[Combat-devs] Threaded rezz

Oskar Linden oskar at lindenlab.com
Mon Nov 15 11:31:02 PST 2010

So here's what we're going to do. We release this code Tuesday morning with
threaded rez disabled. Then sometime Wednesday or Thursday we test enable it
randomly on regions of your choice and see if things get better or worse.

I will now take requests for regions to enable threaded rez on.


On Mon, Nov 15, 2010 at 10:17 AM, Oskar Linden <oskar at lindenlab.com> wrote:

> I can answer the enabling question. The code was released the first week
> with threaded rez disabled to give time to ascterain the stability of the
> code with threaded rez disabled. Then we enabled it to have something to
> compare it to. I believe that when we release the code we are doing it with
> threaded rez disabled. We will then enable it sometime after that. Enabling
> and disabling code like this is a fairly simple procedure and does not
> require grid rollbacks.
> __Oskar
> On Mon, Nov 15, 2010 at 1:33 AM, Tony Dodd <tony.dodd at btinternet.com>wrote:
>> I'm concerned about the timescale proposed for the rollout of threaded
>> rezz.
>> I may have misunderstood this, but the impression I have is that the
>> feature
>> has only just been enabled on Le Tigre, and yet it is proposed to roll it
>> to
>> the whole grid some time this week. Was it disabled earlier because it was
>> regarded as unsafe? If so, is the timescale for rollout now not a little
>> hasty?
>> Combat sims have had issues recently with rezzing objects, and I feel
>> there
>> ought to be some stress tests in very busy sims before this feature is
>> taken
>> any further. I can't find any technical information about what will be
>> involved but let me list some totally un1nformed concerns.
>> 1. Does 'rezz' include attaching and entering a sim with an attachment?
>> Sim
>> incoming attachments are a very major source of worry in CCS sims and any
>> changes in this area need good testing.
>> 2. Is there some tradeoff to avoid creation of a huge number of trivial
>> threads? Fast guns rezz bullets from multiple scripts at a quite alarming
>> rate, and the bullets are usually single prims with minimal texturing, so
>> there is a danger that the overhead of creating and taking down a thread
>> will be greater than the work the thread actually does.
>> 3. Will the server be robust about partially rezzed objects? If the rezz
>> thread suspends and an object collides with the partially rezzed object,
>> will that show up? These are very simple considerations and I don't want
>> for
>> a moment to imply the developers won't have considered them, but some
>> technical information would be useful if we are to monitor these changes
>> for
>> potential issues.
>> 4. Is scripting affected? I would assume, given that the mono evaluator
>> already has a complex scheduling system of its own, that at the end of
>> rezz
>> the invocation of the on_rez or attach handler is simply kicked into the
>> queue and evaluated as before. If the story is any more complicated than
>> this I'd really like to hear it!
>> I'll try later to speak to Oskar in-world and see if he feels that some
>> testing in a busy combat sim would be a good precaution. If this has
>> already
>> happened without my knowledge, or if I've missed information that answers
>> my
>> questions and concerns above, perhaps someone could mail details to this
>> list.
>> Maldoror Bowman
>> _______________________________________________
>> Click here to unsubscribe or manage your list subscription:
>> https://lists.secondlife.com/cgi-bin/mailman/listinfo/combat-devs
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/combat-devs/attachments/20101115/09dda2b9/attachment.htm 

More information about the Combat-devs mailing list