web serving speed thoughts

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

web serving speed thoughts

Paolo Bonzini-2
I pushed a couple of commits to avoid using a #nextPut:/#next loop in
Stream>>#upToEnd.  It's obviously better to just use stream-to-stream
#nextPutAll:, but I had to fix a bug before. :-)

With this change #next moves a bit down in the profile on a Seaside
test, and #nextPut: is also less heavily used.  The latter however
remains the top method in the profile, which makes sense since a web
server will write much more than it reads!

It looks like the parsing of fields in Swazoo is quite expensive because
of the many indirections between SwazooBuffer and SwazooStream.  I'll
probably look at eliminating SwazooBuffer so that operations like
SwazooStream>>#upTo: and SwazooStream>>#nextUnfoldedLine can be
optimized better.

Paolo



_______________________________________________
help-smalltalk mailing list
[hidden email]
http://lists.gnu.org/mailman/listinfo/help-smalltalk