TSE "Milestone 3.5" Update
by Kyle Carter · in Torque Game Engine Advanced · 06/30/2006 (11:06 am) · 246 replies
Hi guys,
We've checked the next TSE update into CVS. It's NOT Milestone 4 - so no fancy lighting - but it does include many bugfixes, a new vastly more efficient rendering infrastructure (ported from Marble Blast Ultra), way faster skinning, and the first drop of the new Atlas codebase.
Brian has more complete info on the delta in this update, and should be posting it sometime in the not too distant future - I just wanted to let you guys know what was up ASAP. :)
Please note that some parts of this are still a little rough, and the new Atlas codebase does not yet have the full complement of art tools, but everything is functional and ready to use.
This thread is for posting general feedback on the release and collaborating on any bugs that might come up. Please be patient, as here in the States we're just about to go into a 4 day weekend - so responses might be a little slow. But we are committed to making everything work 100%, so definitely let us know what your experiences are!
Regards,
Ben Garney
Torque Technologies Director
We've checked the next TSE update into CVS. It's NOT Milestone 4 - so no fancy lighting - but it does include many bugfixes, a new vastly more efficient rendering infrastructure (ported from Marble Blast Ultra), way faster skinning, and the first drop of the new Atlas codebase.
Brian has more complete info on the delta in this update, and should be posting it sometime in the not too distant future - I just wanted to let you guys know what was up ASAP. :)
Please note that some parts of this are still a little rough, and the new Atlas codebase does not yet have the full complement of art tools, but everything is functional and ready to use.
This thread is for posting general feedback on the release and collaborating on any bugs that might come up. Please be patient, as here in the States we're just about to go into a 4 day weekend - so responses might be a little slow. But we are committed to making everything work 100%, so definitely let us know what your experiences are!
Regards,
Ben Garney
Torque Technologies Director
#2
We'll let you know when we've identified the problem, and what the fix is (or that we've put it into CVS).
Thanks guys!
06/30/2006 (12:05 pm)
FYI - we have some reports of oddities w/ the terrain_water_demo. Loss of FPS, other strangeness. The Atlas file we shipped as a demo also MAY be being corrupted by CVS. If it fails to load or has garbage behavior this may be why. Do let us know if you experience these problems (and on what HW), but we have them on our radar and are already working on fixes.We'll let you know when we've identified the problem, and what the fix is (or that we've put it into CVS).
Thanks guys!
#3
06/30/2006 (12:39 pm)
Also note, if you have a 128MB video card, the demos will not fit in that amount of memory at the moment, so you will get texture thrashing and very bad performance. We're working on it
#4
I'm getting alot of compile errors, I must have done something wrong.
Thanks very much !
06/30/2006 (12:43 pm)
Quote:Also note, if you have a 128MB video card, the demos will not fit in that amount of memory at the moment, so you will get texture thrashing and very bad performance. We're working on itWhich gives me another reason to upgrade my video card. :)
I'm getting alot of compile errors, I must have done something wrong.
Thanks very much !
#5
06/30/2006 (12:53 pm)
@Brian - I assume it's an code issue and not an art issue that the demos don't fit in 128MB of video memory? Could you give us a hint at what the problem is?
#6
06/30/2006 (12:55 pm)
Just a matter of tuning the texture blender and esp. it's texture cache. I have an update to it planned that will significantly reduce the runtime resource costs... more on that later, though. :) 
#7
Was this in the CVS yesterday or just now today? Just wondering as I grabbed it yesterday and started porting things over.
06/30/2006 (12:58 pm)
@BenWas this in the CVS yesterday or just now today? Just wondering as I grabbed it yesterday and started porting things over.
#8
06/30/2006 (1:14 pm)
Are the docs intended to be updated with this release or do we need to browse TDN for that? As of now, the docs that come with TSE are the old ones.
#9
06/30/2006 (1:57 pm)
If you regenerate your dOxygen documentation, you'll see that Atlas2 object has quite a bit of source code level documentation. 
#10
06/30/2006 (1:59 pm)
Awesome, download now!
#11
Thanks Stephen. (:
06/30/2006 (2:13 pm)
I haven't touched dOxygen before, but I guess that's another reason to do so.Thanks Stephen. (:
#12
@Brian: Is that a temporary issue with the 128 mbyte requirement and stays like this or will it be still possible to run it flawlessly with 64 mbyte cards again?
06/30/2006 (2:13 pm)
Cool, super cool!@Brian: Is that a temporary issue with the 128 mbyte requirement and stays like this or will it be still possible to run it flawlessly with 64 mbyte cards again?
#13
06/30/2006 (2:17 pm)
Cool... plenty of new things about Atlas ! Yeaaahhh ! A week to explore all of these new features. Let's go!
#14
And obviously for your game you should definitely be budgeting texture memory and making sure that, for the scenes you're rendering, you're spending resources where they need to be spent.
We initially started putting stuff into CVS yesterday, you definitely want to make sure you do an update so you have the latest, though!
06/30/2006 (2:20 pm)
Re: GPU memory size - it all depends. You can't fit the same amount of data in a 64 MB card, but the engine SHOULD run with some degradation in image quality on smaller cards. We're working on resolving that issue.And obviously for your game you should definitely be budgeting texture memory and making sure that, for the scenes you're rendering, you're spending resources where they need to be spent.
We initially started putting stuff into CVS yesterday, you definitely want to make sure you do an update so you have the latest, though!
#15
1. Is the debris rendering code in now? It was commented out in MS3
2. Does the alpha transparency work on the DTS shapes now? ("startFade" for the player shapes)
Edit: Forgot to ask: Does anyone know if a shoreline renderer comes in? This is the only thing in my eyes that keeps Atlas away from beeing super realistic together with the fantastic water.
06/30/2006 (2:24 pm)
Just two more questions:1. Is the debris rendering code in now? It was commented out in MS3
2. Does the alpha transparency work on the DTS shapes now? ("startFade" for the player shapes)
Edit: Forgot to ask: Does anyone know if a shoreline renderer comes in? This is the only thing in my eyes that keeps Atlas away from beeing super realistic together with the fantastic water.
#16
Edit :
06/30/2006 (2:43 pm)
Another question : Can we have a quick list of new features?Edit :
Quote:Forgot to ask: Does anyone know if a shoreline renderer comes in? This is the only thing in my eyes that keeps Atlas away from beeing super realistic together with the fantastic water.You're right. something new about this?
#17
As of right now, don't expect a GG implementation of a new shoreline renderer, but I'm certain that if anyone is willing to do the research and implementation, it would make for a great TDN article.
06/30/2006 (3:22 pm)
I spoke with Brian Ramage a while back about this specific request, and he mentioned that it was one of those systems that could be rather easily implemented for a small design requirement (per project), but could be a nightmare to make generic and agnostic enough to put into the core engine.As of right now, don't expect a GG implementation of a new shoreline renderer, but I'm certain that if anyone is willing to do the research and implementation, it would make for a great TDN article.
#18
06/30/2006 (3:43 pm)
Ya, getting a good looking shoreline in on the water is going to take a lot of work and time that could be better spent elsewhere. That's the main issue there.
#19
Atlas 2.0 update - I'll let Ben elaborate on that
Render Batching - see my latest .plan
Optimized skinning further - fixed some performance issues
Added support for dynamic vertex buffers in addition to the "static" and "volatile" buffers in TSE.
Added a warning material that will show up if a piece of geometry doesn't have a material mapped to it.
Debris is now rendering in TSE.
Tons of bug fixes (this is just a few):
Disappearing water on editing change is fixed
Objects/terrain jumping around in editor is fixed
Fixed XInput init problem
Fixed text edit control problem
06/30/2006 (3:49 pm)
Here's a more complete list of what's in this update:Atlas 2.0 update - I'll let Ben elaborate on that
Render Batching - see my latest .plan
Optimized skinning further - fixed some performance issues
Added support for dynamic vertex buffers in addition to the "static" and "volatile" buffers in TSE.
Added a warning material that will show up if a piece of geometry doesn't have a material mapped to it.
Debris is now rendering in TSE.
Tons of bug fixes (this is just a few):
Disappearing water on editing change is fixed
Objects/terrain jumping around in editor is fixed
Fixed XInput init problem
Fixed text edit control problem
#20
I have also found a compile break in the build
c:\tse\engine\atlas\core\atlasresourcetoc.h(465) : error C3867: 'AtlasTOC::cbOnChunkWriteComplete': function call missing argument list; use '&AtlasTOC::cbOnChunkWriteComplete' to create a pointer to member
c:\tse\engine\atlas\core\atlasresourcetoc.h(451) : while compiling class template member function 'void AtlasResourceTOC::queueChunkUpdate(StubType *)'
with
[
StubType=AtlasResourceConfigStub
]
This will only appear in VC 2005, see http://msdn2.microsoft.com/en-us/library/b0x1aatf.aspx for information. I guess it will be easy to fix but will leave it for you guys. It is a pointer compliance thing.
06/30/2006 (3:54 pm)
I think i was over on the wrong thread. There are some CVS problems the VC8 build is out of date it just needs the Atlas Core and a few other files adding in. I took a clean VC8 folder and it still failed. I have also found a compile break in the build
c:\tse\engine\atlas\core\atlasresourcetoc.h(465) : error C3867: 'AtlasTOC::cbOnChunkWriteComplete': function call missing argument list; use '&AtlasTOC::cbOnChunkWriteComplete' to create a pointer to member
c:\tse\engine\atlas\core\atlasresourcetoc.h(451) : while compiling class template member function 'void AtlasResourceTOC
with
[
StubType=AtlasResourceConfigStub
]
This will only appear in VC 2005, see http://msdn2.microsoft.com/en-us/library/b0x1aatf.aspx for information. I guess it will be easy to fix but will leave it for you guys. It is a pointer compliance thing.
Torque Owner Chris Byars
Ion Productions