[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250321152116.61bc7265@inria.fr>
Date: Fri, 21 Mar 2025 15:21:16 +0100
From: Jₑₙₛ Gustedt <jens.gustedt@...ia.fr>
To: Rich Felker <dalias@...c.org>
Cc: musl@...ts.openwall.com
Subject: Re: implement pthread_cond_clockwait?
Hello Rich,
on Fri, 21 Mar 2025 08:50:08 -0400 you (Rich Felker <dalias@...c.org>)
wrote:
> On Fri, Mar 21, 2025 at 11:05:27AM +0100, Jₑₙₛ Gustedt wrote:
> > Hi,
> > Posix 2024 now has this function and it is mandatory. It seems that
> > it is not very difficult to implement in musl. Would it make sense
> > to send a patch?
>
> I think so, but there might already be one.
I did not find easily on the mailing list. Where should I look
otherwise?
> IIRC there are several other explicit-clock functions to be added.
Yes, seems to be for all relevant lock structures that have interfaces
with timing.
> I was originally hesitant before they were standard because it's an
> extra level of call that can't be a tail-call on many archs (because
> extra argument), but now it's necessary to add them anyway. If the
> cost matters it could probably be alleviated with an explicit
> trylock first, but we already do the equivalent for normal-type
> mutexes which are the main ones whose performance matters.
`pthread_cond_t` is especially simple, because there is already
internal state (coming from `pthread_condattr_setclock`) that is
loaded into a local variable. So basically all is already in place,
we'd just have to split the interface a bit.
For the possible performance impact, at least for that one it would
only be potential reordering of parameters. They had the brilliant
idea of putting this new argument as the one before last. So there
would probably be an overhead of one or two instructions ;-)
So I would see this to move the current implementation of
`pthread_cond_timedwait` to a new `__pthread_cond_clockwait` by
lifting the local clock parameter into the parameters. Then
`pthread_cond_timedwait` and `pthread_cond_clockwait` could be simple
wrappers to that.
Jₑₙₛ
--
:: ICube :::::::::::::::::::::::::::::: deputy director ::
:: Université de Strasbourg :::::::::::::::::::::: ICPS ::
:: INRIA antenne de Strasbourg :::::::::::::::::: Camus ::
:: INRIA PIQ program Strasbourg :::::::::: piq.inria.fr ::
:: :::::::::::::::::::::::::::::::::::: ☎ +33 368854536 ::
:: https://icube-icps.unistra.fr/index.php/Jens_Gustedt ::
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
