help-smalltalk
[Top][All Lists]
Advanced

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

Re: [Help-smalltalk] Roadmap


From: Paolo Bonzini
Subject: Re: [Help-smalltalk] Roadmap
Date: Thu, 17 Jan 2013 15:57:14 +0100
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130110 Thunderbird/17.0.2

Il 17/01/2013 14:58, Janko Mivšek ha scritto:
> Hi Paolo,
> 
> Dne 17. 01. 2013 14:48, piše Paolo Bonzini:
> 
>>> So, a patch would be submitted on GST, later ported to VW and production
>>> tested, then down-ported to all other dialects. This is how is a current
>>> procedure but this can be changed of course.
>>
>> Honestly, I'd be reluctant to do that because I am still waiting for my
>> Swazoo improvements (both portability fixes and performance
>> improvements) to be committed, after 2 years---even though I did the
>> porting to Pharo myself and submitted nice changesets.
> 
> "Don't touch if it works" mentality is in works here, I admit :) That
> is, Swazoo works, and it was spent quite some time in optimizations by
> me before.

For some of them, the point is that GNU Smalltalk doesn't have the same
(Smalltalk-80) implementation of streams as Squeak and VW, so
SwazooBuffer almost needed a rewrite.  I also made it portable and
faster.  And when I sent it, it was against top of tree, now it requires
a refresh and I don't have time to do it.

Streams are very efficient in GNU Smalltalk, so some problems may come
up that aren't visible in other dialects.

http://smalltalk.gnu.org/blog/bonzinip/optimizing-http-header-manipulation-part-2

  "11% of the time is spent converting strings to uppercase. And
  another 8% is spent iterating through a dictionary. Neither thing
  seems very cunning; though in defense of the Swazoo authors, these
  11+8=19% represented a meager 3% before my optimizations on streams."

Also, the other functionality (SCGI) is used by Nicolas Petton.  He's
using all of it in production.

Paolo

> Including input/output streaming support. To include your
> patches would need quite some time carefully integrating, testing etc..
> Are they improve Swazoo so much to be worth spend that time? Not that I
> won't integrate them, but I really don't have much time these days.
> That's why only important things like WebSocket support were integrated
> recently.
> 
> Janko
> 




reply via email to

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