WPF 3.5 SP1 Graphics with David Teitlebaum

Wow!

Love the demo's. Can't help thinking it will probably be compared to 'Compiz Fusion' by some. But this is really good, it aren't 'just effects', these are thing we can accually use !

Where can we download the demo's given and compile/run/render them ourselfs? I for one would love to see the performance of that on my pc.

Realy great work!

Sweet! DSP in WPF will definitely Rock!

Is there source code available for these demos at all? I would love to try these things out and see the code for myself.

Very, very cool.

You guys really deliver! Bring this stuff to Silverlight and watch Adobe Flash go, well, back to the dashboard.

Thats really cool stuff, I think I remember watching a c9 video awhile back about the trick used to get 3D hit detection to work, wasn't it like the 2D layer was essentially invisibly overlayed and moved perfectly to line up against the mouse with fancy maths..

That might of been the 'hack' to do it back then however..

The effects thing is also really slick as well..

Almost makes me sad to be a web dev now


.. so uh.. when will you guys deliver this power via sl?

I can see this being used to create some kind of replacement for the start menu or those clones of the dock like RocketDock.

This is the most interesting technology I've seen featured on C9 in a while.

I really really like the edited format 20 minutes is the perfect length. This could've easily gone an extra 10 minutes, maybe even 20.

Oh, and the content is cool, too.

WOW! Kudos to you WPF SP1 team on delivering all these awesome UI features. We can now get down to building the next generation of software experiences we've only enjoyed in movies and games up to this point.

Thank you Microsoft for investing and following through on this research.

that

is

in

sane

:O :O

 

please pleeeeease let there be more videos about effects :O

Hi 9ers, I'm one of the dev's on the WPF graphics team.  As you can see we're really excited about 3.5 SP1.

We haven't figured out the mechanism yet (msdn, codeplex, blog, etc), but I plan to release the transition demo+source once 3.5 SP1 ships.  We understand that not everyone has experience writing pixel shaders, so samples like this (and others) will be worth looking at.

As David said in the video, the shader effects integrate really well with the WPF platform.  This mean you can take a library of effects, and just drop them into your XAML without knowing a thing about HLSL.

Ion Todirel wrote:

...ensure the WPF meets customers requirements, what customers? How do they decide what features to add/improve in WPF?

Great question. There are lots of different ways we gather feedback: the Connect site, forum and newsgroup posts, support requests, blog comments, requests received via MVPs etc. We also run invite-only labs here in Redmond for some of the very largest customers and partners: we gather lots of feedback at those events also. Occasionally we have customers who are on campus come and deliver short presentations to the team where they demonstrate their own application and talk about some of the best and worst experiences they had in developing the product.

At the same time, we're constantly looking at the broad-scale research we collect about platform adoption - for example, at what point does the size of a runtime have a material effect on its adoption rate? How many people are still using Windows 2000? And so on.

Plus of course, almost all of us here actively use WPF as developers, whether for amazing demos like David produces, for writing technical articles or in building test harnesses to ensure the platform works consistently.

We're always keen to hear real-world experiences with the platform: it's immensely useful to get your feedback. Ian mentions a few ways you can provide that feedback above. If it's actionable feedback and not simply a request for technical support, and you're struggling to be heard through more traditional routes, you're always welcome to drop me an email at tims at microsoft.com and I'll ensure it gets routed in the right direction.

We know that we're only delivering if you are able to deliver on top of us - our success is measured by your success!

Tim

Brandon, thanks for saying you will post the sample code when SP1 ships. For now, even a very simple but complete (.sln to .ps) "hello pixel shaders in WPF" sample would be very helpful.

And of all 52 weeks in the year, our national DSL network chooses C9's WPF Week to go down, and go down hard. Guess I'll have a lot of video catching up to do in the weekend.

robski wrote:

Brandon, thanks for saying you will post the sample code when SP1 ships. For now, even a very simple but complete (.sln to .ps) "hello pixel shaders in WPF" sample would be very helpful.

Hey robski, I think you might well like Greg Schechter's blog series that goes into this in some detail:
http://blogs.msdn.com/greg_schechter/archive/2008/05/12/a-series-on-gpu-based-effects-for-wpf.aspx

Tim

watching the video again a question comes to mind,

can you use other filtering techniques than nearset neighbur on a writable bitmap?  how much work is it to get something like that to work?

I will work on getting these posted soon.  They are almost ready to go.  And I believe Brandon is working on getting his demos (the multi input transitions) posted too! 

DukeNukem wrote:

Love the demo's. Can't help thinking it will probably be compared to 'Compiz Fusion' by some. But this is really good, it aren't 'just effects', these are thing we can accually use !

Where can we download the demo's given and compile/run/render them ourselfs? I for one would love to see the performance of that on my pc.

Realy great work!

You can use any of the three filtering methods we now expose (bilinear, nearest neighbor, and the "Fant scaler").  These are selected from the RenderOptions.BitmapScalingMode attached property, and affect the entire subgraph from which they're applied.

aL_ wrote:

watching the video again a question comes to mind,

can you use other filtering techniques than nearset neighbur on a writable bitmap?  how much work is it to get something like that to work?

Internally the term "customer" is really generic and refers to anyone who uses the thing you build.  Thus you can have "internal customers" who are just another team within the company.

And as Tim mentioned, the process of choosing what features to add/improve is really pretty thorough and includes an assessment of who/what scenarios would benefit from the work, what the risk would be from a performance, stability, security, or architectural standpoint, and an rough idea of how much time that work would require.

Ion Todirel wrote:

I'm a little bit confused, David said he's also responsible for design and ensure the WPF meets customers requirements, what customers? How do they decide what features to add/improve in WPF?

The performance of D3DImage is really scenario dependent, but generally we've had very good results... meaning full frame rates for complex scenes.

David

David Teitlebaum [MSFT] wrote:

You can use any of the three filtering methods we now expose (bilinear, nearest neighbor, and the "Fant scaler").  These are selected from the RenderOptions.BitmapScalingMode attached property, and affect the entire subgraph from which they're applied.

sweeet

also i vaguely remember you saying that there wasnt any multipass support for pixel shaders? but what about the MultipassShaderEffect Class?  is that depricated or something? or am i just wrong about the multipass stuff? i noticed that MultipassShaderEffect doesnt have a RegisterPixelShaderSamplerProperty  method or any pixelshader(s) property though.. can you elaborate on how that works?

(yet) another question how about pixelshading for 3d models? is that possible somehow? since there is now a pixelshaderwrapper with constants and stuff, the step shouldnt be to long i'd imagine (but what do i know:P)

 

David Teitlebaum [MSFT] wrote:

Right, the multipass shader effect class has been cut and won't be in the final builds of WPF.  If any of it is in the beta, it's just a stub.

MultiInput effects *will* be in the final release, but shouldn't be in the beta.

oh ok can you talk anythign about why it was cut? it seemed like a good idea to me but i guess the panel (or something) with an effect containg another panel with another effect works as well but it just seems a little more cumbersome to me but im sure you have another trick up your sleeve

havent found anything about multiInput so far but i figured you'd just register more DPs as sampler properties and hook them onto diffrent registers (can you do that by the way?)

David Teitlebaum [MSFT] wrote:

In WPF3D we haven't yet added a mechanism to use a custom pixel shader when your 3D object is being rasterized.  The new effects are applicable to UIElements -- ie 2D.

That having been said, you could use a our new DX Interop (D3DImage) and do whatever you wanted in 3D, but I gather that's not exactly what you wanted.

ok any thing like that in the works post sp1? i guess perhaps you could apply a visual element and then try to massage the normal data into the shader with a property or something but that sounds hard

David Teitlebaum [MSFT] wrote:

Multi-input effects didn't make it because we ran out of time in our compressed release schedule to get it implemented and tested properly before we released. 

oh, im sorry thats not what i meant i realise that the time to get stuff in the beta was limited and if you had gotten everything in, it whouldnt be a beta no i ment, why was multipass shaders cut (you awnserd this below)

David Teitlebaum [MSFT] wrote:

In the grand scheme of things we nailed the most important effect scenarios with the features we did get, so the loss of multi-pass was not a deal-breaker.

alright, thanks 

Why can't you guys remember to post a link to the LOW-RES file on every video that you do?

It's very irritating when downloading with BITS to have to wait on 600Mb files.

---dup---

We're just now in the early planning stages for the post-SP, so nothing has really been decided yet.

aL_ wrote:



David Teitlebaum [MSFT] wrote: In WPF3D we haven't yet added a mechanism to use a custom pixel shader when your 3D object is being rasterized.  The new effects are applicable to UIElements -- ie 2D. That having been said, you could use a our new DX Interop (D3DImage) and do whatever you wanted in 3D, but I gather that's not exactly what you wanted.

ok any thing like that in the works post sp1? i guess perhaps you could apply a visual element and then try to massage the normal data into the shader with a property or something but that sounds hard

David Teitlebaum [MSFT] wrote: Multi-input effects didn't make it because we ran out of time in our compressed release schedule to get it implemented and tested properly before we released.

oh, im sorry thats not what i meant i realise that the time to get stuff in the beta was limited and if you had gotten everything in, it whouldnt be a beta no i ment, why was multipass shaders cut (you awnserd this below)

David Teitlebaum [MSFT] wrote: In the grand scheme of things we nailed the most important effect scenarios with the features we did get, so the loss of multi-pass was not a deal-breaker.

alright, thanks 

Multi-input effects didn't make it because we ran out of time in our compressed release schedule to get it implemented and tested properly before we released.  In the grand scheme of things we nailed the most important effect scenarios with the features we did get, so the loss of multi-pass was not a deal-breaker.

aL_ wrote:



David Teitlebaum [MSFT] wrote: Right, the multipass shader effect class has been cut and won't be in the final builds of WPF.  If any of it is in the beta, it's just a stub. MultiInput effects *will* be in the final release, but shouldn't be in the beta.

oh ok can you talk anythign about why it was cut? it seemed like a good idea to me but i guess the panel (or something) with an effect containg another panel with another effect works as well but it just seems a little more cumbersome to me but im sure you have another trick up your sleeve

havent found anything about multiInput so far but i figured you'd just register more DPs as sampler properties and hook them onto diffrent registers (can you do that by the way?)

In WPF3D we haven't yet added a mechanism to use a custom pixel shader when your 3D object is being rasterized.  The new effects are applicable to UIElements -- ie 2D.

That having been said, you could use a our new DX Interop (D3DImage) and do whatever you wanted in 3D, but I gather that's not exactly what you wanted.

aL_ wrote:



(yet) another question how about pixelshading for 3d models? is that possible somehow? since there is now a pixelshaderwrapper with constants and stuff, the step shouldnt be to long i'd imagine (but what do i know:P)

 

Right, the multipass shader effect class has been cut and won't be in the final builds of WPF.  If any of it is in the beta, it's just a stub.

MultiInput effects *will* be in the final release, but shouldn't be in the beta.

aL_ wrote:



David Teitlebaum [MSFT] wrote: You can use any of the three filtering methods we now expose (bilinear, nearest neighbor, and the "Fant scaler").  These are selected from the RenderOptions.BitmapScalingMode attached property, and affect the entire subgraph from which they're applied.

sweeet

also i vaguely remember you saying that there wasnt any multipass support for pixel shaders? but what about the MultipassShaderEffect Class?  is that depricated or something? or am i just wrong about the multipass stuff? i noticed that MultipassShaderEffect doesnt have a RegisterPixelShaderSamplerProperty  method or any pixelshader(s) property though.. can you elaborate on how that works?

Adam,

Very nice work indeed. What should I use if I wanted to access TV functionality from a TV Tuner card? Is that a DirectX only route or can something be done from WPF?

Thanks.

Hurry up and post some sample files I want to play with the demos.

It looks like you guys have done some really good to me, I'm really impressed. The next Office, Windows and everything else should be fantastic thanks to this technology.

Cool!

Fantastic demo.  I have a quick question re D3DImage:  Will it work in an xbap?  In other words, if you have a managed dx9 rendering back-end, will it fuse with the new dx interop without requiring full trust?

Thanks.

Unfortunately D3DImage does require full trust.  There are a variety of reasons for this, but among the most fundamental, the WPF team can't guarantee that the D3D APIs are safe for use in a partial trust environment.  Additionally, the SetBackBuffer method has the signature:

public void SetBackBuffer(D3DResourceType backBufferType, IntPtr backBuffer)

which takes an IntPtr as a param, necessitating full trust.

Murrderizer wrote:

Fantastic demo.  I have a quick question re D3DImage:  Will it work in an xbap?  In other words, if you have a managed dx9 rendering back-end, will it fuse with the new dx interop without requiring full trust?

Thanks.

Is there any sample code available with the 3d form stuff?  That just has me itching to try a few things out, and thinking of what might be ahead in 3-5 years if some of this gets into Silverlight - wow, it's going to be cool!
Nice work!

Rob

@David Teitlebaum:

i enjoyed this session on the teched in orlando.
As you said, please provide a download link for this demos.

thx,
Thomas

Did the demos ever get posted by any chance? I'm particularly interested in the audio waveform with fmod demo, but it's all really interesting. Well done guys!

hello again....

i am very disappointed at the moment...
since the teched@orlando 08 i am waiting to get the source code of davids wpf gfx demos...as he promised.

Is anybody here who can talk to david titlebaum?
We need demo code !...

The source code for this demo would be really cool!!!
More votes for source!!

I really like this video! So much! But now it's been more than 2 months, and where's the source code of this demo?

Hey Adam and David, even now .NET 3.5 SP1 and this includes WPF 3.5 SP1, has been released, where's the demo code? Please don't make anymore promises.

We want it now.

Eriawan

Hey, this is really, really cool stuff ! Congratulations to David and all the other developers working on this.  Thank you very much !

We need the source of the demos ,
as promised from david teitlebaum at his teched08@orlando session !

Thanks! But now where are the rest of the sample code of the video?

And again, it's been more than a month since Part 1. I may sound to be bad, but why can't I simply have the codes soon?
I think it doesn't matter for me if the code came directly enedited from David, instead of waiting for you to post the code, and it's been taking too long time to wait.

Eriawan

First, congratulations on doing a great job with SP1.

Now my question:
David, at 21:11 you mentioned that the integrated DirectX could also be used for wireframes.

After some investiation, the only thing I saw was a new D3DImage class (btw. it is very usefull). But it is not appropriate to create a 3d lines and wires.

So is there a way to use the real DirectX 3d lines or do we still have to use workarounds like ScreenSpaceLines3D and similar (use triangles to simulate lines) which are quite slow?

Thanks for the source code. The spinning 3d window thing is begging to be used as a some sort of gratuitus cool looking (and probably incredibly frustrating to use) menu!

I'll get to wasting as much time as quickly as possible! 

This single-handedly inspired me to move from Managed DX to WPF.

Digging up a pretty old thread here because I've just been studying the whole D3DImage thing which is an awesome, but very basic facility if you have a C++ 3d engine that you want to integrate with WPF. For example if you need to write a level editor for your 3d game you can use the actual graphics engine for the game itself in a WPF shell instead of the traditional choice of having to write a separate renderer for the editor that didn't use D3D or trying to code a user interface in D3D (which as David points out in this video, doesn't seem like a major pain until you actually go to do it).

Anyway, I saw in the WPF Futures video from PDC2008 that in the future we'll be able to use XNA as an ImageSource, thus keeping the whole application managed code. Any news on when that's going to be available? Hope we don't have to wait until .NET4. Although the 3d viewport in WPF allows you to do some 3d stuff, it's kindof reminiscent of Retained Mode in old versions of DirectX, i.e. good for putting a few objects on the screen, but nothing too serious. Having a proper 3d api in managed code with a surrounding user interface in WPF would be a dream come true.

Way cool... is there code available for the demo's. I'd like to compile and run them locally.

Cool cool stuff

 

But where is the promised code? The link http://adamkinney.com/blog/357/default.aspx don't work and there have never been anything else.

 

I believe SP1 is released now.. Please... send some code...

Better late than never, eh?

 

I dumped a bunch of the demos + source on skydrive:  http://cid-3ffaec66aac45172.skydrive.live.com/browse.aspx/WPF^_3.5^_PS1^_Demos

It should be publicly accessible.

 

Let me know if you are having trouble accessing it, or if you want some other demos that aren't here.

Thanks!
David

Visit this link to vote for multi-pass shaders which are so desperately needed.
http://dotnet.uservoice.com/forums/40583-wpf-feature-suggestions/suggestions/480961-add-gloweffect-and-multipass-shaders-