Comments on: Performance vs Responsiveness —or— How I Made the Parser Twice As Fast in One Day

By: video converter

video converter — Wed, 02 Dec 2009 13:28:27 +0000

hey!

By: Employment

Employment — Thu, 12 Nov 2009 06:38:49 +0000

I must say that it really feels much more responsive. It felt sluggish before, but it's quick and snappy now.

By: message

message — Sat, 03 Oct 2009 14:29:49 +0000

I thought you were using the DOM setTimeout here. Missed the Utils part. ed hardy UGG Boots

By: Dirk

Dirk — Fri, 04 Sep 2009 08:20:05 +0000

Hey there, it feels much better now. Keep up the good work, I love it.

By: mitcho

mitcho — Tue, 18 Aug 2009 06:01:36 +0000

Thanks! I'm glad you feel it's also improved the keystroke issue as well.

By: Trent

Trent — Tue, 18 Aug 2009 05:45:53 +0000

I notice a marked improvement. The missing keystrokes was a problem before … but so far so good with the new update. Kudos … This is how is should feel.

By: mitcho

mitcho — Fri, 14 Aug 2009 16:42:01 +0000

I'm glad you've been able to notice it!

By: Nick

Nick — Fri, 14 Aug 2009 16:06:05 +0000

Nice work. I must say that it really feels much more responsive. It felt sluggish before, but it's quick and snappy now.

By: Boris

Boris — Thu, 13 Aug 2009 16:25:43 +0000

Ah, for some reason I thought you were using the DOM setTimeout here. Missed the Utils part.

By: mitcho

mitcho — Thu, 13 Aug 2009 16:22:56 +0000

Boris, I apologize if I was unclear. Each timeout is 0s, using the XPCOM timer which we wrapped up as Utils.setTimeout (seen above). The timeouts are not 10ms.

When I said "By moving this into that loop, I was able to bring this initial synchronous time down to around 10 ms." I meant that, due to the overhead of setting up the parse and the yieldingParse, the first time we yield control to the event loop is now 10ms into the parse, rather than 15-20. Is that what you were referring to?

By: Boris

Boris — Thu, 13 Aug 2009 16:08:31 +0000

Do you really want 10ms timeouts between the parse attempts? Or do you just want to yield control to the event loop? If the latter, I'd use a 0-length XPCOM timer instead….