By: Linus Torvalds (torvalds.delete@this.linux-foundation.org), January 4, 2020 3:54 pm
Room: Moderated Discussions
Linus Torvalds (torvalds.delete@this.linux-foundation.org) on January 4, 2020 1:24 pm wrote:
>
> Now, if you didn't get the lock, you might decide that "hey, lots of locks are held only
> for very short times, and I'll try to wait for a very bounded time".
Btw, I do not really suggest you do this unless you have serious numbers that show that it helps on a real-world load (and not just some "lock/unlock" benchmark).
Even a very limited "let's busy-loop" can have serious disadvantages. Doing that "CPU yield" thing ("rep nop" on x86, "yield" instruction on arm64) can trigger the hypervisor to do things in virtualized environments, as can things like doing a cycle counter read to limit the time you wait etc. It might work fine on your machine, but it can cause unexpected behavior elsewhere.
And if you only do a hard busy loop reading the lock value, you might be wasting time, energy and cause unnecessary cache contention. If the loop is limited to a small value, it might have very dubious value, and if you make it longer, the downsides can grow a lot. And the value tuning might be as much about your load as it is about the particular machine you are running on, etc, so you might again be tuning only for your development machine, not any real life situation.
So to a first approximation, start off with that fast-case "trylock/tryunlock" model where you handle the case of absolutely no contention specially yourself. And then fall back to a sleeping real model like futexes immediately after that unless you really know a lot about your workload.
In fact, feel free to skip even that fast case. Yes, it often makes a lot of sense to do a special and simple fast case: in a lot of real-world locking situations you'll hardly ever see any contention at all, and you might be writing a library that is often used in a completely single-threaded manner.
So handling the trivial non-contended locking case with a special fast-case is quite often a very good idea.
But "quite often" doesn't mean "always". Locking might simply either be so rare that it's pointless to have a special fast case in the first place if the fallback isn't actively horrible. Or the lock might be designed to be a resource limiter and the contention case is the expected case because you have a thousand threads that might want to run, but you use the lock as a way to limit them to a much smaller number (ie the GNU make "jobserver" case).
So my suggestion is to basically always start off with just pthread_mutex_lock() or other generic non-spinning locking (std::mutex, whatever you have in your environment).
And if that turns out to have performance problems and you know your behavior, go out and google for somebody that already solved it and wrote a paper about it, and has numbers for a lot of different scenarios. I see that El Presidente already pointed to a report by Waiman Long elsewhere in this thread. Waiman knows what he's doing. He's one of the Linux kernel locking people.
And only after that, and only if you can't for the life of you find something that works for you, and you know that you have very very particular locking behavior (that may be the reason you can't find something that works for you), do you start going down the "let me try to write my own", and start off with the special fast case optimized for your known load, and fall back to something like futexes. But realize that it's a big job, and it will be painful.
And not in a million years will any of those choices hopefully ever include "sched_yield()", although sometimes I might worry about standard libraries (particularly if you're targeting a new architecture or new language environment, so that the standard locking implementation may basically be some stub function - it generally will take many years for this kind of infrastructure to mature).
Linus
>
> Now, if you didn't get the lock, you might decide that "hey, lots of locks are held only
> for very short times, and I'll try to wait for a very bounded time".
Btw, I do not really suggest you do this unless you have serious numbers that show that it helps on a real-world load (and not just some "lock/unlock" benchmark).
Even a very limited "let's busy-loop" can have serious disadvantages. Doing that "CPU yield" thing ("rep nop" on x86, "yield" instruction on arm64) can trigger the hypervisor to do things in virtualized environments, as can things like doing a cycle counter read to limit the time you wait etc. It might work fine on your machine, but it can cause unexpected behavior elsewhere.
And if you only do a hard busy loop reading the lock value, you might be wasting time, energy and cause unnecessary cache contention. If the loop is limited to a small value, it might have very dubious value, and if you make it longer, the downsides can grow a lot. And the value tuning might be as much about your load as it is about the particular machine you are running on, etc, so you might again be tuning only for your development machine, not any real life situation.
So to a first approximation, start off with that fast-case "trylock/tryunlock" model where you handle the case of absolutely no contention specially yourself. And then fall back to a sleeping real model like futexes immediately after that unless you really know a lot about your workload.
In fact, feel free to skip even that fast case. Yes, it often makes a lot of sense to do a special and simple fast case: in a lot of real-world locking situations you'll hardly ever see any contention at all, and you might be writing a library that is often used in a completely single-threaded manner.
So handling the trivial non-contended locking case with a special fast-case is quite often a very good idea.
But "quite often" doesn't mean "always". Locking might simply either be so rare that it's pointless to have a special fast case in the first place if the fallback isn't actively horrible. Or the lock might be designed to be a resource limiter and the contention case is the expected case because you have a thousand threads that might want to run, but you use the lock as a way to limit them to a much smaller number (ie the GNU make "jobserver" case).
So my suggestion is to basically always start off with just pthread_mutex_lock() or other generic non-spinning locking (std::mutex, whatever you have in your environment).
And if that turns out to have performance problems and you know your behavior, go out and google for somebody that already solved it and wrote a paper about it, and has numbers for a lot of different scenarios. I see that El Presidente already pointed to a report by Waiman Long elsewhere in this thread. Waiman knows what he's doing. He's one of the Linux kernel locking people.
And only after that, and only if you can't for the life of you find something that works for you, and you know that you have very very particular locking behavior (that may be the reason you can't find something that works for you), do you start going down the "let me try to write my own", and start off with the special fast case optimized for your known load, and fall back to something like futexes. But realize that it's a big job, and it will be painful.
And not in a million years will any of those choices hopefully ever include "sched_yield()", although sometimes I might worry about standard libraries (particularly if you're targeting a new architecture or new language environment, so that the standard locking implementation may basically be some stub function - it generally will take many years for this kind of infrastructure to mature).
Linus
Topic | Posted By | Date |
---|---|---|
Nuances related to Spinlock implementation and the Linux Scheduler | Beastian | 2020/01/03 12:46 PM |
Nuances related to Spinlock implementation and the Linux Scheduler | Montaray Jack | 2020/01/03 01:14 PM |
Nuances related to Spinlock implementation and the Linux Scheduler | Montaray Jack | 2020/01/03 01:49 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Linus Torvalds | 2020/01/03 07:05 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Beastian | 2020/01/04 12:03 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Malte Skarupke | 2020/01/04 12:22 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Linus Torvalds | 2020/01/04 01:31 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | dmcq | 2020/01/05 07:33 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | smeuletz | 2020/01/06 02:05 AM |
Do not blame others for your unfinished job | smeuletz | 2020/01/06 02:08 AM |
Where did all the experts come from? Did Linus get linked? (NT) | anon | 2020/01/06 04:27 AM |
Phoronix | Gabriele Svelto | 2020/01/06 05:04 AM |
Phoronix | Salvatore De Dominicis | 2020/01/06 07:59 AM |
Do not blame anyone. Please give polite, constructive criticism | Chester | 2020/01/06 09:17 AM |
Do not blame anyone. Please give polite, constructive criticism | smeuletz | 2020/01/06 10:11 AM |
Do not blame anyone. Please give polite, constructive criticism | Chester | 2020/01/06 10:54 AM |
Do not blame anyone. Please give polite, constructive criticism | smeuletz | 2020/01/06 11:33 AM |
Do not blame anyone. Please give polite, constructive criticism | Linus Torvalds | 2020/01/06 12:58 PM |
Do not blame anyone. Please give polite, constructive criticism | Gionatan Danti | 2020/01/06 01:13 PM |
Do not blame anyone. Please give polite, constructive criticism | Linus Torvalds | 2020/01/06 01:28 PM |
Do not blame anyone. Please give polite, constructive criticism | Gionatan Danti | 2020/01/06 01:52 PM |
Do not blame anyone. Please give polite, constructive criticism | John Scott | 2020/01/10 08:48 AM |
Do not blame anyone. Please give polite, constructive criticism | supernovas | 2020/01/10 10:01 AM |
Do not blame anyone. Please give polite, constructive criticism | Linus Torvalds | 2020/01/10 12:45 PM |
Do not blame anyone. Please give polite, constructive criticism | GDan | 2020/04/06 03:10 AM |
Oracle | Anon3 | 2020/04/07 06:42 AM |
Do not blame anyone. Please give polite, constructive criticism | smeuletz | 2020/01/07 04:07 AM |
Do not blame anyone. Please give polite, constructive criticism | Simon Farnsworth | 2020/01/07 01:40 PM |
Do not blame anyone. Please give polite, constructive criticism | Etienne | 2020/01/08 02:08 AM |
Do not blame anyone. Please give polite, constructive criticism | smeuletz | 2020/01/08 02:18 AM |
Do not blame anyone. Please give polite, constructive criticism | Michael S | 2020/01/08 02:56 AM |
Not deprecating irrelevant API: sched_yield() on quantum computers? | smeuletz | 2020/01/07 04:34 AM |
Do not blame anyone. Please give polite, constructive criticism | magicalgoat | 2020/01/09 05:58 PM |
Do not blame anyone. Please give polite, constructive criticism | Linus Torvalds | 2020/01/09 10:37 PM |
Do not blame anyone. Please give polite, constructive criticism | Anon3 | 2020/01/10 04:40 PM |
Do not blame anyone. Please give polite, constructive criticism | rwessel | 2020/01/06 10:04 PM |
Do not blame anyone. Please give polite, constructive criticism | Linus Torvalds | 2020/01/06 12:11 PM |
Do not blame anyone. Please give polite, constructive criticism | Gabriele Svelto | 2020/01/06 02:36 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Howard Chu | 2020/01/09 11:39 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Linus Torvalds | 2020/01/10 12:30 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | president ltd | 2020/01/04 02:44 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Jörn Engel | 2020/01/04 12:34 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Emil Briggs | 2020/01/04 01:13 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Jörn Engel | 2020/01/04 01:46 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Linus Torvalds | 2020/01/04 02:24 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Linus Torvalds | 2020/01/04 03:54 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Jörn Engel | 2020/01/05 10:21 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Linus Torvalds | 2020/01/05 12:42 PM |
FUTEX_LOCK_PI performance | Jörn Engel | 2020/01/05 02:45 PM |
FUTEX_LOCK_PI performance | Linus Torvalds | 2020/01/05 04:30 PM |
FUTEX_LOCK_PI performance | Jörn Engel | 2020/01/05 07:03 PM |
FUTEX_LOCK_PI performance | RichardC | 2020/01/06 07:11 AM |
FUTEX_LOCK_PI performance | Linus Torvalds | 2020/01/06 01:11 PM |
FUTEX_LOCK_PI performance | Gabriele Svelto | 2020/01/06 03:20 AM |
FUTEX_LOCK_PI performance | xilun | 2020/01/06 05:19 PM |
FUTEX_LOCK_PI performance | Konrad Schwarz | 2020/01/13 04:36 AM |
FUTEX_LOCK_PI performance | Gabriele Svelto | 2020/01/13 04:53 AM |
FUTEX_LOCK_PI performance | Simon Farnsworth | 2020/01/13 05:36 AM |
FUTEX_LOCK_PI performance | rwessel | 2020/01/13 06:22 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | rainstar | 2020/01/04 10:58 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Charles Ellis | 2020/01/05 04:00 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Richard | 2020/01/05 09:58 AM |
It's hard to separate | Michael S | 2020/01/05 11:17 AM |
It's hard to separate | rainstared | 2020/01/06 01:52 AM |
It's hard to separate | David Kanter | 2020/01/08 09:27 AM |
It's hard to separate | Anon | 2020/01/08 09:37 PM |
It's hard to separate | none | 2020/01/08 11:50 PM |
It's hard to separate | Anon | 2020/01/09 01:41 AM |
It's hard to separate | none | 2020/01/09 03:54 AM |
It's hard to separate | gallier2 | 2020/01/09 04:19 AM |
It's hard to separate | Anon | 2020/01/09 05:12 AM |
It's hard to separate | Adrian | 2020/01/09 05:24 AM |
It's hard to separate | gallier2 | 2020/01/09 05:58 AM |
It's hard to separate | Adrian | 2020/01/09 07:09 AM |
It's hard to separate | gallier2 | 2020/01/09 05:42 AM |
It's hard to separate | Adrian | 2020/01/09 04:41 AM |
It's hard to separate | Anon | 2020/01/09 05:24 AM |
It's hard to separate | gallier2 | 2020/01/09 06:07 AM |
It's hard to separate | David Hess | 2020/01/09 09:27 AM |
It's hard to separate | Adrian | 2020/01/09 10:15 AM |
It's hard to separate | David Hess | 2020/01/09 10:45 AM |
It's hard to separate | Anon | 2020/01/09 11:15 AM |
It's hard to separate | Adrian | 2020/01/09 11:51 AM |
It's hard to separate | Brett | 2020/01/09 01:49 PM |
Zilog Z8000 | Brett | 2020/01/10 10:53 PM |
Zilog Z8000 | David Hess | 2020/01/11 07:06 AM |
Zilog Z8000 | Adrian | 2020/01/11 07:29 AM |
Zilog Z8000 | David Hess | 2020/01/11 08:45 AM |
Zilog Z8000 | Ricardo B | 2020/01/11 08:04 PM |
Zilog Z8000 | Ronald Maas | 2020/01/12 10:47 AM |
Zilog Z8000 | Ricardo B | 2020/01/12 12:15 PM |
Zilog Z8000 | Anon | 2020/01/12 11:34 PM |
Zilog Z8000 | Jose | 2020/01/13 01:23 AM |
Zilog Z8000 | gallier2 | 2020/01/13 01:42 AM |
Zilog Z8000 | Jose | 2020/01/13 10:04 PM |
Zilog Z8000 | rwessel | 2020/01/13 10:40 PM |
Zilog Z8000 | David Hess | 2020/01/13 11:35 PM |
Zilog Z8000 | Simon Farnsworth | 2020/01/14 03:56 AM |
Zilog Z8000 | Michael S | 2020/01/14 04:09 AM |
Zilog Z8000 | Simon Farnsworth | 2020/01/14 05:06 AM |
Zilog Z8000 | David Hess | 2020/01/14 10:22 AM |
Zilog Z8000 | David Hess | 2020/01/14 10:15 AM |
Zilog Z8000 | rwessel | 2020/01/14 04:12 PM |
286 16 bit I/O | Tim McCaffrey | 2020/01/15 11:25 AM |
286 16 bit I/O | David Hess | 2020/01/15 09:17 PM |
Zilog Z8000 | Ricardo B | 2020/01/13 11:52 AM |
Zilog Z8000 | Anon | 2020/01/13 12:25 PM |
Zilog Z8000 | David Hess | 2020/01/13 06:38 PM |
Zilog Z8000 | rwessel | 2020/01/13 07:16 PM |
Zilog Z8000 | David Hess | 2020/01/13 07:47 PM |
Zilog Z8000 | someone | 2020/01/14 07:54 AM |
Zilog Z8000 | Anon | 2020/01/14 08:31 AM |
Zilog Z8000 | Ricardo B | 2020/01/14 06:29 PM |
Zilog Z8000 | Simon Farnsworth | 2020/01/15 03:26 AM |
Zilog Z8000 | Tim McCaffrey | 2020/01/15 11:27 AM |
Zilog Z8000 | Simon Farnsworth | 2020/01/15 02:32 PM |
Zilog Z8000 | Ricardo B | 2020/01/15 03:47 PM |
Zilog Z8000 | Anon | 2020/01/15 04:08 PM |
Zilog Z8000 | Ricardo B | 2020/01/15 05:16 PM |
Zilog Z8000 | Anon | 2020/01/15 05:31 PM |
Zilog Z8000 | Ricardo B | 2020/01/15 06:46 PM |
Zilog Z8000 | Anon | 2020/01/15 07:04 PM |
Zilog Z8000 | David Hess | 2020/01/15 09:53 PM |
Zilog Z8000 | Ricardo B | 2020/01/16 07:27 PM |
Zilog Z8000 | Anon | 2020/01/16 08:33 PM |
Zilog Z8000 | Ronald Maas | 2020/01/17 12:05 AM |
Zilog Z8000 | Anon | 2020/01/17 08:15 AM |
Zilog Z8000 | Ricardo B | 2020/01/17 02:59 PM |
Zilog Z8000 | Anon | 2020/01/17 07:40 PM |
Zilog Z8000 | Ricardo B | 2020/01/18 08:42 AM |
Zilog Z8000 | gallier2 | 2020/01/19 08:02 AM |
Zilog Z8000 | David Hess | 2020/01/18 07:12 AM |
Zilog Z8000 | David Hess | 2020/01/15 09:49 PM |
Zilog Z8000 | gallier2 | 2020/01/16 12:57 AM |
Zilog Z8000 | Simon Farnsworth | 2020/01/16 02:30 AM |
IBM PC success | Etienne | 2020/01/16 06:42 AM |
Zilog Z8000 | Ricardo B | 2020/01/16 07:32 PM |
Zilog Z8000 | Brett | 2020/01/17 01:38 AM |
Zilog Z8000 | David Hess | 2020/01/18 07:28 AM |
Zilog Z8000 | David Hess | 2020/01/18 07:22 AM |
Zilog Z8000 | David Hess | 2020/01/15 09:30 PM |
Zilog Z8000 | Maxwell | 2020/01/11 09:07 AM |
Zilog Z8000 | David Hess | 2020/01/11 09:40 AM |
Zilog Z8000 | Maxwell | 2020/01/11 10:08 AM |
Zilog Z8000 | Ricardo B | 2020/01/11 08:42 PM |
8086 does NOT have those addressing modes | Devin | 2020/01/12 02:13 PM |
8086 does NOT have those addressing modes | Ricardo B | 2020/01/12 06:46 PM |
8086 does NOT have those addressing modes | Anon | 2020/01/13 05:10 AM |
8086 does NOT have those addressing modes | gallier2 | 2020/01/13 06:07 AM |
8086 does NOT have those addressing modes | Anon | 2020/01/13 07:09 AM |
8086 does NOT have those addressing modes | Ricardo B | 2020/01/13 11:48 AM |
8086 does NOT have those addressing modes | Michael S | 2020/01/13 07:40 AM |
Zilog Z8000 | Ronald Maas | 2020/01/13 09:44 AM |
Zilog Z8000 | Anon | 2020/01/13 04:32 PM |
8086 does NOT have those addressing modes | Ricardo B | 2020/01/13 11:24 AM |
8086 does NOT have those addressing modes | rwessel | 2020/01/13 03:59 PM |
8086 does NOT have those addressing modes | David Hess | 2020/01/13 07:12 PM |
8086 does NOT have those addressing modes | rwessel | 2020/01/13 07:28 PM |
8086 does NOT have those addressing modes | David Hess | 2020/01/13 07:51 PM |
8086 does NOT have those addressing modes | David Hess | 2020/01/13 06:55 PM |
Zilog Z8000 | rwessel | 2020/01/11 01:26 PM |
Zilog Z8000 | Brett | 2020/01/11 03:16 PM |
Zilog Z8000 | rwessel | 2020/01/11 08:20 PM |
Zilog Z8000 | Brett | 2020/01/12 01:02 PM |
Zilog Z8000 | rwessel | 2020/01/12 10:06 PM |
Zilog Z8000 | Brett | 2020/01/12 11:02 PM |
Zilog Z8000 | James | 2020/01/13 06:12 AM |
Zilog Z8000 | Adrian | 2020/01/12 12:38 AM |
PDP-11 | Michael S | 2020/01/12 02:33 AM |
Zilog Z8000 | rwessel | 2020/01/12 07:01 AM |
Zilog Z8000 | Ronald Maas | 2020/01/12 11:03 AM |
Zilog Z8000 | Konrad Schwarz | 2020/01/13 04:49 AM |
Zilog Z8000 | Adrian | 2020/01/14 12:38 AM |
Zilog Z8000 | konrad.schwarz | 2020/01/15 05:50 AM |
Zilog Z8000 | Adrian | 2020/01/15 11:24 PM |
It's hard to separate | David Hess | 2020/01/11 07:08 AM |
It's hard to separate | David Hess | 2020/01/11 07:11 AM |
It's hard to separate | Adrian | 2020/01/09 12:16 PM |
It's hard to separate | David Hess | 2020/01/11 07:17 AM |
It's hard to separate | gallier2 | 2020/01/10 01:11 AM |
It's hard to separate | none | 2020/01/10 02:58 AM |
It's hard to separate | rwessel | 2020/01/09 08:00 AM |
It's hard to separate | David Hess | 2020/01/09 09:10 AM |
It's hard to separate | rwessel | 2020/01/09 09:51 AM |
It's hard to separate | Adrian | 2020/01/08 11:58 PM |
It's hard to separate | rwessel | 2020/01/09 07:31 AM |
It's hard to separate | Adrian | 2020/01/09 07:44 AM |
It's hard to separate | David Hess | 2020/01/09 09:37 AM |
It's hard to separate | none | 2020/01/09 10:34 AM |
Are segments so bad? | Paul A. Clayton | 2020/01/09 03:15 PM |
Yes, they are terrible (NT) | Anon | 2020/01/09 03:20 PM |
Are segments so bad? | Adrian | 2020/01/10 12:49 AM |
Are segments so bad? | Etienne | 2020/01/10 02:28 AM |
Are segments so bad? | gallier2 | 2020/01/10 02:37 AM |
Are segments so bad? | Adrian | 2020/01/10 03:19 AM |
Are segments so bad? | Adrian | 2020/01/10 04:27 AM |
Are segments so bad? | Etienne | 2020/01/10 04:41 AM |
Are segments so bad? | Adrian | 2020/01/10 03:05 AM |
Are segments so bad? | gallier2 | 2020/01/10 03:13 AM |
Are segments so bad? | Anon3 | 2020/01/10 11:37 AM |
Are segments so bad? | Adrian | 2020/01/10 11:47 AM |
Are segments so bad? | Brendan | 2020/01/11 01:43 AM |
Are segments so bad? | Anon | 2020/01/10 06:51 PM |
Are segments so bad? | Adrian | 2020/01/11 01:05 AM |
Are segments so bad? | Jukka Larja | 2020/01/11 08:20 AM |
Are segments so bad? | Brendan | 2020/01/11 10:14 AM |
Are segments so bad? | Jukka Larja | 2020/01/11 09:15 PM |
Are segments so bad? | Brendan | 2020/01/11 11:15 PM |
Are segments so bad? | Jukka Larja | 2020/01/12 04:18 AM |
Are segments so bad? | anon | 2020/01/12 12:30 PM |
Are segments so bad? | Brendan | 2020/01/12 10:19 PM |
the world sucks worse than you're aware of | Michael S | 2020/01/13 01:50 AM |
the world sucks worse than you're aware of | Brendan | 2020/01/13 03:56 AM |
the world sucks worse than you're aware of | Gabriele Svelto | 2020/01/13 04:46 AM |
Are segments so bad? | Jukka Larja | 2020/01/13 07:41 AM |
Are segments so bad? | Brendan | 2020/01/13 08:21 AM |
Are segments so bad? | Jukka Larja | 2020/01/13 09:43 AM |
Are segments so bad? | Brendan | 2020/01/13 01:02 PM |
Are segments so bad? | Anne O. Nymous | 2020/01/13 01:22 PM |
Are segments so bad? | Brendan | 2020/01/13 02:50 PM |
actor of around 200? | Michael S | 2020/01/14 03:58 AM |
Not overcomitting leads to more OOMs, not less | Gabriele Svelto | 2020/01/14 12:50 PM |
Not overcomitting leads to more OOMs, not less | Brendan | 2020/01/14 01:40 PM |
Not overcomitting leads to more OOMs, not less | Gabriele Svelto | 2020/01/15 03:17 AM |
Not overcomitting leads to more OOMs, not less | Anon | 2020/01/15 04:43 AM |
Not overcomitting leads to more OOMs, not less | Gabriele Svelto | 2020/01/15 05:09 AM |
Not overcomitting leads to more OOMs, not less | Anon | 2020/01/15 05:16 AM |
Not overcomitting leads to more OOMs, not less | Gabriele Svelto | 2020/01/15 06:58 AM |
Not overcomitting leads to more OOMs, not less | Anon | 2020/01/15 09:08 AM |
Not overcomitting leads to more OOMs, not less | Gabriele Svelto | 2020/01/16 04:05 AM |
Not overcomitting leads to more OOMs, not less | Michael S | 2020/01/15 04:48 AM |
Not overcomitting leads to more OOMs, not less | Gabriele Svelto | 2020/01/15 05:10 AM |
Not overcomitting leads to more OOMs, not less | Michael S | 2020/01/15 08:13 AM |
Not overcomitting leads to more OOMs, not less | Jukka Larja | 2020/01/15 08:46 AM |
Not overcomitting leads to more OOMs, not less | Jukka Larja | 2020/01/15 06:08 AM |
Thanks for the info (NT) | Gabriele Svelto | 2020/01/15 07:00 AM |
Not overcomitting leads to more OOMs, not less | Linus Torvalds | 2020/01/15 12:30 PM |
OOM killer complains | Anon | 2020/01/15 12:44 PM |
OOM killer complains | anon | 2020/01/15 04:26 PM |
Not overcomitting leads to more OOMs, not less | Brendan | 2020/01/16 07:26 AM |
Not overcomitting leads to more OOMs, not less | Linus Torvalds | 2020/01/16 10:17 AM |
Not overcomitting leads to more OOMs, not less | Linus Torvalds | 2020/01/16 10:48 AM |
Not overcomitting leads to more OOMs, not less | Doug S | 2020/01/16 03:41 PM |
Not overcomitting leads to more OOMs, not less | Doug S | 2020/01/16 03:44 PM |
Are segments so bad? | rwessel | 2020/01/13 04:11 PM |
Are segments so bad? | Jukka Larja | 2020/01/14 07:37 AM |
Are segments so bad? | Brendan | 2020/01/14 08:48 AM |
Are segments so bad? | Jukka Larja | 2020/01/14 11:13 AM |
Are segments so bad? | Brendan | 2020/01/14 02:30 PM |
Are segments so bad? | Brett | 2020/01/14 10:13 PM |
Are segments so bad? | Jukka Larja | 2020/01/15 07:04 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/15 03:35 AM |
Specifying cost of dropping pages | Paul A. Clayton | 2020/01/13 03:00 PM |
Specifying cost of dropping pages | rwessel | 2020/01/13 04:19 PM |
Specifying cost of dropping pages | Gabriele Svelto | 2020/01/15 03:23 AM |
Are segments so bad? | anon | 2020/01/14 02:15 AM |
Are segments so bad? | Brendan | 2020/01/14 06:13 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/14 12:57 PM |
Are segments so bad? | Brendan | 2020/01/14 02:58 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/15 03:33 AM |
Are segments so bad? | Anon | 2020/01/15 05:24 AM |
Are segments so bad? | Jukka Larja | 2020/01/15 06:20 AM |
Are segments so bad? | Etienne | 2020/01/15 05:56 AM |
Are segments so bad? | Jukka Larja | 2020/01/15 08:53 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/16 06:12 AM |
Are segments so bad? | Jukka Larja | 2020/01/16 10:56 AM |
Are segments so bad? | Brendan | 2020/01/15 06:20 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/15 06:56 AM |
Are segments so bad? | Brendan | 2020/01/16 07:16 AM |
Are segments so bad? | Jukka Larja | 2020/01/16 11:08 AM |
Are segments so bad? | Brendan | 2020/01/17 01:52 PM |
Are segments so bad? | Jukka Larja | 2020/01/17 10:08 PM |
Are segments so bad? | Brendan | 2020/01/18 12:40 PM |
Are segments so bad? | Jukka Larja | 2020/01/18 10:13 PM |
Are segments so bad? | Brendan | 2020/01/19 12:25 PM |
Are segments so bad? | Brett | 2020/01/19 03:18 PM |
Are segments so bad? | Brett | 2020/01/19 03:34 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/20 12:57 AM |
Are segments so bad? | Jukka Larja | 2020/01/20 05:54 AM |
Are segments so bad? | Brendan | 2020/01/20 12:43 PM |
Are segments so bad? | Jukka Larja | 2020/01/21 07:01 AM |
Are segments so bad? | Brendan | 2020/01/21 06:04 PM |
Are segments so bad? | Jukka Larja | 2020/01/22 07:30 AM |
Are segments so bad? | Brendan | 2020/01/22 03:56 PM |
Are segments so bad? | Jukka Larja | 2020/01/23 08:44 AM |
Are segments so bad? | rwessel | 2020/01/16 03:06 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/16 03:13 PM |
Are segments so bad? | Brendan | 2020/01/17 01:51 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/17 03:18 PM |
Are segments so bad? | Anon | 2020/01/17 08:01 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/20 01:06 AM |
Are segments so bad? | Brendan | 2020/01/18 03:15 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/20 12:55 AM |
Are segments so bad? | Michael S | 2020/01/20 05:30 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/20 08:02 AM |
Are segments so bad? | Jukka Larja | 2020/01/20 08:41 AM |
Are segments so bad? | Michael S | 2020/01/20 08:45 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/20 09:36 AM |
Are segments so bad? | Brendan | 2020/01/20 11:04 AM |
Are segments so bad? | Michael S | 2020/01/20 01:22 PM |
Are segments so bad? | Brendan | 2020/01/20 02:38 PM |
Are segments so bad? | Simon Farnsworth | 2020/01/20 03:40 PM |
Are segments so bad? | Anon | 2020/01/20 04:35 PM |
Are segments so bad? | Simon Farnsworth | 2020/01/20 05:30 PM |
Are segments so bad? | Michael S | 2020/01/20 05:20 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/21 05:08 AM |
Are segments so bad? | Brendan | 2020/01/21 06:07 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/22 01:53 AM |
Are segments so bad? | Brendan | 2020/01/22 04:32 AM |
Are segments so bad? | Jukka Larja | 2020/01/22 07:12 AM |
Are segments so bad? | Brendan | 2020/01/22 04:28 PM |
Are segments so bad? | Jukka Larja | 2020/01/23 07:36 AM |
Are segments so bad? | Brendan | 2020/01/24 07:27 PM |
Are segments so bad? | Jukka Larja | 2020/01/24 10:42 PM |
Are segments so bad? | Brendan | 2020/01/25 02:46 AM |
Are segments so bad? | Jukka Larja | 2020/01/25 08:29 AM |
Are segments so bad? | Brendan | 2020/01/26 11:17 PM |
Are segments so bad? | Jukka Larja | 2020/01/27 07:55 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/27 04:33 PM |
Are segments so bad? | Jukka Larja | 2020/01/28 06:28 AM |
DDS assets and MipMap chains | Montaray Jack | 2020/01/29 03:26 AM |
Are segments so bad? | gallier2 | 2020/01/27 03:58 AM |
Are segments so bad? | Jukka Larja | 2020/01/27 06:19 AM |
Are segments so bad? | Anne O. Nymous | 2020/01/25 03:23 AM |
Are segments so bad? | Anon | 2020/01/22 05:52 PM |
Are segments so bad? | Anne O. Nymous | 2020/01/23 01:24 AM |
Are segments so bad? | Anon | 2020/01/23 05:24 PM |
Are segments so bad? | Anne O. Nymous | 2020/01/24 12:43 AM |
Are segments so bad? | Anon | 2020/01/24 04:04 AM |
Are segments so bad? | Etienne | 2020/01/24 06:10 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/23 01:48 AM |
Are segments so bad? | Michael S | 2020/01/23 03:48 AM |
Are segments so bad? | Jukka Larja | 2020/01/23 07:38 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/23 01:29 PM |
Are segments so bad? | Anon | 2020/01/23 06:08 PM |
Are segments so bad? | Jukka Larja | 2020/01/24 09:51 PM |
Are segments so bad? | Anon | 2020/01/23 06:02 PM |
Are segments so bad? | Gabriele Svelto | 2020/01/24 03:57 AM |
Are segments so bad? | Anon | 2020/01/24 04:17 AM |
Are segments so bad? | Gabriele Svelto | 2020/01/24 09:23 AM |
Are segments so bad? | Anon | 2020/02/02 10:15 PM |
Are segments so bad? | Gabriele Svelto | 2020/02/03 01:47 AM |
Are segments so bad? | Anon | 2020/02/03 02:34 AM |
Are segments so bad? | Gabriele Svelto | 2020/02/03 05:36 AM |
Are segments so bad? | Anon3 | 2020/02/03 08:47 AM |
Are segments so bad? | Anon | 2020/02/04 05:49 PM |
Are segments so bad? | Jukka Larja | 2020/01/24 10:10 PM |
Are segments so bad? | Jukka Larja | 2020/01/17 10:26 PM |
Are segments so bad? | Anne O. Nymous | 2020/01/12 04:18 AM |
Are segments so bad? | Jukka Larja | 2020/01/12 08:41 AM |
Are segments so bad? | rwessel | 2020/01/11 01:31 PM |
Are segments so bad? | Anne O. Nymous | 2020/01/11 08:22 AM |
Are segments so bad? | Ricardo B | 2020/01/11 08:01 PM |
Are segments so bad? | Adrian | 2020/01/12 12:18 AM |
Are segments so bad? | Michael S | 2020/01/12 02:43 AM |
Are segments so bad? | Adrian | 2020/01/12 04:35 AM |
Are segments so bad? | Ricardo B | 2020/01/12 12:04 PM |
Are segments so bad? | Anon3 | 2020/01/12 05:52 PM |
Are segments so bad? | Brendan | 2020/01/12 09:58 PM |
Are segments so bad? | Paul A. Clayton | 2020/01/13 09:11 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | rainstared | 2020/01/06 01:43 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Foo_ | 2020/01/06 05:33 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | dmcq | 2020/01/06 06:03 AM |
changes in context | Carlie Coats | 2020/01/09 09:06 AM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | rainstar | 2020/01/09 10:16 PM |
No nuances, just buggy code (was: related to Spinlock implementation and the Linux Scheduler) | Montaray Jack | 2020/01/09 11:11 PM |
Suggested reading for the author | anon | 2020/01/04 11:16 PM |
Suggested reading for the author | ab | 2020/01/05 05:15 AM |
Looking at the other side (frequency scaling) | Chester | 2020/01/06 10:19 AM |
Looking at the other side (frequency scaling) | Foo_ | 2020/01/06 11:00 AM |
Why spinlocks were used | Foo_ | 2020/01/06 11:06 AM |
Why spinlocks were used | Jukka Larja | 2020/01/06 12:59 PM |
Why spinlocks were used | Simon Cooke | 2020/01/06 03:16 PM |
Why spinlocks were used | Rizzo | 2020/01/07 01:18 AM |
Looking at the other side (frequency scaling) | ab | 2020/01/07 01:14 AM |
Cross-platform code | Gian-Carlo Pascutto | 2020/01/06 08:00 AM |
Cross-platform code | Michael S | 2020/01/06 09:11 AM |
Cross-platform code | Gian-Carlo Pascutto | 2020/01/06 12:33 PM |
Cross-platform code | Michael S | 2020/01/06 01:59 PM |
Cross-platform code | Nksingh | 2020/01/07 12:09 AM |
Cross-platform code | Michael S | 2020/01/07 02:00 AM |
SRW lock implementation | Michael S | 2020/01/07 02:35 AM |
SRW lock implementation | Nksingh | 2020/01/09 02:17 PM |
broken URL in Linux source code | Michael S | 2020/01/14 01:56 AM |
broken URL in Linux source code | Travis Downs | 2020/01/14 10:14 AM |
broken URL in Linux source code | Michael S | 2020/01/14 10:48 AM |
broken URL in Linux source code | Travis Downs | 2020/01/14 04:43 PM |
SRW lock implementation - url broken | Michael S | 2020/01/14 03:07 AM |
SRW lock implementation - url broken | Travis Downs | 2020/01/14 11:06 AM |
SRW lock implementation - url broken | gpderetta | 2020/01/15 04:28 AM |
SRW lock implementation - url broken | Travis Downs | 2020/01/15 11:16 AM |
SRW lock implementation - url broken | Linus Torvalds | 2020/01/15 11:20 AM |
SRW lock implementation - url broken | Travis Downs | 2020/01/15 11:35 AM |
SRW lock implementation - url broken | Linus Torvalds | 2020/01/16 11:24 AM |
SRW lock implementation - url broken | Konrad Schwarz | 2020/02/05 10:19 AM |
SRW lock implementation - url broken | nksingh | 2020/02/05 02:42 PM |
Cross-platform code | Linus Torvalds | 2020/01/06 01:57 PM |