Tagged with opengl-es - NVIDIA Developer Forums

Profiling OpenGLES2 applications using a Acer ICONIA A500 device

sylvainmeunier — Tue, 24 Apr 2012 12:04:52 -0400

Hello,

I try to develop benchmarks for profiling accurately an Acer ICONIA A500 device (Tegra and Android 3.2.1). I currently focus on its graphics performances.

OpenGL provides the GL_ARB_timer_query extension for profiling the graphic pipeline. Unfortunately OpenGLES2 doesn't provide a similar extension. Am I wrong ?

I worked on two approaches:
- EGL_NV_system_time
- NVPerfHUD ES

EGL_NV_system_time

The first approach uses EGL_NV_system_time functions for querying the system time and glFinish for synchronizing driver and pipeline times.


startDate = eglGetSystemTimeNV();
//
// Some OpenGLES2 calls
// ...
glFinish();
endDate = eglGetSystemTimeNV();
//
// Some profiling stuffs
// ...

Unfortunately, a large part of the results seem distorted by some vsync issues. Do you have any ideas for dodging them ?

NVPerfHUD ES

The second approach is based on PerfHUD tools. This tool requires a recent Tegra Android OS image which has PerfHUD support but NVidia just provides some images for Ventana and Cardhu devices and Acer doesn't provide anything until now. Am I wrong ?

Following several posts on this forum (Post, Post...), I tried to retrieve missing libraries and scripts from Ventana system image but failed on a file system issue. This file isn't an ext4 file system nor an yaffs2 file system but a "basic" file system (see flash.cfg in Ventana image). Do you have any tips for solving it ?

Optimal vertex attribute formats for Tegra2/3?

clb — Thu, 10 May 2012 12:26:39 -0400

Hi,

I am trying to figure out what would be the most optimal vertex format to use for my vertex data. Does there exist somewhere a guide/tips on how to lay out data for best performance?

I am mostly comparing the performance guidelines to ones published for iOS ( http://developer.apple.com/library/ios/documentation/3DDrawing/Conceptual/OpenGLES_ProgrammingGuide/OpenGLES_ProgrammingGuide.pdf , page 65+), since I cannot find a similar document for Tegra2/3. Do such guidelines exist somewhere?

Specific questions:
- Does GL_FIXED carry a penalty as opposed to GL_FLOAT? Should I always use GL_FLOAT instead of GL_FIXED?
- Is there a performance difference on the order of laying out data with different semantics inside a single vertex. E.g. having pos/normal/uv versus pos/uv/normal, and loading them in different orders in the shaders. The iOS guide lists a preferred order, do similar rules apply here?
- Is there a performance hit if GL_UNSIGNED_BYTE data is not aligned to two bytes? or four bytes?
- Is there a performance hit if GL_UNSIGNED_SHORT data is not aligned to two bytes? or four bytes? iOS guides mention a performance hit if GL_UNSIGNED_SHORT is not aligned to four bytes, and suggests adding two bytes of padding. How about Tegra2/3?
- Is there a performance hit if GL_FLOAT data is not aligned to four bytes?
- Is it faster to pack vertex data fields inside a same attribute? E.g. with two sets of 2D UV coordinates, should I pack them into a single 4D {uv0.x uv0.y uv1.x uv1.y} attribute and separate in the shader?

What kind of other tips would you recommend? I am currently packing as much of my data into GL_UNSIGNED_BYTEs as possible, and am now worried about the misalignment of some data fields it causes.

If nothing else, I'll probably run some profiling to investigate this, but if people have done this already, would be great to hear the results.

Other related questions:
- Is there any practical difference between GL_DYNAMIC_DRAW and GL_STREAM_DRAW in the Tegra2/3 drivers? I am currently doing double-buffered (swapping between two created VB objects each frame) updates to a GL_STREAM_DRAW vertex buffer for my particles. Should I be double-buffering myself, or are the Tegra2/3 drivers already double-buffering/optimizing for me for per-frame updates if I specify GL_STREAM_DRAW?
- Do Tegra2/3 support Vertex Array Objects?

Thanks for the help!

Is 1080p HDMI output possible on Tegra3?

clb — Thu, 10 May 2012 05:24:48 -0400

Hi,

I am investigating the different options that are possible when using a HDMI link to an external display. Looking at how it currently works, it seems that the main display is always mirrored to the HDMI display, giving a 1280x800 res. This is quite surprising, since the Asus Transformer Prime adverts claim it to be able to play back 1080p video on HDMI. What gives?

I'm now pondering whether any of the following scenarios are possible:
- Extend/separate display mode. Use Prime screen as 1280x800 display and use the other display as an extended monitor in 1920x1080 or 1920x1200 mode, whatever is the max. res for that display. Can I simultaneously render different content to both displays using GLES2? This would give full control and support to specify what content is displayed on which screen.
- Turn off the small 1280x800 display, only use the large external 1920x1080 display. Render natively to the 1920x1080 display without upscaling from a smaller resolution. Is this possible? I notice that Shadowgun is not doing this at least, but the game is played in 1280x800 res on the HDMI display, which the monitor stretches to fit the screen.

If neither of these scenarios are doable, do you know if that is only due to software limitations, or are there potentially some hardware restrictions that will limit Tegra3 to always mirroring a single max. 1280x800 display?

Thanks for any help!

How to access NV_ OpenGL ES extension for stereoscopic output?

dsewtz — Tue, 24 Jan 2012 06:41:49 -0500

I would be very much interested in a documentation on how to access the opengl es extension for stereoscopic output. I assume it's NV_MULTIVIEW_DRAW_BUFFERS ?
(As seen on the CES: Attached Asus Transformer Prime through HDMI1.4 to 3DMonitor)

Tegra 2-3 ICS driver memory leak?

Gary13579 — Mon, 02 Apr 2012 20:26:20 -0400

It appears that the Tegra driver for Android 4.x introduced some sort of memory leak which has broken several applications. Googling it a bit, it seems to specifically effect Asus Eee Pad Transformer and the Transformer Prime, but there are also occurrences of the problem on the Galaxy Tab 10.1. I personally own the Transformer TF101.

What happens from an end user standpoint is the application will go to a black screen and become unresponsive. This happens to several select applications, the main ones that I have noticed being Chrome, GMail, and AndChat (an IRC client). The only way to fix the problem is a full tablet reboot.

Here is a logcat output of when the application AndChat freezes. https://gist.github.com/2259510

This logcat was taken whilst running a completely stock ROM, however I should note that this problem occurs on every ROM I have tried, including several AOSP builds. I have tried clean installs of Android, by wiping everything from the tablet and only installing AndChat, and the problem still occurs.

What makes me think it is a GPU driver memleak are 1) the error "OpenGLRenderer is out of memory", 2) the fact that it only occurs after using the tablet for a while, 3) a reboot fixes it temporarly.

I do not know how to proceed in getting this fixed. I could contact ASUS, but it would likely just result in me talking to someone that barely speaks English, convincing me that I need to reboot the tablet and it'll magically fix itself. If I managed to get a developer at ASUS, they would likely just forward the problem to you guys.

Please direct me to who I should contact in getting this fixed. I am more than happy to test/debug things to the best of my capabilities to help resolve this issue. As of now, though, I have a tablet which is completely broken for the past few months since ICS has been released. If this is the quality of drivers that nVidia releases, I will likely not be purchasing Tegra based devices for my company any more.

Revisiting NPOT, ETC1, MipMapping

Rob Nugent — Thu, 22 Mar 2012 09:26:42 -0400

Hi,

I've finally managed to distill the problems I reported in the following conversation into a simple test case:

http://forums.developer.nvidia.com/devforum/discussion/596

Essentially I'm trying to use non-power-of-two texturing with ETC1 compression and mipmapping. If the texture size *is* a power of two (say 128x128), it works OK and gives results as shown in good_128x128.png (attached). This is a synthetic example with the mipmap levels showing as green/red/blue squares each with a white centre.

If the texture size is *not* a power of two (say 124x124) I get a bad image as shown in bad_124x124.png (attached)

I think I've followed all the suggestions in the earlier conversation. I'll append the source in a subsequent append. If anyone can help or offer suggestions, I'd much appreciate it.

This is all running on my Asus EPad Transformer TF101 now running Android 4.0.3

The source code can be modified from the working to failing cases by changing line 93 of C4Cube.java from:

createMipMapCompressed(128);

to

createMipMapCompressed(124);

Many thanks in advance.
Rob

Tegra ETC1 NPOT MipMaping

Rob Nugent — Tue, 20 Sep 2011 04:56:51 -0400

Hi,

I'm experimenting porting one of my existing PC opengl mapping applications to Android. My development platform is a Tegra (Asus EEE Pad Transformer).

I'd like to use ETC1 texture compression with non-power-of-two textures and mipmapping but can't get this to work at all. I can get mipmapping working with non-compressed textures, and I can get npot-ETC1 textures working without mipmapping.

If anyone can tell me if this is actually supported on Tegra, I'd really appreciate it.

Regards,
Rob

Tegra 3 Tablet Camera API

MichaelD — Mon, 26 Mar 2012 03:39:57 -0400

Hi,

i am doing my master thesis with the Tegra 3 tablet development kit. I want to do real time image processing on the stereo camera the tablet is providing (calibration, stereo correspondence, ...).

The FCamPro example does something like that - it gives an OpenGL texture to a SurfaceView. Is there an existing API-function that allows me to get the current (stereo)frame and manipulate it with some algorithms and pass it on to an (GL)SurfaceView - like the onPreviewFrame() function does on the Android API?

Thanks,
Michael

Tegra 2 GLES2.0 shader question

dashxdr — Tue, 20 Mar 2012 16:17:43 -0400

Does the Tegra2 fragment shader support flow of execution instructions? A piece of fragment shader code I want to use has a for() loop with a conditional break, but the behaviour is as if the entire loop has been unwound, which destroys performance.

I really want to know if the ULP GeForce used on the Tegra2 even supports the conditional flow of execution ARB instructions at all, and I can't seem to find any documentation or anything that has that particular detail.

If the tegra2 does support those instructions,
1) How can I tell the GLES20 compiler to make use of them when it is compiling my shader code?
2) If I can't do it with GLSL, can I wrote the ARB assembly language myself?

This is on an android samsung galaxy 10.1 tablet, by the way.

Thanks for any and all replies.
-Dave

Future of Harmony-based devices

dpol — Tue, 21 Feb 2012 06:04:46 -0500

Hi...
The part of relnotes L4T beta - "...developers should note that support for Harmony devices shall be deprecated following this release."
If yours company is't want (or have not any resources for this) to support this devices, where may I get hardware specific documentation to write own drivers ?..

Best regards

L4T beta for hardfp

dpol — Mon, 20 Feb 2012 05:00:45 -0500

Hi... I've some issues with alpha L4T (bug in libcgdrv.so... It's unuseable for compositing with gl-rendering, only software-x11). Where may I download L4T beta for hardfp ?..

Best regards

Use CgFX on Android ?

alexandretrog — Fri, 24 Feb 2012 18:35:56 -0500

Hi,

I was wondering if we can use the CgFX libs on Android (ARM) ?

20 Euro Pico-ITX Tegra 2 Computer System Promotion

ToradexDan — Thu, 23 Feb 2012 19:57:31 -0500

I'm sorry for the commercial post, but maybe some of you appreciate information about this deal.

Exclusive to Embedded World 2012, Toradex is offering a limited number of Performance Kits, consisting of the Colibri T20 embedded computer module, based on the NVIDIA Tegra 2 processor, and a corresponding Iris Carrier Board. The Performance Kit comes pre-installed with Windows Embedded Compact 7 or you can download Linux form our webpage.
The kits are available to qualified customers for a donation of only €20 to the Red Cross.

For more information see our promo video:

instrumented OpenGL ES drivers for end user device to use PerfHUD ES?

dsuoch — Thu, 15 Sep 2011 11:43:51 -0400

hello,

is there any chance to get instrumented OpenGL ES drivers for end user devices (like LG SU 660)? to be able to finally use PerfHUD ES?

in the old forums there was a thread about some hack, but i am not sure if i should try it ...
http://www.enovapp.com/tips/how-to-enable-perfhud-es-on-all-android-tegra-devices/

all the competitors have something like this working OOTB. powervr, adreno. why not nvidia,
when they apparently set so much on tegra?!? sorry, but without proper GPU tools we cannot make games that shine on nvidia hardware ...

so come on, please show us devs some love!

OpenGL ES 1.0 illumination sphere - request for help

MajsteR — Sat, 21 Jan 2012 09:04:36 -0500

Hello.

At the start let me say that I am very glad that I came here. I am writing a project in Java and Android exactly. I also use a microprocessor Tegra2. I use the graphics library OpenGL ES 1.0.
Unfortunately, some time ago I came across a problem with the lighting sphere, which I can not solve. I searched and asked probably everywhere, so much please help.

The error is somewhere in the programming of lighting.
Despite the fact that all parameters, seems to me to initiate properly.

Parameter screen when turned on gl.glEnable (GL10.GL_LIGHTING);
remains completely black.

The method, which draws each frame.
public void rysuj(GL10 gl) {/ / Draw
gl.glFrontFace (GL10.GL_CW);
gl.glEnable (GL10.GL_TEXTURE_2D);
gl.glEnable (GL10.GL_NORMALIZE);

gl.glEnableClientState (GL10.GL_VERTEX_ARRAY);
gl.glEnableClientState (GL10.GL_NORMAL_ARRAY);
gl.glEnableClientState (GL10.GL_TEXTURE_COORD_ARRAY);

gl.glVertexPointer (3, GL10.GL_FLOAT, 0, vertexBuffer);
gl.glNormalPointer (GL10.GL_FLOAT, 0, normalBuffer);
gl.glTexCoordPointer (2, GL10.GL_FLOAT, 0, textureBuffer);

gl.glDrawArrays (GL10.GL_TRIANGLE_STRIP, 0, liczbaWierzcholkow);

gl.glDisableClientState (GL10.GL_VERTEX_ARRAY);
gl.glDisableClientState (GL10.GL_NORMAL_ARRAY);
gl.glDisableClientState (GL10.GL_TEXTURE_COORD_ARRAY);
}

In this function, I set all the parameters of OpenGL:

public void onSurfaceCreated (GL10 gl, EGLConfig config) {
mSphere = new Sphere (1, 0, 0, 0);

/ / Initialize all configuration settings for OpenGL ES 1.0
gl.glClearColor (0.0f, 0.0f, 0.0f, 0.0f);

gl.glHint (GL10.GL_PERSPECTIVE_CORRECTION_HINT, GL10.GL_NICEST);
gl.glEnable (GL10.GL_DEPTH_TEST);

gl.glDepthFunc (GL10.GL_LEQUAL);
gl.glEnableClientState (GL10.GL_VERTEX_ARRAY);

/ / Enable auto-normalize the normal vectors
/ / Gl.glEnable (GL10.GL_NORMALIZE);
/ / Enable lighting and configuration of a single light source
oswietlenieDook float [] = new float [] {1.0f, 1.0f, 1.0f, 1};
oswietlenieRozp float [] = new float [] {1.0f, 1.0f, 1.0f, 1};
oswietlenieKier float [] = new float [] {1.0f, 1.0f, 1.0f, 1};
pozSwiatla float [] = new float [] {2, 0, 0, 1};
gl.glEnable (GL10.GL_COLOR_MATERIAL);
gl.glShadeModel (GL10.GL_SMOOTH);
gl.glEnable (GL10.GL_LIGHTING);
gl.glEnable (GL10.GL_LIGHT0);
gl.glLightfv (GL10.GL_LIGHT0, GL10.GL_AMBIENT, oswietlenieDook, 0);
gl.glLightfv (GL10.GL_LIGHT0, GL10.GL_DIFFUSE, oswietlenieRozp, 0);
gl.glLightfv (GL10.GL_LIGHT0, GL10.GL_SPECULAR, oswietlenieKier, 0);

gl.glLightfv (GL10.GL_LIGHT0, GL10.GL_POSITION, pozSwiatla, 0);

matDook float [] = new float [] {0.73f, 0.72f, 0.73f, 1};
matRozp float [] = new float [] {0.5f, 0.55f, 0.5f, 1};
matKier float [] = new float [] {1.0f, 0.95f, 0.8f, 1};
gl.glMaterialfv (GL10.GL_FRONT_AND_BACK, GL10.GL_AMBIENT, matDook, 0);
gl.glMaterialfv (GL10.GL_FRONT_AND_BACK, GL10.GL_DIFFUSE, matRozp, 0);
gl.glMaterialfv (GL10.GL_FRONT_AND_BACK, GL10.GL_SPECULAR, matKier, 0);
gl.glMaterialf (GL10.GL_FRONT, GL10.GL_SHININESS, 27.8974f);

/ / / / Starts the texture handling in
/ / Gl.glEnableClientState (GL10.GL_TEXTURE_COORD_ARRAY);
/ / Gl.glEnable (GL10.GL_TEXTURE_2D);

/ / Load a bitmap texture of the ball
if (texture == null) {
texture = BitmapFactory.decodeResource (kontekst.getResources (), R.drawable.kula_textura);
}
Kula.wczytajTeksture (gl, texture);
}

Function DrawFrame

public void onDrawFrame (GL10 gl) {

/ / Write the drawing code needed
gl.glClear (GL10.GL_COLOR_BUFFER_BIT | GL10.GL_DEPTH_BUFFER_BIT);
gl.glLoadIdentity ();
setxRot (30f / 1000f);
setyRot (15f / 1000f);
/ **
* Change this value in the if you want to see the image zoomed in
* /
gl.glTranslatef (0.0f, 0.0f, 5.0f-);

gl.glRotatef (xRot, 0.0f, 1.0f, 0.0f);
gl.glRotatef (yrot, 1.0f, 0.0f, 0.0f);
mSphere.rysuj (gl);
}

And in this way generates a buffer of normal:
normalBuffer.put (x / R);
normalBuffer.put (y / R);
normalBuffer.put (z / R);
x, y, z - coordinates of the apex; R - radius of the sphere;

I ask that if possible someone tried to solve it.

Super Fast Booting Tegra 2

ToradexDan — Thu, 01 Dec 2011 13:02:08 -0500

Toradex just released a video of a Colibri T20 computer module which boots Windows CE 6 and Windows Compact 7 in less than 1 Second. Supporting OpenGL, Touch, USB, ...

You will be able to download the binary Image, BSP and the Workspace file to try it by yourself.

Daniel

Is NVIDIA complying with its Open Source obligations?

Core Memory — Fri, 02 Dec 2011 11:54:01 -0500

On xdadevelopers, I read NVIDIA's announcement about making Tegra binary code and drivers, etc. available to independent developers but no where is the release of source code (un-obfuscated human readable code) stated. Although developers appreciate binary code, it isn't enough. The concept of Open Source is openness so that there's a level playing field of opportunity for every developer which I still don't observe with NVIDIA. Is NVIDIA complying with its Open Source obligations?"

No Tegra 2 driver source code = No independent developer support for NVIDIA Tegra 2, 3, etc.

Core Memory — Fri, 25 Nov 2011 14:15:27 -0500

Over at xdadevelopers.com, developers and consumers are complaining about the lack of support for the Tegra 2 from NVIDIA which is preventing the development of kernels to upgrade phones such as the LG Optimus 2X and G2X which phone manufacturers, such as LG, appear to have abandoned support for. NVIDIA must provide the information and tools necessary for independent developers or they will lose that too and, when consumers realize that they bought phones which wouldn't be updated and also don't use the full capabilities of the NVIDIA GPU, they will stop buying phones with NVIDIA chips inside. I own a G2X which LG appears to have abandoned, they were forced to provide Android 2.3 in the U.S.A. but, apparently, that's it for updates to anything newer. NVIDIA, you can't depend on phone manufacturers to support your technology, you have to do it yourself by openly providing the source code to your drivers, including the hardware accelleration code, or you will be left behind in the dust by Texas Instruments and other companys which do supply what independent developers need to fully develop software. Beware NVIDIA, you could be headed into oblivion as RIM appears to be and as already happened with NORTEL which was so big and successful it didn't think it could fail.

Poor performance of Tegra 2 in general case.

stevehaggerty — Tue, 29 Nov 2011 06:09:13 -0500

Hi there,

We've been getting a few emails from users commenting on the generally poor frame rate of our game on Tegra 2 devices. This game has been run a great deal (it reached #2 in all paid apps) so there has been a large amount of testing.

We're doing all the obvious:
native app,
compressed textures,
optimised precision friendly shaders,
avoiding dependent texture reads,
caching states, etc

But the title still performs worse on Tegra 2 than say a SGX540 based system.

Is this to be expected? I've read nothing in the docs to suggest there are any "easy wins"
to be had either. Maybe I'm over-estimating the hardware capability?

Any suggestions?

Many thanks,

Steve.

OpenGL-ES and Zune HD

yalfonsom — Mon, 28 Nov 2011 13:50:15 -0500

I have a Zune HD and I downloaded the OpenZDK framework and also installed the prerequisites suggested, but when I deployed an OpenZDK app, the device tried to launch it, but after a few seconds the device restarted itself. Then I went into the code and commented all the gl* funcions call and deployed a test application. That time the device didn't restart. Later on, I discovered that a single call to a gl* function, even glClear, causes the device to restart itself. So far, my question is: do I need to install a specific feature to use OpenGL-ES on my Zune player??? Any ideas will be well appreciated. Thanks in advance. (Sorry my English)

Running Tegra2 samples on Android Emulator - Invalid APK?

Spaddlewit — Sat, 26 Nov 2011 14:32:35 -0500

I installed the Tegra2 Android Development Pack.
When I try to 'Run As->Android Application' for the Tegra2 samples, I get this:

[2011-11-26 12:57:58 - es2_water] Uploading es2_water.apk onto device 'emulator-5554'
[2011-11-26 12:57:58 - es2_water] Installing es2_water.apk...
[2011-11-26 12:58:12 - es2_water] Installation failed due to invalid APK file!
[2011-11-26 12:58:12 - es2_water] Please check logcat output for more details.
[2011-11-26 12:58:13 - es2_water] Launch canceled!

What am I doing wrong?

Tegra 2 Source Code

Ts0cha0tik — Tue, 22 Nov 2011 13:28:26 -0500

Are you ever going to release the Tegra 2 Source code so developers can actually work with the supporting android devices? Its because of this that I refuse to buy any more Nvidia related devices

Tegra 2: Slow floating point texture operations

agso0 — Wed, 16 Nov 2011 17:11:34 -0500

Hi guys.

I am stuck in the middle of a problem, I hope someone can help me with this one:

Currently I have an application developed on android NDK targeting android 3.2 honeycomb.
Everything went well until I decided to include support for floating point textures to add image postprocessing functionality to my effects framework.

The thing is that when I create floating point render targets (textures with GL_FLOAT as the type of the components), the application performance drops insanely.
If I change the type of the format to GL_UNSIGNED_BYTE, everything goes back to normal frame frecuency.

I tested with a sample program with just a textured quad and the results are similar: a significant drop in the frame rate.

The target hardware is a ViewSonic Viewpad 10S with a Tegra 2 processor.

The requested format for the back buffer is GL_RGBA8888.

I appreciate your time. Thanks.

NOTE: Currently, the message edition control in this forum has a bug (it does not let you edit the message content, at least in Internet Explorer 9), so, a workaround is to write your message in
notepad (or some other editor) and then paste the content in the editor (though still the control does not show the content, but is there).

OPENGL ES 2.0; Tegra2; GPU; How much GPU cores are used into glDrawArrays/glDrawElements functions;

GreenTroll — Thu, 10 Nov 2011 07:58:29 -0500

Hi everybody!

Do Anybody have information about how much GPU cores will be used when I call glDrawArrays/glDrawElements???

A bit more details to explain my question.
Processor Tegra2 has 4cores GPU. To work libGLESv2.so is used.
After all preparatory works have been done (create and link shaders; upload textures and etc), I call DRAW function which started rasterization and create image in framebuffer.
I think, DRAW function has to use as more cores as possible to do rasterizarion more fast.
But I can't found any documents which confirm my theory.
Description of OPENGL has only information about there own level API, and, understandably, not any information about below levels. NVIDIA don't present description how libGLESv2.so is realized.

Thanks for your attention.
With Best Wishes.

inout function parameters with precision qualifiers are not supported on Tegra?

miai — Mon, 14 Nov 2011 07:51:34 -0500

Hello!

I am developing a OpenGL ES 2.0 app for Samsung Galaxy Tab and I have encountered a strange problem while compiling my shaders.

My fragment shader looks like this:

...
void test(in float arg1, out float arg2)
{...}
...
mediump float var1 = 1.0;
mediump float var2;
test(var1, var2);

When I am calling the method test(...) with mediump float variables I get a compile error. Also adding mediump to the params from the method declaration, there is another error saying that inout with precision qualifiers params are not supported. The GLSL ES standard document says that this should be valid.

Do you have any advices for this situation?

Non-blocking work mode for OPENGL ES function.

GreenTroll — Thu, 10 Nov 2011 08:21:15 -0500

Hi everybody!

Is it possible to use functions of libGLESv2.so in non-blocking mode to do upload textures and DRAW process simultaneously? (libGLESv2.so from NVIDIA SDK for Tegra processors).
I'm speaking about glDrawArrays/Elements and glTexSubImage2D functions.

For example, I had uploaded texture1 and call DRAW_1. While rasterization is executing I could start to upload texture2 to have ready data in GPU memory when rasterization_1 will have finished.

Is it possible? How can I realized it or somethink like that?

Thanks for your attention.
With Best Wishes.

Tegra dev headers for EGL/OpenGL ES 2 development

vatjjar — Mon, 10 Oct 2011 06:12:17 -0400

Hello,

I'm a happy owner of Trimslice Tegra device, and starting my tests with the OpenGL ES capabilities of the device. I have managed to setup Ubuntu as my development platform and I'm running the alpha 1 release of the driver base there with kernel 2.6.38.3.

I have also managed to compile and run OpenGL ES 2 / EGL programs in the platform, so in that respect everything is nice and smooth. However, two things require a little clarification:

1) I've tried to seek NVidia developer headers (egl.h, gl2.h, gl2ext.h etc etc) for Tegra without success. I've managed to compile my programs against nvidia binary GLESv2 and EGL libraries, but with official Khronos developer headers. Now, I want to clarify whether those are sufficient, or are there some NVidia specifics which would not be correctly defined in Khronos files. One thing I'm especially interested is the list of OpenGL ES 2 extensions which are supposed to work with this driver release. In Khronos header files they are practically all defined to "1", meaning they are all supported.

2) Release notes say that the driver release supports EGL 1.3 with image extension, but the driver itself reports at runtime to be EGL 1.4 compliant. Which one is an error?

These clarifications will make my development tests more secure, but as I have managed to run already my benchmarks this is not a showstopper, so I will continue exploration. Any further info is appreciated.

GLSL shader compatibility on 'highp' between GLES and OpenGL4.1

JCommins — Fri, 23 Sep 2011 07:27:13 -0400

I am wondering if anyone could shed some light on the following:

In a fragment shader, the following compiles on GLES2:

varying highp float lightIntensity;

However, compiling it on an OpenGL 4.1 system (Linux in this case, with nVidia 280.x drivers supporting OpenGL4.1) gives this error:

warning C7022: unrecognized profile specifier "highp"

I have seen that using the #ifdef GLES precision highp float; #endif (and removing highp from the offending line) resolves the problem - however, I have read that OpenGL4.1 can run GLES2 programs without modification. Am I missing anything here or is that #define a compulsory requirement?

Regards,
Jose.

EDIT: Moved to Mobile Category

black textures after 2 "battery low" messages

dsuoch — Thu, 13 Oct 2011 14:37:43 -0400

hello,

our testers reported that on 2 consecutive "battery low" messages some textures are reloaded in black. i thought our interrupt handling works, because on every other interrupt all our resources are successfully destroyed and recreated ...
but this seems like invalid textures because of lack of memory, right?!

my phones:
LG star optimus 2x
firmware 2.2.1 and 2.2.2

any input on how we could tackle this issue?
thanks in advance!

Math on the shaders; How to do; Tegra2

GreenTroll — Fri, 14 Oct 2011 03:37:56 -0400

Hi guys!

Could you help me, please.
I need to use shaders for direct math computing. For example, my function gets array with numbers and has to return arithmetic mean.
How can I do it? Any examples with shaders and math?

All I've found out it descriptions how use high level function from library OpenGL. But I need to realized low level math function to increase performance already existed "C" soft.
Nothing 3D object, textures and something like that!

Thanks

Tegra 2/Android screenshot capability

orrebmas — Thu, 06 Oct 2011 16:46:35 -0400

Is there a way to programmatically (?spell?) take screenshots on a Tegra 2 based platform?
I've searched everywhere for answers to this, and the answers I found only mention that Tegra 2 does not maintain a framebuffer in Android like other SoCs. If the framebuffer cannot be dumped to save a screenshot, what is the best way to go about this? A link to code sample will be amazing too.

Thanks.

Android Tegra Driver Crash

workless — Mon, 26 Sep 2011 06:29:06 -0400

Hi, I'm an android app developer and I'm having problems with tegra phones: they just crash.

09-25 21:00:27.255 994 994 I DEBUG : #00 pc 0000e816 /system/lib/egl/libGLESv1_CM_tegra.so
09-25 21:00:27.255 994 994 I DEBUG : #01 pc 0000e988 /system/lib/egl/libGLESv1_CM_tegra.so (glDisable)

The game is fully tested on x86(with opengl) using valgrind(memcheck ptrcheck DTD) and gDEBugger.

All other phones don't have stability issue, I was wondering if:
1) Is there any android native tools that i may use to debug openglES1?
2) Any known issue in the openglES1 driver of tegra devices?
3) Any advise is welcome! :)

Thanks for spending your time helping me :)

PerfHUD ES displays one fullscreen quad and only a couple of draw calls

Andrew Edelsten — Wed, 21 Sep 2011 13:05:54 -0400

When I run PerfHUD ES and examine my OpenGL ES application running under Android, all I see in PerfHUD is one giant quad that takes up the entire screen and the Frame Debugger only shows two draw calls.

What's going on?

android/tegra - binary shaders - example? shader compiler?

dsuoch — Fri, 09 Sep 2011 10:16:58 -0400

hello,
i am currently working on a LG Star SU-660 w nvidia tegra hardware.

now i would like to include precompiled binary shaders and check the speedup.
according to the tegra opengles2 development guide it seems you need to be very careful
about a few things ... so is there an example available?
also, where from do i get the shader compiler? there was none in the SDK ...

thx,
Dietmar

What's the maximum polygon count for a Tegra game?

Andrew Edelsten — Mon, 22 Aug 2011 20:35:24 -0400

What's the highest number of polygons Tegra can handle in a 3D scene?

Why don't the NVIDIA Android samples work in the Android emulator?

Andrew Edelsten — Fri, 19 Aug 2011 16:42:10 -0400

I'm trying to run the samples in the NVIDIA Android samples pack that I downloaded from the Tegra Resources page. When I try to run some of the apks, I get a runtime error. Why is that?