[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gforth] Working example of tasker.fs use

From: Bernd Paysan
Subject: Re: [gforth] Working example of tasker.fs use
Date: Tue, 16 Dec 2014 14:48:43 +0100
User-agent: KMail/4.14.3 (Linux/3.16.6-2-desktop; KDE/4.14.3; x86_64; ; )

Am Dienstag, 16. Dezember 2014, 08:31:39 schrieb Anton Ertl:
> On Mon, Dec 15, 2014 at 10:45:09PM +0100, Bernd Paysan wrote:
> > BTW: Modern OSes support nanosecond sleeps,
> > so what about having : ms ( n -- ) 1000000 m* ntime d+ ns ; and putting in
> > the actual replacible sleep logic into ns?
> > 
> > Actually, the right spec is to define an absolute deadline instead of a
> > relative; we should do that with ns (ANS/Forth200x failed with MS); maybe
> > the name might need to change (ABSTIME-NS? DEADLINE-NS?).
> DEADLINE sounds good.  One would naturally use it with UTIME (or
> something like it), so either we make DEADLINE accept microseconds, or
> we add NSTIME to produce nanoseconds.

I've already added NTIME ( -- d ); the consensus between different API makers 
seems to be that a nanosecond is a good unit, even though current CPUs and 
OSes don't even deliver microsecond accuracy for waiting, and take nearly half 
a microsecond for reading out the timer... but that might change.

> My guess is that the idea is that there are usually no threads that
> wait for both an absolute deadline and some other event; so threads
> that wait for an absolute deadline, use pthread_cond_timedwait(),
> others use select() or poll() or something.

No, unfortunately, the typical case is that you wait for an event *and* have a 
deadline.  Typically, you want to wait for the deadline, but you need to 
respond e.g. to inter-task communication, too, or accept other asynchronous 

> We can implement the absolute timeout with a relative one:
> : deadline ( d -- )
>   utime d- dup 0> if us else 2drop then ;
> where US ( d -- ) is a microsecond variant of MS.


> Are we in the word-savers club?  That's as idiotic as F~.  Such
> trickery is only justified if we want to reconcile two existing
> conflicting usages of NS, but that's not the case here.  Just have NS
> (or US) for relative deadlines (analogous to MS), and DEADLINE for
> absolute ones.


Bernd Paysan
"If you want it done right, you have to do it yourself"

reply via email to

[Prev in Thread] Current Thread [Next in Thread]