Re: [Algorithms] Per-pixel (normapped) lighting of environments on lower-end HW
Brought to you by:
vexxed72
From: Emanuele S. <in...@fo...> - 2004-05-29 14:34:06
|
Can't remember in which Nvidia slides I've seen this...but they recomend = a batch size between 2000 and 10000 triangles...GPU tends to perform to = its maximum polygon throughput. If your polygon batches are below = 800/600 triangles you get poor performances. So, mapping does matter...and a lot, IMHO. As well as number of = materials/object. Every time you get discontinuous UV mapping you break your triangle = list, or vertex array or whatever else you use.... thus reducing the = size of a batch. As said before, best performances with 1 material/object -> 1 object =3D = 1 batch. This of course depends very much from your graphics workflow...but you = can optimize in several ways... I'd suggest beginning with getting = continuous UV mapping. ...I remember our artists faces when we exported our geometry into the = engine and noticing the number of vertices was triplicated... (that was = obviously due to "not careful mapping"....and use many tiny textures - = materials - on a few polygons....that's pretty bad!) Best, Emanuele Salvucci Maya|Lightwave Game Technical Artist Lscript developer www.forwardgames.com ema...@fo... "#3. He causes things to look different so it would appear time has = passed." -------------------------------------------------------------------------= ------- ----- Original Message -----=20 From: Andrey Iones=20 To: gda...@li...=20 Sent: Friday, May 28, 2004 9:43 PM Subject: Re: [Algorithms] Per-pixel (normapped) lighting of = environments on lower-end HW I think our biggest concern is batch-limiting. For some reason we find = it really hard to stay=20 near the recommended 500-700 batches - and that's why we're so much = concerned=20 with fitting everything in one pass. I don't know whether it's our = texturing (probably not though) or=20 what... But in the detailed scenes with numerous dynamic objects = (physics controlled, or simply animated)=20 we are often staying over 1000-1300 batches, and we don't want to go = even higher.=20 Andrey.=20 Gary McTaggart wrote:=20 =20 Why is the goal single-pass? Are you batch (driver) or fillrate = limited? If not, then I would release this restriction. Gary -------------------------------------------------------------------------= - From: gda...@li... = [mailto:gda...@li...] On Behalf Of = Andrey Iones=20 Sent: Friday, May 28, 2004 9:43 AM=20 To: gda...@li...=20 Subject: Re: [Algorithms] Per-pixel (normapped) lighting of = environments on lower-end HW=20 =20 Not really - the goal is definitely to come up with a single-pass = solution which would enable=20 per-pixel lighting at least for diffuse effects (diffuse + albedo = in one pass). It should not also=20 require tesselation (i.e. light directions / colors should be = stored for the most part in lightmap=20 texels rather then in colors). It should also work on GF3-style = hardware....=20 Andrey.=20 Tom Forsyth wrote:=20 If your alternative is another pass (that needs lighting all = over again),=20 I'd say yeah, it's trivial.=20 TomF.=20 > -----Original Message-----=20 > From: gda...@li...=20 > [mailto:gda...@li...] On=20 > Behalf Of Jon Watte=20 > Sent: 27 May 2004 11:22=20 > To: gda...@li...=20 > Subject: RE: [Algorithms] Per-pixel (normapped) lighting of=20 > environments on lower-end HW=20 >=20 >=20 >=20 > > Who would even notice another 1000 tris in their scenes?=20 >=20 > Aaah, to be living in the future!=20 >=20 > Meanwhile, on planet earth, CPU-transform solutions sit in 50% = of all=20 > machines currently sold (and the hardware-transform = capabilities of=20 > maybe another 10% might be in question, truth be told).=20 >=20 > With a current scene load of 50,000 tris per frame, another = 1,000=20 > tris would matter. Another 1,000 tris PER STREET per frame = would be=20 > even worse -- because, in reality, your scene isn't just a = single=20 > 50x10 meter street ;-)=20 >=20 > Cheers,=20 >=20 > / h+=20 >=20 >=20 >=20 > -------------------------------------------------------=20 > This SF.Net email is sponsored by: Oracle 10g=20 > Get certified on the hottest thing ever to hit the market...=20 > Oracle 10g.=20 > Take an Oracle 10g class now, and we'll give you the exam = FREE.=20 > http://ads.osdn.com/?ad_id=3D3149&alloc_id=3D8166&op=3Dclick=20 > _______________________________________________=20 > GDAlgorithms-list mailing list=20 > GDA...@li...=20 > https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list = > Archives:=20 > http://sourceforge.net/mailarchive/forum.php?forum_id=3D6188=20 >=20 -------------------------------------------------------=20 This SF.Net email is sponsored by: Oracle 10g=20 Get certified on the hottest thing ever to hit the market... = Oracle 10g.=20 Take an Oracle 10g class now, and we'll give you the exam FREE.=20 http://ads.osdn.com/?ad_id=3D3149&alloc_id=3D8166&op=3Dclick=20 _______________________________________________=20 GDAlgorithms-list mailing list=20 GDA...@li...=20 https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list=20 Archives:=20 http://sourceforge.net/mailarchive/forum.php?forum_id=3D6188 --=20 Andrey Iones=20 Project Leader PhD in Computer Science=20 www.saber3d.com=20 =20 --=20 Andrey Iones=20 Project Leader PhD in Computer Science=20 www.saber3d.com=20 =20 |