[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CA+TsHUByjjjXC4U9k4+o=Op-QLg+V3xgjTjjF6ziOf-BZBM0zA@mail.gmail.com>
Date: Sun, 9 Mar 2014 01:25:34 +0530
From: Sayantan Datta <std2048@...il.com>
To: john-dev <john-dev@...ts.openwall.com>
Subject: Re: MPI signalling
On Sun, Mar 9, 2014 at 12:25 AM, magnum <john.magnum@...hmail.com> wrote:
> Lukas, Sayantan, or anyone else knowing MPI -
>
> Do you know what would be the right MPI mechanism for just sending a
> trival "signal" to other processes, much like a SIGUSR1 but across network?
> There's got to be something suitable.
>
> Background: When we use --fork and the mother needs to tell children they
> should re-sync the pot file, it sends them a SIGUSR2 (in latest bleeding).
> If a child wants to initiate the same thing, it sends the mother a SIGUSR2,
> and mother relays to other children.
>
> With MPI, I thought I'd use MPI_Isend() for the sender, and receivers
> check for that with MPI_Irecv() once in a while using a timer, but it turns
> out that while it's asynchronous, it's still not of the kind I want because
> if I do an MPI_Irecv(), it will wait forever (though non blocking) for a
> message. That is not what I want: What I need is more of a "was there
> anything sent to me?" function that may return "nope, it wasn't".
>
> magnum
>
>
Hi magnum,
Do you need MPI_Iprobe ?
Regards,
Sayantan
Content of type "text/html" skipped
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.
