Khronos.org Public Message Boards

Hey I'm new

Mon Nov 09, 2009 1:32 pm

by Teelfluhnub (Posted Mon Nov 09, 2009 1:32 pm)
What's up everyone, I'm new to the forum and just wanted to say hey. Hopefully I posted this in the right section!

gl_PointSize problem

Mon Nov 09, 2009 12:14 pm

by jgrarod (Posted Mon Nov 09, 2009 12:14 pm)
I have been trying to write some code to draw lines and points with different sizes.

I have tried this in the vertex shader:
//Vertex shader
void main()
{
gl_Position = ...;
gl_PointSize = 3.0;
}

if I modify the value of gl_PointSize anything happens (between 1 and 63), the points are always drawn in the same way. What am I doing wrong?

Changes in Firefox Nightly 2009/11/02

Mon Nov 09, 2009 9:30 am

by giles (Posted Mon Nov 09, 2009 9:30 am)
Yup. On the other hand, the data must be getting stored somewhere, even if the [] operator can't retrieve it -- otherwise the call to gl.bufferData wouldn't get passed anything in the example code I gave...

Using AMD SDK with Nvidia driver and Nvidia samples with AMD

Mon Nov 09, 2009 7:00 am

by Executor (Posted Mon Nov 09, 2009 7:00 am)
Nice job!
Tnx for info about interop - useful...

OpenCL + OpenGL (interop)

Mon Nov 09, 2009 5:16 am

by Executor (Posted Mon Nov 09, 2009 5:16 am)
I found answer here:
viewtopic.php?f=28&t=2174

I need use cl_gl.h from ATI Stream SDK, in official headers tokens not founds.

Loading large Meshes

Mon Nov 09, 2009 4:11 am

by Coolcat (Posted Mon Nov 09, 2009 4:11 am)

Quote:

and would scale roughly linearly?

yes, at least it should be in O(n) class

How to change text color

Mon Nov 09, 2009 1:17 am

by flashscan (Posted Mon Nov 09, 2009 1:17 am)
Is there a formal tool to convert glyphs in font files to openvg source codes?
I have tried font2openvg. It works but only for ASCII characters currently. Also, it is not quite stable either.
Thanks

How to Include Common File into OCL file!

Mon Nov 09, 2009 12:23 am

by dbs2 (Posted Mon Nov 09, 2009 12:23 am)
That should work fine. If it doesn't, then you should check the path to the file and make sure it is relative to the built binary. (E.g., some build systems put your program in Debug/bin or something, so you need to make sure your included file is in that directory as well.) If it doesn't work then you should file a bug against the vendor as that is supposed to be included. BTW -- if you convert your error code to a signed int you will get a nice small negative number that you can use to compare against the header file.

LLVM compilation failure -- why?

Mon Nov 09, 2009 12:19 am

by dbs2 (Posted Mon Nov 09, 2009 12:19 am)
That may or may not be a failure in Apple's compiler, actually. LLVM is usually really good about telling you what failed and where. If you get less descriptive errors it may be because LLVM generated code and passed it on to the GPU vendor's compiler which choked on it. That tends to get reported back in a less-than-helpful manner as some generic failure.

WebGl sheat sheet

Sun Nov 08, 2009 4:33 pm

by tuto (Posted Sun Nov 08, 2009 4:33 pm)
i found this

http://www.nihilogic.dk/labs/webgl_chea ... _Sheet.htm

maybe is good for someone

tuto

Double buffer

Sun Nov 08, 2009 11:09 am

by Coolcat (Posted Sun Nov 08, 2009 11:09 am)
Once again...the browser does swap buffers, when it does leave the mainloop() function. After that your interval timer does call mainloop() again. Double buffering is used automatically.

WebGL deliverables and plan

Sun Nov 08, 2009 10:28 am

by learnwebgl (Posted Sun Nov 08, 2009 10:28 am)
I have just found out this website http://webkit.org/blog/603/webgl-now-av ... nightlies/ which says,
"WebGL is a new standard being worked on in the Khronos consortium. The work done in Khronos is only available to its members, so I can’t show you the spec just yet. But it will become public within the next few months after a review by Khronos members.". It was posted on October 19th, 2009.

Where can I find the following info?

What are the key deliverables? Specification, utilities, tools, (conformance) test suites, etc. Many OpenGL apps use the glu.h and Windowing specific extensions (WGL, GLX, etc).
When are they going to be available for public (beta, final release, etc)? I only found http://www.nihilogic.dk/labs/webgl_chea ... _Sheet.htm

Those info would be useful so we could plan accordingly.
Thanks in advance for your help.

CLH_ERROR_NO_BINARY_FOR_GPU for image2d_t

Fri Nov 06, 2009 12:24 pm

by rob (Posted Fri Nov 06, 2009 12:24 pm)
Thanks, I have filed bug ID 7372367. Let's hope we see some action from Apple on this!

A WebGL cookbook

Fri Nov 06, 2009 8:46 am

by giles (Posted Fri Nov 06, 2009 8:46 am)
One common problem for people starting out in a new language is finding snippets of code to do specific simple tasks. A while back a friend of mine set up a "Cookbook" wiki to help people learning IronPython and it turned out to be very useful to a lot of people. So I've set up something similar for WebGL. It's only got a few recipes so far, but I'll be adding to it as I develop my tutorials -- and if you have something you'd like to add, you're very welcome to do so!

Here's the link: http://learningwebgl.com/cookbook/index.php/Main_Page

Cheers,

Giles

Vertex Shaders built-in variables

Fri Nov 06, 2009 4:50 am

by Coolcat (Posted Fri Nov 06, 2009 4:50 am)
I think you are searching for the current OpenGL ES Shading Language Specification. Take a look into chapter 7 ("Built-in Variables").

If there is in future a new version of that specification, you will find it here:
http://www.khronos.org/registry/gles/

RGB Internal Texture formats in SC standard

Thu Nov 05, 2009 6:45 pm

by jpilon (Posted Thu Nov 05, 2009 6:45 pm)
Your right, I believe your pretty much forced to do format conversions to comply with the specification. What you really want to avoid is doing multiple conversion, and sometimes that requires application developers to be aware of this issue. Some hardware support certain texture compression formats and expose the additional formats with extension. Regardless, the optimal texture format tends to be platform specific.

About Khronos ICD model..

Thu Nov 05, 2009 1:36 pm

by oscarbg (Posted Thu Nov 05, 2009 1:36 pm)
To Khronos
Is ICD in Win7 with multiple OpenCL implemenations selecting one or all?
Also exposing in 1 joint platform as Apple or multiple platforms as is we made a wrapper?

I want to know how to use webGL in my web program?

Thu Nov 05, 2009 9:25 am

by giles (Posted Thu Nov 05, 2009 9:25 am)
Hey Pion,

Glad you like the website!

Cheers,

Giles

Array reduction example

Thu Nov 05, 2009 6:02 am

by mathew (Posted Thu Nov 05, 2009 6:02 am)
Hello All
I am Mathew I have problem of Array,I am getting its properly work in program,Anyone can help me give me the some basic program Example of Array list..
Example of Array Reduction Is Really very Nice..I like this method.
Thanks..

Question about "Definitions", Basic Ideas for MD-Simulation

Thu Nov 05, 2009 1:16 am

by dbs2 (Posted Thu Nov 05, 2009 1:16 am)
You shouldn't need to use clFinish. You should either use events to make sure that each kernel waits for the one before it, or just use an in-order queue so they will execute in-order. clFinish is extremely expensive as you have noted. In general you want to enqueue as much work as possible to the device (multiple kernels, double-buffer if possible to overlap computation) and use events to determine when data is ready to be read back.

OpenCL-z Khronos ICD compatible version

Wed Nov 04, 2009 5:48 pm

by oscarbg (Posted Wed Nov 04, 2009 5:48 pm)
See for more info: http://oscarbg.blogspot.com/2009/11/ope ... -here.html

Builtin OCL wrapper and autodetects AMD and Nvidia implementations (195.39)..
Detecting up to 4 devices in my in Windows 7 and Linux (AMD 5850,Nvidia GTX 275,Intel CPU, AMD 4850)

Notes
=====

Works with Nvidia 195.39 OpenCL ICD or higher
Works with CPUs and AMDs GPUs with AMD Stream 2.0beta4 and Catalyst 9.11 or higher
Autodetects in displays info of AMD and Nvidia OpenCL drivers in a single launch in current form..
For detecting AMD drivers you need to define the enviroment variable ATISTREAMSDKROOT where the SDK is ..
I think in Windows this is done in the installer in Linux use something as
export ATISTREAMSDKROOT..

This build is hopeful futureproof searching OpenCL ICD and also AMD while it's not ICD..

Port of Apple demos to Windows..

Wed Nov 04, 2009 5:43 pm

by oscarbg (Posted Wed Nov 04, 2009 5:43 pm)
See http://oscarbg.blogspot.com/2009/11/por ... ndows.html

Works Nvidia GPUs, AMD GPUs and CPUs.. Needs Forceware 195.39

This package has 5 Apple OpenCL demos:

* Procedural Geometric Displacement
* Raytraced Quaternion Julia Set
* Procedural Grass and Terrain
* Procedural_Noise32
* NBody Simulation Example

Key points:
*It has 32 and 64 binaries.
*OpenGL interop! up to 2X in perf! (only on Nvidia 195.39 drivers currently) :
World first OpenCL OpenGl interop in Windows!?
*Select CPU and GPU and OGL interop or not in one exec
*Linux perhaps coming soon..
* Source code (depends on feedback on my blog): I have aVS2008 solution with project files!

Working Group Info

Wed Nov 04, 2009 2:42 pm

by Coolcat (Posted Wed Nov 04, 2009 2:42 pm)

giles wrote:

Quote:

I think I will do some performance tests over the weekend.

I'd love to see the results!

I put this into an new thread.

Moving data from global to private

Wed Nov 04, 2009 11:44 am

by gigurra (Posted Wed Nov 04, 2009 11:44 am)

dbs2 wrote:

Image objects are cached, so any spatial locality will get you a big boost in performance. Note that variables are by default private so you don't have to put __private (or just "private") in any of those cases.

I'd also be a bit careful here. It looks like you are reading the same data everywhere. That will mean that the first image read will load the cache and every single read thereafter will hit. The other ones are not cached so they will have to do the read each time. This appears to cover the case where you have extremely good locality. If your real code is accessing all over the place you will see far less benefit from the image access.

Thanks. That is good to know.
The important part is just that it wont be significantly slower than reading it element by element.

At least the grid which is only about 20x10x20 should be cached to some degree, as
it is quite sparse. (each entry in the grid holds 2 float4s)

Right now I just let every work item read the same data to test performance.
How much data do you think could be cached?

How to feed timing info into IL component?

Wed Nov 04, 2009 1:26 am

by andyquan (Posted Wed Nov 04, 2009 1:26 am)
Hi,
I want to feed timing info, e.g. playback latency (playback clock - timestamp), into IL component so that I can apply frame dropping or other dynamic configuration to codecs dynamically.

If I want to use BufferHeader as the media for transfer, which variable is recommended? Can I use "nTickCount"?

Any other suggestion?

Thanks,
Andy