwvVxtVAu1vJTFg-1AXNEXeX8ojMqazg%40mail.gmail.com?utm_medium=email&utm_source=footer> >>>> . >>>> >>> -- >>> -- >>> v8-dev mailing list >>> v8-dev@googlegroups.com >>> http://groups.google.com/group/v8-dev >>> --- >>> You received this message because you are subscribed to the Google >>> Groups "v8-dev" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to v8-dev+unsubscr...@googlegroups.com. >>> To view this discussion visit >>> https://groups.google.com/d/msgid/v8-dev/CAJFGqfO8RRmOh6sbeu4AckMf%2B0FqNoh%3D-7W%3DAn1SjNGgo6o0wQ%40mail.gmail.com >>> <https://groups.google.com/d/msgid/v8-dev/CAJFGqfO8RRmOh6sbeu4AckMf%2B0FqNoh%3D-7W%3DAn1SjNGgo6o0wQ%40mail.gmail.com?utm_medium=email&utm_source=footer> >>> . >>> >> -- >> -- >> v8-dev mailing list >> v8-dev@googlegroups.com >> http://groups.google.com/group/v8-dev >> --- >> You received this message because you are subscribed to the Google Groups >> "v8-dev" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to v8-dev+unsubscr...@googlegroups.com. >> To view this discussion visit >> https://groups.google.com/d/msgid/v8-dev/CAHZxHpjKD%3DUAWs5E3yvP9%3Douc%2BA2-6xcqBc6%3DS-HG9z0s4n6Pw%40mail.gmail.com >> <https://groups.google.com/d/msgid/v8-dev/CAHZxHpjKD%3DUAWs5E3yvP9%3Douc%2BA2-6xcqBc6%3DS-HG9z0s4n6Pw%40mail.gmail.com?utm_medium=email&utm_source=footer> >> . >> > -- > -- > v8-dev mailing list > v8-dev@googlegroups.com > http://groups.google.com/group/v8-dev > --- > You received this message because you are subscribed to the Google Groups > "v8-dev" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to v8-dev+unsubscr...@googlegroups.com. > To view this discussion visit > https://groups.google.com/d/msgid/v8-dev/CAJFGqfPGM33z%3DF9QX%3Du1pUXf_39nfJPSvMA03SEeUd%3D0VD8nEA%40mail.gmail.com > <https://groups.google.com/d/msgid/v8-dev/CAJFGqfPGM33z%3DF9QX%3Du1pUXf_39nfJPSvMA03SEeUd%3D0VD8nEA%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > -- -- v8-dev mailing list v8-dev@googlegroups.com http://groups.google.com/group/v8-dev --- You received this message because you are subscribed to the Google Groups "v8-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to v8-dev+unsubscr...@googlegroups.com. To view this discussion visit https://groups.google.com/d/msgid/v8-dev/CAHZxHpiQJMG68dp66ydki7uQVVc2ga_%3Djkr54ercKpYdYNE9OA%40mail.gmail.com.
