Wednesday, 2007-05-02

*** cgTobi has quit IRC00:02
*** Pseudonym has joined #Aqsis00:58
*** tuga3d has joined #aqsis02:12
*** cgTobi has joined #Aqsis05:15
*** ChanServ sets mode: +o cgTobi05:15
PseudonymG'day.05:15
cgTobihi Andrew05:28
*** cgTobi sets mode: +vv Pseudonym tuga3d05:28
*** Alex_K has joined #aqsis05:33
*** tuga3d has quit IRC05:58
cgTobiI need to shoot06:10
cgTobioff to germany06:10
cgTobicya06:10
*** cgTobi has quit IRC06:10
*** Alex_K has quit IRC06:49
*** Anteru has joined #aqsis07:58
*** pgregory has joined #aqsis07:58
*** ChanServ sets mode: +o pgregory07:59
*** render has joined #aqsis08:03
pgregorymorning all08:06
*** pgregory sets mode: +vv Anteru render08:06
*** Anteru has quit IRC08:08
*** Alex_K has joined #aqsis08:08
PseudonymG'day.08:10
pgregoryhi Andrew08:10
*** joeedh is now known as joeedh_sleep08:26
PseudonymGotta go.  Night.08:30
*** Pseudonym has quit IRC08:30
*** renderguy has joined #aqsis09:23
renderguyMorning all.09:23
*** pgregory sets mode: +v renderguy09:29
*** pgregory sets mode: +v Alex_K09:29
*** Anteru has joined #aqsis10:03
AnteruMorning10:03
*** pgregory sets mode: +v Anteru10:04
renderguyAnteru: Moin.10:04
AnteruWhat's up?10:07
renderguyAnteru: Nothing much.10:09
Anteruhm10:11
AnteruThis secure STL stuff bothers me :/10:12
*** Anteru2056 has joined #aqsis10:14
*** Anteru has quit IRC10:14
*** Anteru2056 is now known as Anteru10:14
Anterure10:15
Anterunext try to disable that warning10:33
*** ShortWave has quit IRC10:35
*** ShortWave has joined #aqsis10:36
AnteruI don't get it10:37
Anterumy first line in aqsis.cpp is #define _SECURE_SCL 010:38
Anteruthe STL has: #if !defined (_SECURE_SCL)10:38
Anteru#define _SECURE_SCL 110:38
Anteru#endif10:38
AnteruI can click on "ignore" but it stops in that assertion anyway10:41
Anteruand the docs here http://msdn2.microsoft.com/en-us/library/aa985896(VS.80).aspx cleary state that I should have them disabled10:42
AnteruAh10:45
AnteruI've built boost with _SECURE_STL on :/10:46
AnteruWhich is the default10:46
AnteruSo the workaround is to rebuild boost with _SECURE_STL off10:46
Anterucatch you later10:52
*** Anteru has quit IRC10:52
*** Anteru has joined #aqsis11:07
*** Joron has joined #Aqsis11:11
*** Anteru has quit IRC11:24
*** Joron has quit IRC11:35
*** Anteru has joined #aqsis11:44
*** Anteru has quit IRC11:51
*** Anteru has joined #aqsis12:10
*** Anteru has quit IRC12:38
*** Alex_K has quit IRC13:10
*** AlexK has joined #aqsis13:25
*** mafm has joined #aqsis13:28
pgregoryhi mafm13:30
*** pgregory sets mode: +vvv ShortWave AlexK mafm13:30
mafmhello13:31
pgregoryhow goes it?13:41
mafmquite busy with school projects and so on :)13:42
mafmI hope to finish most of them to have the summer free13:42
mafmanything new around here?13:44
pgregoryvarious interesting things.13:47
pgregoryI'm working on a new FB13:47
pgregoryA new developer has stepped up to implement point cloud/brickmap support.13:47
pgregoryLots of interesting discussion about the other SoC project (Deep Shadows).13:48
mafmgoody :)13:50
*** Anteru has joined #aqsis14:02
mafmhi anteru14:06
Anteruhi mafm14:15
*** Sh0rtWave has joined #aqsis14:31
*** Anteru has quit IRC14:45
*** c42f has joined #aqsis15:17
pgregoryneed to reboot, brb15:17
*** pgregory has left #aqsis15:17
*** pgregory has joined #aqsis15:24
*** ChanServ sets mode: +o pgregory15:25
pgregoryback15:28
c42fHi Paul15:28
pgregoryhi Chris15:28
c42fpgregory: do we do any kind of anisotropic filtering ATM?  As far as I can tell the SampleMap() function(s) don't really do it.15:29
pgregoryno15:29
c42fOk, that's what I gathered.  Do you think it's worth investigating?15:31
pgregorysure15:31
pgregoryif you can identify a valid case where it would provide a significant advantage, then it's bound to be worth it.15:32
c42fCool.  I'll have to understand it a bit better first...15:32
c42fAlso, what do you think about the bilinear vs trilinear defaults?  I'm quite strongly in favour of making trilinear the default.  Bilinear filtering often seems to cause some nasty artefacts.15:33
pgregorygets my vote, as long as it's not an order of magnitude slower, quality is always primary.15:35
pgregoryof course speed is a very close second ;-)15:35
pgregorygotta go, maybe back later, cya15:38
*** pgregory has left #aqsis15:38
c42fAh damn, hope to see you later Paul ;-)15:38
c42f(damn that Paul went before I could distract him with further questions that is)15:49
mafm:D15:56
*** AlexK has quit IRC16:19
c42fHmm, with a simple scene containing only a textured linear patch (ie, should be heavily texture-limited if anything is), bilinear filtering took 2m26s, while trilinear took 3m8s.  That's only a ~30% speed increase, for double the number of texture accesses.16:24
c42fs/double/half/16:25
* c42f certinally doesn't consider this an "order of magnitude" ...16:25
mafman order of magnitude would be like 20min16:28
c42fExactly.16:29
c42fBetter that we get something correct here I feel, rather than a bodgy job at only 30% speed increase...16:30
*** Anteru has joined #aqsis16:34
mafmI think that too16:35
renderguyc42f: Hi Chris... next step would be (on-going) optimisation anyway.  ;-)16:38
renderguyc42f: Weren't you thinking of changing the default texture filter to 'sinc' rather than 'box' too?16:42
c42fYes, I was thinking of doing that for mipmap generation.16:43
renderguyc42f: I'm assuming there's little/no penalty in doing that either?16:44
c42fThe new mipmap generation is _at least_ an order of magnitude faster than the old (really), so speed shouldn't be an issue.16:44
renderguyc42f: 8-)16:44
c42fI can run an 8x8 sinc filter over an 800x800 image in ~1sec.16:44
c42fNot blindingly fast, but certinally usable.16:44
c42fThe old code took ages to do that - at least a minute, say.16:45
renderguyc42f: Seems you've really been getting into the texturing stuff, with good results, congrats.16:45
c42fWell, thanks :-)  You should see some nicer results soon, when I've finally fixed the trilinear filtering properly.  (I have a solution ATM, but it's a nasty hack ;-)16:46
renderguyc42f: Still being influenced by 'The Dark Side' (Pixie)?  ;-)16:48
c42fI'd *really* like to know how 3delight does it's gaussing filtering without stochastic sampling...16:48
c42frenderguy: The pixie codebase will be a useful source of ideas at times I think.16:49
renderguyc42f: I think we'd like to know how 3Delight does a lot of things.  :-D16:49
c42fI'd like to think that16:49
c42foops16:49
c42fI'd like to think that the open source world represents a whole pool of solutions, and people can mix and match from here and there to get inspiration.16:50
c42fIt's hard not to think of Pixie as "the competition" sometimes though ;-)16:50
c42fHopefully they can get inspiration from parts of our codebase at times.16:51
renderguyc42f: I'm sure they have, as have others.16:51
c42fand regarding 3delight, yep it'd be very nice to peek at thier code twice in a while :-)16:52
renderguyc42f: Perhaps we need to get a sleeper in there.16:53
* renderguy goes covert.16:53
c42fhehe16:53
renderguyI've been getting into Paul's eqshibit (new FB) work, it's already my prefered FB!16:55
renderguy... it's seqsy.  :-)16:56
c42f:-)16:56
* renderguy makes note of that name for potential use in future.16:57
c42fAll these words with q's in them are so darn slow to type :-(16:59
*** Anteru has quit IRC16:59
c42faqsl aqsis eqshibit teqser... Argh!16:59
c42f;-)16:59
renderguyc42f: I still trip-up on eqshibit occasionally, but getting use to it.16:59
c42fWhen I first saw that I couldn't help reading it as eqsh*t all the time :-/  I think I've got used to it now though.17:01
renderguyc42f: Maybe we'll drop the 'h', so it scans better.17:01
c42fyeah, actually eqsibit seems much nicer to me17:02
c42fmuch more obvious what it's meant to represent phonetically.17:02
renderguyc42f: That said, if it changes I'll spend a while getting used to typing the *right* (new) name.  ;-)17:06
renderguyc42f: BTW... would you be interested in (simple) documenting your texture/filter work on our blog?17:09
renderguyc42f: http://aqsis.blogspot.com17:09
c42fSeems like for bilinear interpolation, the artefacts can be made much less objectionable if the mipmap level is chosen more appropriately...17:09
* c42f has just been playing.17:10
c42fOh, I didn't know we had a blog.17:10
* c42f goes to look17:10
renderguyc42f: Not many do, as it's still a WIP.17:10
renderguyc42f: I need to update my posts, but other such contributions by the team *are* welcome... to show what goes-on behind the scenes.17:11
* renderguy next update will regard eqshibit.17:12
c42fSounds cool.  I'd have to find time to write it though.17:14
c42fI guess the tone is light.  I wrote rather extensively on my earlier tests with mipmapping on the wiki.  That was rather detailed though.17:14
c42fs/tone is light/tone should be light/17:15
c42fAnyway, I'd better go.17:15
renderguyc42f: Yeah... simple/light would be the goal here.17:15
renderguyc42f: K, cya.17:15
*** c42f has quit IRC17:15
renderguyGotta shoot too, laterz.17:19
*** renderguy has left #aqsis17:22
*** AlexK has joined #aqsis17:23
*** Joron has joined #Aqsis17:33
*** Anteru has joined #aqsis17:44
Anteruhoi17:46
*** mafm has quit IRC18:16
*** AlexK has quit IRC18:42
*** pgregory has joined #aqsis20:54
*** ChanServ sets mode: +o pgregory20:54
pgregoryevening all21:01
*** pgregory sets mode: +vv Anteru Sh0rtWave21:02
*** joeedh_sleep is now known as joeedh21:38
|rt|hey guys21:39
|rt|hey joeedh how is DSM going?21:39
joeedhSoC hasn't started yet.  so I'm doing another project in the meantime.22:00
AnteruGood evening Paul22:03
*** Anteru has quit IRC22:13
*** pgregory has left #aqsis22:29
*** Sh0rtWave has quit IRC22:45
*** joeedh is now known as joeedh_class22:58
*** render has quit IRC23:32

Generated by irclog2html.py 2.5 by Marius Gedminas - find it at mg.pov.lt!