Using a lot of CPU

[RoLex 1]

Hi.

Congratulations on new release. But there is one but as always, it's exactly the same for my part, as was in beta version of 1.2.0.

This new version constantly uses alot CPU, this never happens in 1.1.0. I have started the client, opened my regular 30 hubs and waiting. It takes about 15 seconds, the CPU usage is normal, about 0-5%, that is what I always have with 1.1.0. Then after those 15 seconds the CPU usage raises up to 80-90%, it holds there for about 5-8 seconds and then goes back to 0-5% of usage. Then again 15-20 seconds passes, I get exactly same thing, CPU usage raise to 80-90% and stays there for another 5-8 seconds, whereafter it goes back to normal. This is happening constantly, no matter what I do, no matter the chat activity.

I should mention that there is not a single file in download queue, no searches or dowloads going on, only a single upload with regular speed of 50-60 KiB/s, and ofcourse about 25 users in waiting list for upload. When I open CMD debug window without filter I see regular NMDC protocol messages from all hubs, both incoming and outgoing, there are most mixed searches and MyINFO's, nothing special, it's not like there are 50 search or CTM requests incoming at one moment every 15 seconds. Completely normal process of 30 opened hubs.

It takes me less than 1 minute to close down 1.2.0, replace the executables and open 1.1.0 with same settings, same hubs, same time, same data going through the client sockets. But 1.1.0 never has this problem with high CPU usage, there I have constant 0-5% CPU usage no matter what I do. 1.2.0 is evil.

I have managed to take a screenshot with exact thread and thread stack which constantly creates the huge CPU usage using Process Explorer, it's always same thread that is created, causes the CPU usage, and is destroyed again, as you can see on the screenshot attached below. There is another screenshot with graphical CPU usage history attached, so you can see that it happens constantly.

Please, what function in ApexDC++ 1.2.0 is responsible for mentioned thread? Let's find out, I'm ready to help as much as I can and I really want to know due which reason this behaviour is caused. Any answers appreciated. I'm using Windows XP Professional with final SP3 installed.

Thank you.

[Crise 26]

Are you using any plugins with 1.2 series?

[Crise 26]

Are you using any plugins with 1.2 series?

[RoLex 1]

Sorry, forgot to mention that. Yes, I'm using latest Lua and MediaPlayer plugins. Should I completely remove them from Plugins folder and restart ApexDC++ to see if it gets better? Scripts used by Lua plugin are all default, as it comes with LuaPlugin 2.0 release.

[Crise 26]

Sorry, forgot to mention that. Yes, I'm using latest Lua and MediaPlayer plugins. Should I completely remove them from Plugins folder and restart ApexDC++ to see if it gets better? Scripts used by Lua plugin are all default, as it comes with LuaPlugin 2.0 release.

Try removing the plugins, though if you just use them with default settings I see no reason for such extreme change in CPU usage...

[RoLex 1]

Did so, completely removed the DLL files, restarted the client and have been watching it over 10 minutes. Exactly the same behaviour, high CPU changes of doing nothing. When I open Process Explorer I can see same thread causing all the load, and exactly the same stack order appears. There in stack window I can see Win32Plugins::closeHubs function being called. What does this means, when is that function called? Maybe all this has to do with the amount of hubs I have open? There are 24 hubs where I'm operator and 4 hubs where I'm registered.

And about plugin settings.. Yes, all plugin settings are default.

Thank you.

[RoLex 1]

Good morning.

Today when I woke up the first thing I did was checking on ApexDC++ again, and guess what.. As you can see on the screenshot below there is no such extreme CPU change anymore, it works as it should be, 0-5% constantly. This gives me only one idea.. The amount of users in the hubs, now early in the morning there are 75% less users on hubs than there is in evening. So, in hope of ability to reproduce it you will have to "play" my situation. You have to login to atleast 30 hubs, 5 of those hubs have atleast 2000 users, the rest have from 50 to 800 users, 50/50. Please try to reproduce this "play". I'll try it too later in evening when there are more users on hubs.

Thank you.

[RoLex 1]

Yeah, I can reproduce this right now, exactly same behavoiur. Today there are even more users on the hubs, so I get higher hangs, even up to 100%. :)

[RoLex 1]

Bump. :)

[air 0]

i second that

i have 7 hubs online; 2 of them have 3500 to 5000 users 24/7, one - 800 to 1200, one - 300 to 600 and the last three 50 to 200. i have intel c2q 9300@2500 mhz and the usage graph looks like the first shown here. it even freezes the whole system for 50-300 millisecs from time to time (typical lag: opera or qip im input windows freezes so i can't see what i've typed in for some time). when i log off the bigger hubs, the graph looks like the second one from this thread and i have no lags

if i can get more info on this please tell what to do

[Solnce2 2]

I have problem 100% load CPU too in Apex 1.2.0. I have XP SP3, Athlon x2 5000@2500mhz. I open 13 hubs, all heve 1000-9000 users.

[User007 0]

I confirm that. Very high cpu load! but on 1.1.0 everything was fine so I returning on it.

Please fix it. Cpu load is VERY HIGH! even another application running slow.

Windows XP SP2. Apex 1.2.0+GeoIP+no plugins