Bug#907429: neverball: Constant fsync calls seriously degrade performance
Am 28.11.2018 um 06:26 schrieb Ryan C. Gordon: > >> Cool, I know how to fix this now, thank you! > > This is now fixed in upstream PhysicsFS here for the 3.0 branch: > > https://hg.icculus.org/icculus/physfs/rev/c17f025e7a92 > > ...and here for the default (what will eventually become an unstable > 3.1.0) branch... > > https://hg.icculus.org/icculus/physfs/rev/8b3cc36531c6 > > PhysicsFS versions before 3.0.0 do not have this issue. > > This fix will be in 3.0.2, as a minor bugfix update, which is coming > sooner or later. Thank you very much. I have uploaded version 3.0.1-3 with your patch! :) > > My (untested) belief is that this will resolve the Neverball issue, > and we can probably close this bug report now. Thanks for tolerating > me abusing Debian's bug tracker like this! :) > > --ryan. You are welcome, but now I am on vacation for the rest of this year ;-) -- /* Mit freundlichem Gruß / With kind regards, Patrick Matthäi GNU/Linux Debian Developer Blog: http://www.linux-dev.org/ E-Mail: pmatth...@debian.org patr...@linux-dev.org */
Bug#907429: neverball: Constant fsync calls seriously degrade performance
Cool, I know how to fix this now, thank you! This is now fixed in upstream PhysicsFS here for the 3.0 branch: https://hg.icculus.org/icculus/physfs/rev/c17f025e7a92 ...and here for the default (what will eventually become an unstable 3.1.0) branch... https://hg.icculus.org/icculus/physfs/rev/8b3cc36531c6 PhysicsFS versions before 3.0.0 do not have this issue. This fix will be in 3.0.2, as a minor bugfix update, which is coming sooner or later. My (untested) belief is that this will resolve the Neverball issue, and we can probably close this bug report now. Thanks for tolerating me abusing Debian's bug tracker like this! :) --ryan.
Bug#907429: neverball: Constant fsync calls seriously degrade performance
I tried this and can confirm that removing setBuffer calls avoids the fsyncs. I verified this with strace too, which should be doable on any machine (just "strace -o /tmp/output ./neverball", and after playing check that /tmp/output contains no fsync calls). Cool, I know how to fix this now, thank you! --ryan.
Bug#907429: neverball: Constant fsync calls seriously degrade performance
On Tue, 16 Oct 2018 11:32:34 -0400 "Ryan C. Gordon" wrote: > > Can someone humor me and make a quick change to Neverball for me? > > In neverball/share/fs_physfs.c, there are three calls to > PHYSFS_setBuffer(). Just comment them out and rebuild Neverball with > PhysicsFS support and see if the problem goes away on the machine with > the slow disk. > > I suspect I know what the problem is (which would be a bug in PhysicsFS, > not Neverball, to be clear), but this will confirm it for me. I tried this and can confirm that removing setBuffer calls avoids the fsyncs. I verified this with strace too, which should be doable on any machine (just "strace -o /tmp/output ./neverball", and after playing check that /tmp/output contains no fsync calls).
Bug#907429: neverball: Constant fsync calls seriously degrade performance
Hi Ryan, On Tue, 16 Oct 2018 11:32:34 -0400 "Ryan C. Gordon" wrote: > > Can someone humor me and make a quick change to Neverball for me? > > In neverball/share/fs_physfs.c, there are three calls to > PHYSFS_setBuffer(). Just comment them out and rebuild Neverball with > PhysicsFS support and see if the problem goes away on the machine with > the slow disk. > > I suspect I know what the problem is (which would be a bug in PhysicsFS, > not Neverball, to be clear), but this will confirm it for me. > > Thanks! > > --ryan. I'm adding the original bug submitter to CC who is hopefully be able to respond to this request. Bug reporters are not automatically subscribed to bug reports. Like I previously wrote I don't have such a slow disk anymore (I own some 20 year old laptops but none of them is capable of running Neverball). I could upload a new neverball package to Debian experimental if that helps to debug the issue though. Regards, Markus signature.asc Description: OpenPGP digital signature
Bug#907429: neverball: Constant fsync calls seriously degrade performance
This is not actionable by Debian itself. (But if you don't mind, we'll organize on this bug report briefly to get this fixed upstream. Apologies in advance for a few emails worth of extra chatter!) --ryan.
Bug#907429: neverball: Constant fsync calls seriously degrade performance
Can someone humor me and make a quick change to Neverball for me? In neverball/share/fs_physfs.c, there are three calls to PHYSFS_setBuffer(). Just comment them out and rebuild Neverball with PhysicsFS support and see if the problem goes away on the machine with the slow disk. I suspect I know what the problem is (which would be a bug in PhysicsFS, not Neverball, to be clear), but this will confirm it for me. Thanks! --ryan.