how to interpret these results ?
Mon Aug 7 23:51:06 CEST 2006
Speaking out of near total ignorance when it comes to how the Erlang
VM is implemented: can't the Erlang emulator spawn a new OS thread(s)
to handle STDIN/STDOUT more efficiently? Using a few OS threads for
special purposes and a single OS thread with non-blocking IO for
everything else seems like it would give the best of both worlds for a
very small cost of a few extra threads.
On 8/7/06, Ulf Wiger (TN/EAB) <> wrote:
> For one thing, Erlang applications traditionally haven't been that
> dependent on stdin/stdout. They've been embedded systems managing tens
> of thousand signaling sessions at a time. Optimizing the performance of
> stdin/stdout hasn't been an important requirement for any (paying)
> This also means that Erlang can never do a blocking wait on stdin.
> Non-blocking I/O is much slower than blocking I/O, esp in low-level
> benchmarks. But soft real-time characteristics is much more important
> than raw speed in telecoms applications.
> The SMP version of Erlang may be able to improve things, I think. You
> can have I/O threads in blocking wait, and won't have to suspend the
> Erlang processes in order to poll descriptors.
> Just my half-educated guesses. Someone else may have a better theory.
> Ulf W
> > -----Original Message-----
> > From:
> > [mailto:] On Behalf Of James Hague
> > Sent: den 7 augusti 2006 22:47
> > To: Erlang Users' List
> > Subject: Re: how to interpret these results ?
> > >erlang is REALLY bad at using stdin/stdout
> > I've wonderered before why this is. Ditto for line-oriented
> > I/O being slow in Erlang. What's the fundamental reason for this?
More information about the erlang-questions