Much higher CPU utilization in 2.42+ vs. 2.42?
Much higher CPU utilization in 2.42+ vs. 2.42?
Running Transmission on an embedded device (read: MIPSEL-based router), I've noticed that CPU utilization during moderate torrenting with 2.42 ranges anywhere from 35-90%, depending.
2.42+, however, with the exact same torrents (mid-stream, literally switching between the two engines), consistently uses pretty close to 100% CPU. This appears to be with or without PEX, DHT, LPD, and uTP, and the web interface being used or not.
Were there significant changes to the core that would account for this? I can't really see that the actual transfer performance of the app improved at all (it appears to be roughly the same), so I'm not quite sure what the spike in processor utilization is all about.
Both engines were built with the same SDK/compiler, the _only_ variance is the Transmission source itself.
Thoughts?
Rodney
2.42+, however, with the exact same torrents (mid-stream, literally switching between the two engines), consistently uses pretty close to 100% CPU. This appears to be with or without PEX, DHT, LPD, and uTP, and the web interface being used or not.
Were there significant changes to the core that would account for this? I can't really see that the actual transfer performance of the app improved at all (it appears to be roughly the same), so I'm not quite sure what the spike in processor utilization is all about.
Both engines were built with the same SDK/compiler, the _only_ variance is the Transmission source itself.
Thoughts?
Rodney
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
No verification. It's r13156 (latest SVN). I figured debugging code would be handled by a compile flag, no?
Rodney
Rodney
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Sure enough - NDEBUG is set for release builds only, it appears. The binary is still a LOT bigger than 2.42, but significantly smaller with NDEBUG enabled, and while CPU utilization is also still observably higher than 2.42, it's not pegged at 100% anymore. =)
Rodney
Rodney
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Hi,
I too am noticing that Transmission 2.42 is eating up all my CPU! It's running on DNS-323 and I only found out when my accessing files from was very slow unless I stop transmission. I was on 2.2 and I had no issues before I upgraded to 2.42.
I have:
transmission-remote 2.42 (13028)
I downloaded from transmission website, is there something I can do. Not sure how I can check for this debugging mentioned here?
Thanks,
I too am noticing that Transmission 2.42 is eating up all my CPU! It's running on DNS-323 and I only found out when my accessing files from was very slow unless I stop transmission. I was on 2.2 and I had no issues before I upgraded to 2.42.
I have:
transmission-remote 2.42 (13028)
I downloaded from transmission website, is there something I can do. Not sure how I can check for this debugging mentioned here?
Thanks,
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
X190...
All configuration, active connections and number of torrents are the same. In fact much less torrents.. more CPU.
I will try the link u sent, although it looks the same place I downloaded my version. The plus means no debugging enabled?
All configuration, active connections and number of torrents are the same. In fact much less torrents.. more CPU.
I will try the link u sent, although it looks the same place I downloaded my version. The plus means no debugging enabled?
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Also,
When I do top command, I see four Instances of transmission-daemon -g running?
When I do top command, I see four Instances of transmission-daemon -g running?
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
x190 is correct -- the nightly builds contain significantly more built-in correctness tests that try to smoke out problems before the stable release. Some of these are expensive in terms of CPU use, so the nightlies by design take up more CPU cycles than the official releases.rhester72 wrote:Running Transmission on an embedded device (read: MIPSEL-based router), I've noticed that CPU utilization during moderate torrenting with 2.42 ranges anywhere from 35-90%, depending.
2.42+, however, with the exact same torrents (mid-stream, literally switching between the two engines), consistently uses pretty close to 100% CPU. This appears to be with or without PEX, DHT, LPD, and uTP, and the web interface being used or not.
Were there significant changes to the core that would account for this?
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Some versions of top show threads as separate processes, and it's likely that's what's happening here. If so, it will appear in top that transmission is using 4x more memory than it actually is.ssss25 wrote:When I do top command, I see four Instances of transmission-daemon -g running?
On the other hand, maybe you've got four copies of transmission-daemon running for some reason. Try killing a few to find out

Re: Much higher CPU utilization in 2.42+ vs. 2.42?
But in fairness, aren't those "correctness tests" disabled with NDEBUG?Jordan wrote:x190 is correct -- the nightly builds contain significantly more built-in correctness tests that try to smoke out problems before the stable release. Some of these are expensive in terms of CPU use, so the nightlies by design take up more CPU cycles than the official releases.rhester72 wrote:Running Transmission on an embedded device (read: MIPSEL-based router), I've noticed that CPU utilization during moderate torrenting with 2.42 ranges anywhere from 35-90%, depending.
2.42+, however, with the exact same torrents (mid-stream, literally switching between the two engines), consistently uses pretty close to 100% CPU. This appears to be with or without PEX, DHT, LPD, and uTP, and the web interface being used or not.
Were there significant changes to the core that would account for this?
Rodney
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Even 2.42 release build uses 100% CPU for me.
Transmission 2.22 (the latest stable version in Gentoo portage tree) didn't exhibit such behaviour. The BitTorrent file I'm downloading is pretty large though (both in terms of absolute size and number of files).Re: Much higher CPU utilization in 2.42+ vs. 2.42?
As I mentioned, my 100% CPU results were coming from 13156, though I concur it does seem somehow linked to the size of the torrent (absolute file size, not # of files in the torrent).
Rodney
Rodney
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
All versions have problems with very large torrents (we* tested one of 1.2 TB composed of a large number of files).
The only solution was to use the daemon, its better than the GUI applications using about 50% of the resources used by the application.
* "We" means a user and me testing the Windows port of Transmission-Qt, and later the daemon running on Windows (actually 2 versions of the daemon, one built for Cygwin which has the better performance, and another build with MinGW).
The only solution was to use the daemon, its better than the GUI applications using about 50% of the resources used by the application.
* "We" means a user and me testing the Windows port of Transmission-Qt, and later the daemon running on Windows (actually 2 versions of the daemon, one built for Cygwin which has the better performance, and another build with MinGW).
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Correction: The solution was to use the daemon, and NOT use the Web client but instead use Transmission-Qt with a remote session.
The reason for not using the Web client was the number of files, since it was a big number and the Web client shows a flat list which makes it useless, things get even worse when the browser takes 100% CPU (on one core of a multi-core CPU, otherwise is browser+transmission trashing your CPU) when trying to show that file list.
The reason for not using the Web client was the number of files, since it was a big number and the Web client shows a flat list which makes it useless, things get even worse when the browser takes 100% CPU (on one core of a multi-core CPU, otherwise is browser+transmission trashing your CPU) when trying to show that file list.
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Thanks for all your replies..
Currently I have:
transmission-remote 2.42 (13028)
I didn't get any nightly build, I just got it fom the Trasmission main download site, and clicked on latest:
http://forum.dsmg600.info/t2719-%5BREL% ... ssion.html
Back then, I will get the most recent one, or try the link in this thread and report back.
Thanks,
Currently I have:
transmission-remote 2.42 (13028)
I didn't get any nightly build, I just got it fom the Trasmission main download site, and clicked on latest:
http://forum.dsmg600.info/t2719-%5BREL% ... ssion.html
Back then, I will get the most recent one, or try the link in this thread and report back.
Thanks,
Re: Much higher CPU utilization in 2.42+ vs. 2.42?
Ok, so whatever is posted at:
http://forum.dsmg600.info/t2719-%5BREL% ... ssion.html
is build 13028
Where do I get the famous r13113?
Thx.
http://forum.dsmg600.info/t2719-%5BREL% ... ssion.html
is build 13028
Where do I get the famous r13113?
Thx.