i noticed that SVN Version now plays a lot slower that plain vanilla 2.5.2, with identical settings.
so i compared the same (online multiplayer mentalspace) demofile in both versions:


SVN-Version (16.-18.Nov)

Code: Select all

result 6853 frames 109.5050000 sec 62.5816 fps,  fps min/avg/max: 13 71 190 (315 sec)
result 6853 frames 125.7910000 sec 54.4792 fps,  fps min/avg/max: 22 61 195 (315 sec)
result 6853 frames 96.8190000 sec 70.78156 fps,  fps min/avg/max: 42 77 193 (315 sec)
result 6853 frames 97.2600000 sec 70.46062 fps,  fps min/avg/max: 42 77 193 (315 sec)
result 6853 frames 128.3960000 sec 53.3739 fps,  fps min/avg/max: 32 58 164 (315 sec)
result 6853 frames 98.2880000 sec 69.72362 fps,  fps min/avg/max: 42 76 193 (315 sec)
result 6853 frames 98.4400000 sec 69.61608 fps,  fps min/avg/max: 42 76 195 (315 sec)
result 6853 frames 98.6650000 sec 69.45723 fps,  fps min/avg/max: 42 75 189 (315 sec)


vanilla version 2.5.2

Code: Select all

result 6853 frames 66.6160000 sec 102.873183 fps,  fps min/avg/max: 59 112 239 (315 sec)
result 6853 frames 69.8840000 sec 98.0625036 fps,  fps min/avg/max: 50 107 218 (315 sec)
result 6853 frames 68.3060000 sec 100.327936 fps,  fps min/avg/max: 56 109 239 (315 sec)
result 6853 frames 74.2140000 sec 92.3410677 fps,  fps min/avg/max: 50 100 214 (315 sec)
result 6853 frames 73.7920000 sec 92.8691457 fps,  fps min/avg/max: 56 101 232 (315 sec)
result 6853 frames 73.7660000 sec 92.9018789 fps,  fps min/avg/max: 57 101 230 (315 sec)

(the very different values within a version might be the first-load and/or external reasons).
but still you see the huge overall difference, for me about 20 FPS!

my settings are low, no gibs or such neat effects ^^
(cleaned config.cfg :http://pastebin.com/m34f595c)


Is this the price for implementing ODE?

I would vantage to guess this is related to collision handling, although I could be wrong. Please try dp revision 9398 vs 2.5.2 vs dp revision 9406 engine for me.

thanx for your answer, that might be the case.

now i dont know how to change version to the ones you mentioned, please tell.
i am using your 'Nexuiz Build System v0.1 - Quick Build'.


there is also some staying 'side effect', meaning jerky (visible opponents) player movement, like something continues to use much processor time or bandwidth.
(i currently dont have a clue about that..)

yes, i am quite sure i already ruled out external reasons like suddenly starting Automatic Updates, Scheduled Backup jobs, virii, trojans or such.

Samual,
seriously, i looked into your nice batch files fo Compiling Nexuiz on Windows, but i dont see a point where i could drop in a 'SVN UP ####', i dont understand that code quite much.
(only see references to 'call getsvn svn://' but that refers to downloading [how, which version?] and some 'make' stuff'..)
I dont see how to fiddle with svn at all, no GUI, the nested folder structures, make, etc. pp.


so please help me, maybe with a new system.bat with version control option?


Can't be that am i the only one with such performance drops!?


PS: for me personally the 'fix' is using plain original 2.5.2 version, but i really liked to try SVN version and scream bug! at you..

thanx to Spaceman (who built it for me) i was able to test the forementioned builds:

Code: Select all

enginedate 02:52:53 Oct  1 2009 9271 release
result 6853 frames  73.7660000s 92.9018789 fps, [..] min/avg/max: 57 101 230 (315s)
enginedate 12:38:51 Nov 20 2009 9398 release
result 6853 frames  69.4940000s 98.6128299 fps, [..] min/avg/max: 55 108 232 (315s)
enginedate 13:04:38 Nov 20 2009 9406 release
result 6853 frames  69.1420000s 99.1148651 fps, [..] min/avg/max: 57 108 235 (315s)
enginedate 13:29:54 Nov 20 2009 9450 release
result 6853 frames 72.1210000s 95.0208677 fps, [..]  min/avg/max: 55 104 219 (315s)
enginedate 12:56:00 Nov 20 2009 9507 release
result 6853 frames 107.4890000s 63.7553610 fps, [..] min/avg/max: 32 70 235 (315s)


first one is release version, last one latest svn, in between the ones you mentioned, which work without flaws as you can see.

Some Windows builds cross compiled on Linux

Darkplaces_r9398

Darkplaces_r9406

Darkplaces_r9450

Darkplaces_r9507

fronten wrote:thanx to Spaceman (who built it for me) i was able to test the forementioned builds:

Code: Select all

enginedate 02:52:53 Oct  1 2009 9271 release
result 6853 frames  73.7660000s 92.9018789 fps, [..] min/avg/max: 57 101 230 (315s)
enginedate 12:38:51 Nov 20 2009 9398 release
result 6853 frames  69.4940000s 98.6128299 fps, [..] min/avg/max: 55 108 232 (315s)
enginedate 13:04:38 Nov 20 2009 9406 release
result 6853 frames  69.1420000s 99.1148651 fps, [..] min/avg/max: 57 108 235 (315s)
enginedate 13:29:54 Nov 20 2009 9450 release
result 6853 frames 72.1210000s 95.0208677 fps, [..]  min/avg/max: 55 104 219 (315s)
enginedate 12:56:00 Nov 20 2009 9507 release
result 6853 frames 107.4890000s 63.7553610 fps, [..] min/avg/max: 32 70 235 (315s)


first one is release version, last one latest svn, in between the ones you mentioned, which work without flaws as you can see.

Meh, that could be anything (After 9450)........ Lots and lots of changes after that which I don't understand in the svn log... So LordHavoc will have to sort this one out.

By the way, my build script release that I gave you (v3 I presume) does not support revision picking, that is something I added later -- But well, I haven't taken the time to release the new version as i'm rather busy and it seems to suffice for now.

Meh, you might say that!

still it would be nice if you would release the updated build system (isnt it mainly the batch files changed?) as i really don't know how to use SVN or that tortoise SVN client to work with fteqc, darkplaces, nexuiz repos (tried; failed;).
i am willing to test/hunt this bug to a certain version change, if i only knew how^^

Code: Select all

enginedate 20:20:48 Nov 20 2009 9404 release
result 6853 frames 69.5230000s 98.5716957 fps, [..] min/avg/max: 55 108 233 (315s)

enginedate 13:29:54 Nov 20 2009 9450 release
result 6853 frames 72.1210000s 95.0208677 fps, [..] min/avg/max: 55 104 219 (315s)

denginedate 21:16:51 Nov 20 2009 9485 release
result 6853 frames 69.8030000s 98.1762961 fps, [..] min/avg/max: 55 107 226 (315s
)
enginedate 21:47:58 Nov 20 2009 9486 release
result 6853 frames 67.028000s 102.2408546 fps, [..] min/avg/max: 58 112 253 (315s)

enginedate 22:22:13 Nov 20 2009 9487 release
result 6853 frames 84.6660000s 80.9415822 fps, [..] min/avg/max: 48 88 240 (315s)
enginedate 22:22:13 Nov 20 2009 9487 release
result 6853 frames 73.4900000s 93.2507824 fps, [..] min/avg/max: 53 101 237 (315s)
enginedate 22:22:13 Nov 20 2009 9487 release
result 6853 frames 73.4800000s 93.2634731 fps, [..] min/avg/max: 56 101 237 (315s)
enginedate 22:22:13 Nov 20 2009 9487 release
result 1551 frames 17.3400000s 89.4463668 fps, [..] min/avg/max: 56 101 240 (70s)
enginedate 22:22:13 Nov 20 2009 9487 release
result 6853 frames 73.2840000s 93.5129087 fps, [..] min/avg/max: 57 102 237 (315s)

enginedate 22:08:35 Nov 20 2009 9488 release
result 6853 frames 105.1560000s 65.1698429 fps, [..] min/avg/max: 38 71 240 (315s)
enginedate 22:08:35 Nov 20 2009 9488 release
result 6853 frames 73.3120000s 93.4771934 fps, [..] min/avg/max: 57 101 237 (315s)
enginedate 22:08:35 Nov 20 2009 9488 release
result 6853 frames 105.6580000s  64.8602094 fps, [..] min/avg/max: 37 71 237 (315s)
enginedate 22:08:35 Nov 20 2009 9488 release
result 6853 frames 73.5490000s  93.1759779 fps, [..] min/avg/max: 55 101 240 (315s)
enginedate 22:08:35 Nov 20 2009 9488 release
result 6853 frames 73.6510000s  93.0469376 fps, [..] min/avg/max: 56 101 237 (315s)

enginedate 22:35:09 Nov 20 2009 9489 release
result 6853 frames 105.1220000s 65.1909210 fps, [..] min/avg/max: 37 71 240 (315s)
enginedate 22:35:09 Nov 20 2009 9489 release
result 6853 frames 105.080000s 65.2169775 fps, [..] min/avg/max: 36 72 240 (315s)
enginedate 22:35:09 Nov 20 2009 9489 release
result 6853 frames 79.1330000s  86.6010388 fps, [..] min/avg/max: 7 100 237 (315s)
enginedate 22:35:09 Nov 20 2009 9489 release
result 6853 frames 74.1230000s  92.4544338 fps, [..] min/avg/max: 56 100 230 (315s)

enginedate 21:53:10 Nov 20 2009 9490 release
result 6853 frames 109.0810000s 62.8248733 fps, [..] min/avg/max: 36 69 228 (315s)
enginedate 21:53:10 Nov 20 2009 9490 release
result 6853 frames 104.3850000s  65.6511951 fps, [..] min/avg/max: 15 74 230 (315s)
enginedate 21:53:10 Nov 20 2009 9490 release
result 6853 frames 73.2480000s  93.5588685 fps, [..] min/avg/max: 55 102 240 (315s)

enginedate 22:02:46 Nov 20 2009 9491 release
result 6853 frames 109.3240000s 62.6852292 fps, [..] min/avg/max: 36 68 225 (315s)
enginedate 22:02:46 Nov 20 2009 9491 release
result 6853 frames 104.7710000s  65.4093213 fps, [..] min/avg/max: 14 74 240 (315s)
enginedate 22:02:46 Nov 20 2009 9491 release
result 6853 frames 73.3340000s  93.4491505 fps, [..] min/avg/max: 55 101 240 (315s)

enginedate 21:37:21 Nov 20 2009 9497 release
result 6853 frames 108.8350000s 62.9668765 fps, [..] min/avg/max: 36 69 228 (315s)
enginedate 21:37:21 Nov 20 2009 9497 release
result 6853 frames 105.2010000s  65.1419663 fps, [..] min/avg/max: 37 71 235 (315s)
enginedate 21:37:21 Nov 20 2009 9497 release
result 6853 frames 73.3500000s  93.4287662 fps, [..] min/avg/max: 57 101 242 (315s)

enginedate 12:56:00 Nov 20 2009 9507 release
result 6853 frames 105.6170000s  64.8853878 fps, [..] min/avg/max: 38 71 235 (315s)
enginedate 12:56:00 Nov 20 2009 9507 release
result 6853 frames 73.4970000s  93.2419010 fps, [..] min/avg/max: 57 101 237 (315s)

now i got dissappointed and very upset.
timedemo just doesnt seem to be very reliable.

for the last times i tested like this:
started Nexuiz, chose demo file, played it, broke, timedemo'd it.

but as it seems the timedemo results change a lot if you run a demo repeatedly.
memory / cache issue?
within these results i do NOT find the bug i was intending to find.
a minimum of above 55 fps is considered good.

for now i cannot call this problem a bug or found whatsoever.
i was able to repeat the buggy action, but not in any case.

still there remains one big thing with all this:

playing a game live online cannot be compared to replaying a demo.

so how does bad caching (if this is) inflict online game play?
in one word: badly.

a 'fix' like for demos, 'just load it again' ist not possible for online game play.

i remain speechless.
this is a problem i noticed (quite heavily) when playing online - huge framerate drops - it is like watching a demo a first, or second time. at third time, most probably, it will be fine.
but as said, again, online game play is not this.


right now it stays:
with 2.5.2 version all is more or less fine, playable.

2 days ago i got rid of my super messy nexuiz-svn install (messy by my fault). I downloaded a vanilla 2.5.2 binary version from the nexuiz homepage. I must admit that it was a very bad version for me. water is fucked up and performance in general is bad. Upgrading with the nov 21st patch fixed all this and made the game go from "not cool at all" to "awesome, like it has ever been".

thanx for reply, liolak.
(because of the forementioned problems i stopped using SVN.)

unfortunately my problem remains unsolved until now (r9530).
i have tried locating the bug, maybe did, maybe not.

i would feel glad if one dev could give an answer to how to correctly find the bug.
(or how timedemo works concerning loading stuff, averaging framerate, and in what it differs from online game play) i would like to find the bad mark and erase it, - but please, please, with cream on top, give me means to do so.

obviously the method of using timedemo is unreliable.
so what next?

yep i stopped using svn as a "daily player" too because from one build to another the game feeling changes. of course this is how its supposed to be since its the development version. I do indeed like a repetable experience for playing , and svn is for testing only xD
I am trying to find what is the revision number of the patch im using atm...(dont know how to find it ), but as soon as i find it, I will tell you so that you can enjoy playing again. but im pretty sure an official patch is on the way

liolak,
in console type 'ver' to get the revision.
(or look into the benchmark.log files, if you have run a timedemo)


i am happy playing vanilla 2.5.2 for now, but i fear a now major release without this bug found and fixed. this missing 20 fps are vital for me.

no suggestions?

Something quite similar was brought up in the dev channel some days ago, it appears that the problem is likely to be a change to the sound handling - which lowers memory usage overall.

Problem with timedemos is that the action is fast forwarded, and the sound handlers are bombarded with more effects than they can handle - causing a significant performance drop. A fix for this is somewhere in the pipeline, but it remains unclear when this will be finalized.

For now can you please retest some of the bad revisions of darkplaces, but starting them with the -nosound flag. Hopefully this will bring back better fps on timedemos.

As a side note, this bug ONLY affects timedemos, regular playing or playing back demos at normal speed should not be affected.

Code: Select all

enginedate 13:19:13 Dec  4 2009 9534
commandline ..\darkplaces\darkplaces-sdl -nexuiz -mygames -condebug +cvar_check_default bypass
result 6853 frames 129.0840000s 53.0894611 fps, [..] min/avg/max: 25 59 185 (315s)
result 6853 frames 126.7010000s 54.0879709 fps, [..] min/avg/max: 11 61 180 (315s)

enginedate 13:19:13 Dec  4 2009 9534
commandline ..\darkplaces\darkplaces-sdl -nexuiz -mygames -condebug -nosound +cvar_check_default bypass
result 6853 frames 75.8750000s 90.3196046 fps, [..] min/avg/max: 51 100 231 (315s)
result 6853 frames 75.5030000s 90.7646054 fps, [..] min/avg/max: 50 100 231 (315s)



for the timedemos it seems to work without sound.
i have yet to play it live, - online with -nosound though..

for me it seems that for live play (online) it all is buggy as before.

even worse (and that made it hard to test) the -nosound parameter messes with my mouse control *wtf* but the low fps and sudden jerks are still the same regardless of nosound and mouse.

so whilest timedemo seems fixed it remains for normal play.


(liolak: you type 'version' not just 'ver' to get revision, sorry.)

what do you suggest for the problem with normal gameplay?

you know, it is really hard to test with no reliable and verifiable, repeatable base of comparism (what the timedemo was thought for).

i mean, i can try to play for myself with the three different versions (vanilla 2.5.2, SVN+/-Sound) , and then judge by my experience, if the behaviour occurs or not, but this is sometimes hard to judge, lags happen, framerate drops all the time, anytime. and even worse, it is no fun playing the retarded (sorry) versions at all, with 20 frames (drops, average) less, fucked senses, and (maybe) no sound..


no fun makes me no play ;-/

hello?
anyone there at the other end of the pipe?

can you please give a hint or such on how to fix or, at least, how to trial this bug?

any suggestion?

fronten wrote:but as it seems the timedemo results change a lot if you run a demo repeatedly.
memory / cache issue?
within these results i do NOT find the bug i was intending to find.
a minimum of above 55 fps is considered good.

still there remains one big thing with all this:

playing a game live online cannot be compared to replaying a demo.

so how does bad caching (if this is) inflict online game play?
in one word: badly.

a 'fix' like for demos, 'just load it again' ist not possible for online game play.

[01:44] <Willis> I was playing around, and no matter what settings (playing with snd_streaming and r_font_hinting) on the second run, performance was fine, meaning even if it is using more cpu, it's not enough that it's cpu bound
[01:54] <LordHavoc> Willis: I've been really tempted to make DP somehow do an automatic twice-run timedemo, reporting time for only the second
[01:55] <Diablo-D3> why?
[01:55] <LordHavoc> because the engine uploads and compiles shaders on first use
[01:55] <LordHavoc> it also uploads lightmaps in q1bsp on first use

[01:44] <Willis> I was playing around, and no matter what settings (playing with snd_streaming and r_font_hinting) on the second run, performance was fine, meaning even if it is using more cpu, it's not enough that it's cpu bound
[01:54] <LordHavoc> Willis: I've been really tempted to make DP somehow do an automatic twice-run timedemo, reporting time for only the second
[01:55] <Diablo-D3> why?
[01:55] <LordHavoc> because the engine uploads and compiles shaders on first use
[01:55] <LordHavoc> it also uploads lightmaps in q1bsp on first use

vid_restart causes the shaders and lightmaps to be flushed from this cache, hence if you do vid_restart between timedemos, you get the same result as if you restart.

Would it be worth using Phoromatic?
http://www.phoromatic.com/
That way an automated build of each DP engine code revision could be benched via the Nexuiz benchmark in PTS. We'd then have an official measure for performance rather than occasional claims for drops in performance which we can't necessarily depend upon.

This same approach is being done by Phoronix to track performance in the kernel Git repository which should bring payoffs for the whole Linux community.

besides anything my SVN now looks like that :

(http://omploader.org/vMzZicQ/nexuiz000005.jpg)

even with *cleanup* and completely new download 'n' build.

yeah, i am fucked.

fronten wrote:besides anything my SVN now looks like that :

Please type this when that problem occurs:
condump whitesurfaces.log

Then upload that somewhere, I'll look at the shader compile errors.

fronten wrote:besides anything my SVN now looks like that :

(http://omploader.org/vMzZicQ/nexuiz000005.jpg)

even with *cleanup* and completely new download 'n' build.

yeah, i am fucked.

I have the same problem and caused by GLSL. If I disable GLSL, the problem is gone. But I can't turn off GLSL forever, it will always reset after entering the game. I have once filed a ticket.

Can you paste a copy of your game console here, this will give the developer(s) some clues why GLSL is failing.

Open your game console and enter condump fle_name.txt open fle_name.txt with your favourite text editor and copy and the paste the contents here.

Knowing what video card and operating system is used would help also.

It is known to work on NVIDIA and ATI fglrx drivers on Linux, it is known to work on NVIDIA and ATI drivers on Windows, I am less certain about Mac OS X drivers, and Linux Mesa DRI drivers.

We really need that condump output.

for the last months i dropped the whole thing because of no clues whatsoever.

now, i replaced my old Ati Radeon 9200 with an Radeon HD 4850, and ran this tests (SVN vs vanilla 2.5.2) again.

this time, demo002 with ULTRA settings, resulted in about 55-60 FPS for standard 2.5.2 and ~40 FPS for latest SVN release..
SVN did much more 'stutter', meaning partly running fast through the demo, but then going very delayed and such.

the new card got more RAM, the Shader Model 2 now works, and so on, still the (bad) differences stay.