|
From: | Eric Hughes |
Subject: | Re: [Gnash-dev] performance and blocking I/O |
Date: | Fri, 04 May 2007 14:11:25 -0600 |
At 01:37 PM 5/4/2007, you wrote:
I haven't measured blocking reads specifically, but I think you're right that we have a problem there.[...] What I can tell is that I see (visually, looking at the system monitor), that each curl-based load gets CPU usage very high, possible due to the horrible poll loop in ::fill_cache.
A polling loop for I/O will show much different real-time performance for local vs. remote network activity, since all the network latencies appear in the timings. My recommended next step is to profile at a fine enough granularity to determine the amount of time spent blocking. This will verify the problem and gauge just how big the problem is. On risk management grounds, it will be totally worth while to do this, since changing the control flow for I/O basics can be (perhaps, "usually is"?) a major change. The other reason to get this working is to be able to run it again, after changes are made, to measure the benefit of the change.
Eric
[Prev in Thread] | Current Thread | [Next in Thread] |