HOME | DD

XnaFreak In reply to dylanja [2019-06-17 06:32:42 +0000 UTC]

I am the wrong person to ask.! lmgtfy.com/?q=at+Microsoft.Xna…


You should contact the developer or you can make a BUG report here: Bug Reports on any XPS version
(Please deliver all 5 required information.)





My 2 cents:
" with the program "
With what version of what program ????????
a) my XNALara Update: version 9.8  ?
b) my New XNALara version 9.9  ??
c) something else ???

1th screen  ibb.co/KN8McN1
Your first screen shows a "Microsoft XNA" error message, so the XNA Framework is installed and works (and because the XNA framework required .NET, the net frameworks is also installed and should  works too)

The error is
Exception Type:

System.TypeInitializationException

Message:
The type initializer for 'Microsoft.Xna.Framework.Graphics.GraphicsAdapter' threw an exception.

Source:
Microsoft.Xna.Framework

Stack Trace:
   at Microsoft.Xna.Framework.Graphics.GraphicsAdapter.get_Adapters()
   at Microsoft.Xna.Framework.GraphicsDeviceManager.CheckForAvailableSupportedHardware()
   at Microsoft.Xna.Framework.GraphicsDeviceManager.ChangeDevice(Boolean forceCreate)
   at Microsoft.Xna.Framework.GraphicsDeviceManager.Microsoft.Xna.Framework.IGraphicsDeviceManager.CreateDevice()
   at Microsoft.Xna.Framework.Game.Run()
   at WindowsGame1.Program.Main(String[] args)

👍: 0 ⏩: 0

XnaFreak In reply to spiralsharingan [2019-03-28 13:34:46 +0000 UTC]

And some more




There is an anniversary version that addresses memory management to some extent.

👍: 0 ⏩: 1

SeyssInquart In reply to XnaFreak [2019-03-31 05:19:39 +0000 UTC]

I haven't tried to reproduce your scene with that exact model but i'll assume it was using 1024x1024 textures. I use the same resolution textures and can only upload 5-6 of my characters on XNALara. I suspect the problem is that they have a lot of render groups. On that subject, i need to keep most groups and i doubt UV remapping will help either.

Anyways, i wanted to post something about XPS:

It's great that you were able to take up the mantle from Dusan and Co. I assume that you made the current iterations of XNALara from Ver. ~9.9 onwards? I had a quick look at the retro releases and the difference in quality is huge (in a good way). Have you ever considered releasing a 64-bit version and monetising that version of XNALara?

I ask because i would like very much to buy a version of XNALara that will only give that 'outofmemory.exception' problem due to hardware limitations; instead of software.

I've hunted around for a long time for something else that allows for such ease with posing models. But there's nothing that comes close. I read somewhere that, apparently, deforming the armature in such a direct way is unusual. So that's why there aren't really any other programmes for doing it. Blender, Unity, Daz studio, they can be used for posing; but it's a hundred times more bothersome for different reasons.

No idea if you've ever considered taking XNALara to a different level. I saw your post about coding and how expensive writing software is. But i'm a big fan that likes to produce large scenes with lots of models (e.g. Market village) and who's been messing with XNALara for years. So i'm hopeful.

👍: 0 ⏩: 1

XnaFreak In reply to SeyssInquart [2019-03-31 09:51:07 +0000 UTC]

A 64-bit version would be a new program, because the "Application Programming Interface" (API) of XNA 4.x (64 bit) is not compatible to 3.x (32 bit)
  www.deviantart.com/comments/1/…

I first have to probe (sort out) my thoughts.

One (memory) problem is loading the data, the other problem is rendering the scene. The rendering problem could be solved by invoking an external 64 bit renderer (external renderer). Even a 32 bit renderer would works, because the scene is always rendered "surface" (Mesh eq Render Group) after surface.
The problem of loading the models could be solved by outsourcing (swap) the textures (load by reference), and reloading one "texture set" (diffuse, normal, specular ...) after the other on rendering time.

Another possible way is to remove the XNA Framework from XNALara. XNALara needs this framework to start, but it does not really use it. So XNALara inherits all the disadvantages of the 32 bit Microsoft XNA Framework, but it does not take advantage of XNA. Porting the XNA code to DirectX may be the least amount of effort.

👍: 0 ⏩: 1

SeyssInquart In reply to XnaFreak [2019-03-31 22:38:20 +0000 UTC]

The only thing i clearly understood is that some innovative solution may be possible, which is excellent.

Probably the most complicated things that i've ever attempted with XNALara was to maximise available virtual memory and also increase cache size in properties menu. It didn't work of course.

Like i said, i hope someday to buy a user license from you or sign up as a patron. If the latter, probably better to avoid Patreon. They'd be liable to censure you for that picture of Lara. Too much skin, you see.

👍: 0 ⏩: 1

XnaFreak In reply to SeyssInquart [2019-04-01 06:35:22 +0000 UTC]

:::::::
That's right.
If I find the time, I will explain in great detail what XNALara does (and what it does not do), what Porters do (and what they fail to do), and where are ways to do it more efficient.
 As a thought, think of cars (car 'AA', car 'BBB', car 'CC', car 'DDD', car 'E' ....) which are one behind the other (with a small gap between the cars).
AA BBB CC DDD E
Even if the (not so long) cars 'BBB' and 'DDD' free the space,
AA     CC     E
then no gap is big enough to park a long vehicle 'FFFFFF' in it! Although there is enough free space overall.
AA     CC     E
   FFFFFF
If from the begin, the vehicles were arranged differently (load order), then the problem would not exist.
BBB DDD AA CC E
        AA CC E
FFFFFF  AA CC E
The biggest elements in a 3D scene are not the 3D geometry data, but the 2D textures (a texture in memory is at least four times the size of the uncompressed image on the disk).
Some user like  RadiantEld  try to tweak the 3D data, to make the models good looking (right looking), like here XPS Transparency , by ordering the meshes (3D data).
Some experienced porters are already aware of this when porting the models.
Hardly anyone takes into account to order the textures in the models correctly. So meshes with big textures first. XNALara do also not arranged the meshes by texture size.
XNALara load at first the (diffuse, Normal, specular ...) images into to memory (like the short car 'AA' and later car 'BBB').
AA
Then the image 'AA' is encoded into the texture format  (like the long vecicle 'FFFFFF'). 
AA FFFFFF
Now, the space for the image 'AA' is discarded.
   FFFFFF
 (XPS do discard the image space, XNALara do not even free the image space -> memory leaks. See .NET unmanaged resources ") and the part about "slideshow" and "out of memory exception" here .
The same for image 'BBB' and the resulting texture 'GGGGGGGG'.
   FFFFFF BBB GGGGGGGG
   FFFFFF     GGGGGGGG
You can see already the trouble: Fragmentation If you now to try to load a new model (or a mesh from the same model), then there is no space left to load a big image 'HHHH' (or to convert it to a much bigger texture 'IIIIIIIIIIIII'), although there is enough free space overall. You get a "out of memory exception"
 If from the begin, the models would use "textures" and not a image format, then the problem would not exist. Read more  MTL OBJ materials file - Paul Bourke   " Image files and texture files can be used interchangeably. If you use an image file, that file is converted to a texture in memory and is  discarded after rendering."(by the application Lightwave)
 If from the begin, the models would orders the data, used for textures, differently (load order big data at first), then the problem would not exist.

::::::::::
In my opinion, maximise available virtual memory or increase cache size do not help at all. UV remapping makes no difference in memory space. Model size makes no difference in memory space. High poly models makes only a small difference in memory space ...
+ Optimizing texture resolutions makes a great difference.
+ Giving more memory space to 32bit process   makes a great difference. Be aware that you needs two steps: the "3 GB boot switch" for you OS, and the "/LARGEADDRESSAWARE" flag for your XNALara application!


:::::::::
You can never " buy a user license from me". I can earn more money in my day job than it is worth sacrificing my leisure time for money.
But, the XNALara source code is there  https://github.com/AerysBat/XNALara
You can always hire a developer to implement the changes, which you want to have in it.
BTW: " to censure you for that picture of Lara. Too much skin" ? LOL, I have never heard before that TRA Lara shows too much digital skin.

👍: 0 ⏩: 1

SeyssInquart In reply to XnaFreak [2019-04-02 04:20:37 +0000 UTC]

Your car analogy certainly explains why i'm only able to use about 1gb out of 12gb of memory (Ram? Vram? I don't know what exactly ctrl+r shows).

That's quite interesting about the distinction between textures and images.

I assume you basically mean to say that the images (JPEG, PNG etc) are converted to textures but XPS won't discard the images after rendering.

To summarise: It's a two-pronged problem.

This memory leak issue is especially horrible because, usually, the solution is to save your file/progress, close the programme, restart the programme, done. Obviously, that doesn't really help with XPS.

-----------------

Heck, it didn't take me long to figure out that it was the textures causing issues. I was never sure if high-poly models affected performance but you've confirmed for me that it does make a small difference.

Random question: Is it possible for an app (E.g game) to initially, and only, use 'textures' when rendering?

Thanks for the link. I had a read and i'll have to more carefully re-read it at some later date. Looks like it might be worth it. However, the 3Gb switch is probably no good to me because i'm using a 64bit OS (article says it's for 32bit).

-----------------

(No, my face didn't darken for a moment. It was just your imagination)

The day job pays, eh? You've got your living expenses covered, yes? But what if some unhinged Deviant found a way to trash Italy's economy and cause rampant inflation? You wouldn't be able to sustain yourself with just one line of work. A second job, possibly involving a certain piece of software, will be necessitated...

How? I guess i'll just make a really strong case for quadrupling the size of the public sector and taking as many predatory foreign loans as needed to sustain it.

----------------

But i'm being facetious. Like when i mentioned Patreon (But not by a lot – they've gotten somewhat puritanical in recent times).

Thanks for the lengthy responses. It's more than i would have expected. Always a pleasure to communicate with someone who's really into something, whether it's trees, hydrology or current affairs. In your case, it's computer science.  

👍: 0 ⏩: 1

XnaFreak In reply to SeyssInquart [2019-04-02 08:02:55 +0000 UTC]

 I also don't know what exactly Ctrl+R show.
Maybe, if you having 12gb of memory, "Ctrl+R" see this 12gb, but the .NET Framework 3.5 cannot allocate this memory.Without the 3gb boot switch, the 32 bit .NET Framework get at maximum 2gb. Around 1gb from this 2gb are already waste by the .NET and the XNA framework. Around 220 MB are required for XPS and some other 32bit apps. Summary 750-800 MB for your scene.

Right, the images (JPEG, PNG etc) are converted to textures.
+ XPS discard the images after converting it to textures and use the textures for rendering (your video card eccept only textures). But now there is a gap (fragmentation) in your memory. Only "small" textures can fill this gap. BTW: For .NEt is 85kb already a huge object.
- The old XNALara won't discard the images! (XPS do it). This "old XNALara" memory leak issue is horrible becauseo bviously, that doesn't really help with XNALara.

"Random question: Is it possible for an app (E.g game) to initially, and only, use 'textures' when rendering?"
When rendering, every app can ONLY use textures. A good app should never allow to use images. A very good app should also not use any 3D model format. A very good app (E.g "The Witcher 3") has the model data (geometry and textures) saved (dump) from the memory to the disk. The technical term is "Serialization and Marshaling ... ". This apps just mapping the data from the hard disk directly into the memory, without any converting. So no gaps and better performance (by loading the data). Maybe you played the games "Skyrim" and "The Witcher 3"?
-- If you enter a city or a house in Skyrim, you get a "loading screen" (different "World Space"). The data must be loaded and converted into the memory. This take some time.
++ If you enter a city or a house in The Witcher, you get not a "loading screen" (same "World Space"). The data are just blending (mapped) into the memory. This is efficiency. The downside of this is marshaling / unmarshalling is that (as Ex) two characters have a different internal structure, so even CD PROJEKT RED, the owner of The Witcher, cannot provide a perfect "uncooking" tool to extract the model data. Every model looks somewehre different in the memory and there is no unique algorithm to decode / encode the data. The algorithm to do it are the different classes inside the complete game engine. This is also one of the reasons why any update / patch / DLC breaks the existing modding tools on such apps. The data in the memory "shifts" during an update.
I know, Game Design is a long story

The 3Gb switch IS  good for you (using a 64bit OS).  The article says it's for 32bit APPLICATIONS, and XPS is a 32bit application. Of course, with a 3gb switch no 32 bit application can use your complete 12gb. Only a maximum of 3gb raw, also 1.7Gb for XPS; instead of 700Mb.


A second job is always good. But no XPS user can pay my fees
Take a look at  en.wikipedia.org/wiki/Sketchfa… 
Some news on Sketchfab + dAHey deviants!
It’s been two weeks since the integration of Sketchfab into deviantArt journals, already!


Workaround Sta.sh + Sketchfab embed :
If you want to embed a 3D graphic in a Deviation you'll have to use a workaround for now: go to sta.sh writer, paste the iframe embed code and submit. Give it a try! P.S. : You don’t need a premium account to use sta.sh.
Few updates for you guys :
#1 We reached 100,000 3D models uploaded to Sketchfab, thanks to you all! Some of those are coming from deviant users
Here’s a nice infographic to put that in context : https://sketchfab.com/100k
#2
@ sketchfab  is a "small" app to show, turn and zoom in / out a 3D model. 
The numbers:
In February 2013, Sketchfab raised an angel round of 370,000 euros.
Less than one year after, in December 2013, Sketchfab raised $2 million from Balderton Capital with Partech Ventures, Borealis Ventures, David Cohen and existing angel investors also participated in the funding.
In June 2015, Sketchfab raised a $7 Million Series A led by FirstMark Capital.
There is always a second job somewhere
  Some idealists or students work for free. Often, this even pays off.
Ex Tomas Pettersson a kind of a computer guy:
www.drpetter.se/project_sculpt… >>  pixologic.com/sculptris/

computer science, 3d models, video games, modding, chaos theory, binary trees, matrices & linear algebra, music, cars, girls, buddhism, free time ... scare time

👍: 0 ⏩: 0