[App + Src + Game] Test OpenGL and OpenVG performance; Game using OpenGL

[heliosdev]

Test OpenGL performs different tests to show the 3d rendering performance on PPCs. It cycles through all tests by rendering 500 frames.

The tests are:
1 alpha texture: eight textured (128x256 RGBA) rotating rectangles with alpha testing
2 torus light: torus with 3k triangles and one lightsource (visual glitches)
3 color blending: blending of four rectangles and a 256x256 texture
4 sub texture: copy a 15x15 rgba texture into a 128x128 rgba texture

Currently, there are discussions (http://forum.xda-developers.com/showthread.php?t=496528) on different drivers.

The images show the four test results on a HTC Touch HD (Blackstone) running the latest official HTC ROM.
Please post your frame rates and the model/ROM running this test.
More tests to come.

----------------Updates-----------------
18 January 2010
print information about available hardware acceleration in OpenVG post #47 (http://forum.xda-developers.com/showpost.php?p=4305005&postcount=47)

17 January 2010
added TestOpenGLVGResolutionHD2.cab to test OpenVG on new devices like HD2 post #47 (http://forum.xda-developers.com/showpost.php?p=4305005&postcount=47)

5 November 2009
Watch the test running on a HTC HD and a HTC HD2! video (http://www.youtube.com/watch?v=2htj8r1gWgc)

10 August 2009
different resolutions test added post #47 (http://forum.xda-developers.com/showpost.php?p=4305005&postcount=47)

24 June 2009
OpenVG test added post #33 (http://forum.xda-developers.com/showpost.php?p=4016546&postcount=33)

17 May 2009
OpenGL game 'Lights On' added
Turn all tiles from black to yellow. Clicking on a tile turns it and all its adjacent neighbours (up/down/left/right)
It starts with a 2x2 grid and automatically grows after completion.
Anybody solved the 6x6 grid? :confused:

09 May 2009
source code released (C# .net cf 2.0)
to add a new test add a new class (in namespace OpenGLTest), implement IOpenGLTest and add a new instance to the list _openglTests in MainForm.InitGL().
To run the app in the emulator copy the file libGLES_CM.dll (software implementation (http://www.vincent3d.com/Vincent3D/software/software.html)) to /windows or the deployment directory of the emulator (not on the device!).
The managed OpenGL ES wrapper, Font and Texture classes are from here (http://www.koushikdutta.com/search/label/Managed%20OpenGL%20ES).
The Blackstone texture is a print screen of the xda main page (with additional alpha layer):D
The text texture is out of one of the OpenGL ES specs (http://www.khronos.org/developers/specs/)

08 May 2009
more accurate performance measurement (taking screen size into account)
show report when closing the test (performance and OpengGL ES information)

02 May 2009
initial version

[incubus26jc]

says its unable to choose config. anyone know why

[uniqueboy]

Hi there, I am using a Diamond with my favourite rom Vic 3.4, 3.7 is old but for me this was best, I have pissed about a bit with some of the the ati drivers found in the thread linked above and these are my scores coresponding to your screenies.:)
1:68
2:21
3:78
4:49

[heliosdev]

incubus26jc, you'll need libgles_cm.dll in the /windows directory. Try to find the one for your device.
uniqueboy, not bad ;)

[heliosdev]

current results on Blackstone (stock rom)

Results
--------------------------------
alpha texture (128x256 rgba)
30340 kPixels/s

torus (3k triangles) one light
6272 kPixels/s

color blending
31498 kPixels/s

sub texture (128x128 15x15)
10762 kPixels/s

OpengGL ES info
--------------------------------
Vendor
QUALCOMM, Inc.

Version
OpenGL ES-CM 1.0

Renderer
Q3Dimension MSM7500W 01.02.03 0 5.1.2

Extensions
GL_OES_byte_coordinates
GL_OES_read_format
GL_OES_fixed_point
GL_OES_single_precision_commands
GL_OES_compressed_paletted_texture
GL_OES_matrix_palette
GL_OES_point_sprite
GL_OES_point_size_array
GL_ARB_vertex_buffer_object
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirrored_repeat
GL_EXT_stencil_wrap
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_blend_func_separate
GL_EXT_blend_equation_separate
GL_ATI_texture_compression_atitc
GL_ATI_extended_texture_coordinate_data_formats
GL_OES_draw_texture

[adrianmak]

mine result without any graphic performance dll added.
HD English stock rom

Results
--------------------------------
alpha texture (128x256 rgba)
20260 kPixels/s

torus (3k triangles) one light
5683 kPixels/s

color blending
24444 kPixels/s

sub texture (128x128 15x15)
10449 kPixels/s

OpengGL ES info
--------------------------------
Vendor
QUALCOMM, Inc.

Version
OpenGL ES-CM 1.0

Renderer
Q3Dimension MSM7500W 01.02.03 0 5.1.2

Extensions
GL_OES_byte_coordinates
GL_OES_read_format
GL_OES_fixed_point
GL_OES_single_precision_commands
GL_OES_compressed_paletted_texture
GL_OES_matrix_palette
GL_OES_point_sprite
GL_OES_point_size_array
GL_ARB_vertex_buffer_object
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirrored_repeat
GL_EXT_stencil_wrap
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_blend_func_separate
GL_EXT_blend_equation_separate
GL_ATI_texture_compression_atitc
GL_ATI_extended_texture_coordinate_data_formats
GL_OES_draw_texture

[NuShrike]

Any source code available?

[heliosdev]

source code released!

[shadowmike]

When you say "more accurate performance measurement (taking screen size into account)" Does this auto detect screensize for phones like the x1 that has a thinner screen then the HD or is it set to the HD screen size to measure performance.

[chalid]

SE X1


Results
--------------------------------
alpha texture (128x256 rgba)
24313 kPixels/s

torus (3k triangles) one light
7374 kPixels/s

color blending
24279 kPixels/s

sub texture (128x128 15x15)
11585 kPixels/s

[heliosdev]

When you say "more accurate performance measurement (taking screen size into account)" Does this auto detect screensize for phones like the x1 that has a thinner screen then the HD or is it set to the HD screen size to measure performance.

Yes, it uses the current screen size. It takes the average fps after rendering 50 frames and multiplies it by the number of pixels (screen size width x height).

[NuShrike]

source code released!Thanks! I'll see if I have the time to produce a C++ optimized version.

Minor improvements I can suggest for the torus test. :)

[malloth]

heliosdev, do You use diffrent thread for pixels/s counter?
I have made this test and each time the result label changed I got a little slowdown.

I tried it on XDA Atom Life with software rendering (it works on devices without gfx accelerator :D) and got those results:


Results
--------------------------------
alpha texture (128x256 rgba)
1128 kPixels/s

torus (3k triangles) one light
666 kPixels/s

color blending
1039 kPixels/s

sub texture (128x128 15x15)
954 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Hans-Martin Will

Version
OpenGL ES-CL 1.1

Renderer
Software

Extensions
GL_OES_fixed_point
GL_OES_single_precision
GL_OES_read_format
GL_OES_query_matrix
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_compressed_paletted_texture

[heliosdev]

No threading. It's because of (slow) software rendering. As I'm updating the performance number every 50 frames you don't see the 'slowdown' during these 50 frames because the number doesn't change!;)
Btw, looking at your results, also in software rendering the torus test is the evil one!:D

[malloth]

Yeah, torus test is a masacre for my 640MHz XScale CPU alone :).
I'm waiting for GSmart S1204. When it will appear my benchmark will be simillar to that of Diamond (I hope).

[NeoS]

Results
--------------------------------
alpha texture (128x256 rgba)
25128 kPixels/s

torus (3k triangles) one light
4701 kPixels/s

color blending
27114 kPixels/s

sub texture (128x128 15x15)
9299 kPixels/s

OpengGL ES info
--------------------------------
Running on my HD with OpenVG driver, without ATI driver.

Results
--------------------------------
alpha texture (128x256 rgba)
25630 kPixels/s

torus (3k triangles) one light
4793 kPixels/s

color blending
27267 kPixels/s

sub texture (128x128 15x15)
9300 kPixels/s

OpengGL ES info
--------------------------------

Running on my HD with OpenVG and ATI driver.

[sc00basteve]

Results
--------------------------------
alpha texture (128x256 rgba)
24159 kPixels/s

torus (3k triangles) one light
6188 kPixels/s

color blending
22541 kPixels/s

sub texture (128x128 15x15)
12201 kPixels/s

OpengGL ES info
--------------------------------
Vendor
QUALCOMM, Inc.

Version
OpenGL ES-CM 1.0

Renderer
Q3Dimension MSM7500W 01.02.03 0 5.1.2

Extensions
GL_OES_byte_coordinates
GL_OES_read_format
GL_OES_fixed_point
GL_OES_single_precision_commands
GL_OES_compressed_paletted_texture
GL_OES_matrix_palette
GL_OES_point_sprite
GL_OES_point_size_array
GL_ARB_vertex_buffer_object
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirrored_repeat
GL_EXT_stencil_wrap
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_blend_func_separate
GL_EXT_blend_equation_separate
GL_ATI_texture_compression_atitc
GL_ATI_extended_texture_coordinate_data_formats
GL_OES_draw_texture

[heliosdev]

Added a small game using OpenGL.

[uniqueboy]

Thanks for the game I will be playing it shortly, 6x6, bah I hope to do better than that!:)

[ponchofiesta]

HTC Touch Diamond2 (Default ROM, German)

Results
--------------------------------
alpha texture (128x256 rgba)
20323 kPixels/s

torus (3k triangles) one light
8422 kPixels/s

color blending
19615 kPixels/s

sub texture (128x128 15x15)
12010 kPixels/s

OpengGL ES info
--------------------------------
Vendor
QUALCOMM, Inc.

Version
OpenGL ES-CM 1.0

Renderer
Q3Dimension MSM7500W 01.02.03 0 5.1.2

Extensions
GL_OES_byte_coordinates
GL_OES_read_format
GL_OES_fixed_point
GL_OES_single_precision_commands
GL_OES_compressed_paletted_texture
GL_OES_matrix_palette
GL_OES_point_sprite
GL_OES_point_size_array
GL_ARB_vertex_buffer_object
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirrored_repeat
GL_EXT_stencil_wrap
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_blend_func_separate
GL_EXT_blend_equation_separate
GL_ATI_texture_compression_atitc
GL_ATI_extended_texture_coordinate_data_formats
GL_OES_draw_texture

[sc00basteve]

http://forum.ppcgeeks.com/showpost.php?p=909692&postcount=10

Just got a replacement Sprint Touch Pro. Stock rom nothing additional installed:

Results
--------------------------------
alpha texture (128x256 rgba)
33903 kPixels/s

torus (3k triangles) one light
6656 kPixels/s

color blending
27930 kPixels/s

sub texture (128x128 15x15)
13493 kPixels/s

[heliosdev]

That's an improvement!
What is the version/renderer reported by the tool?

[sc00basteve]

That's an improvement!
What is the version/renderer reported by the tool?

IDK. I just quoted user "nswenson"
Thats why I put the link up for everyone.
It is a significant difference though, right?

[xeirwn]

Tried this on my Flame and Omnia.

On Flame nothing works (I get error).

On Omnia runs (bad..) with dll from 'Sensory Overload' game.

[heliosdev]

You'll need libGLES_CM.dll in the windows directory! Have a look in the report at the line 'Renderer'. If it says 'Software' then this is the reason for the bad performance and you should replace it by the one appropriate for your device. I'm sure this is available for the Omnia, no idea about the Flame.

[johnpatcher]

What is "lightson.cab" good for?

[heliosdev]

Just a small game using OpenGL. (see first post for 'instructions')

[cadenza]

here my results with blackstone, topix 2.1.1. Im impressed, I really like this! Keep working;) greets, cad

[NuShrike]

Thanks! I'll see if I have the time to produce a C++ optimized version.Slightly faster torus test in C/C++ (and some ATI-specific acceleration) running ~ 22fps for the 4500 vertices on my FUZE.

I also built a version in Qt 4.5.x, but that ran @ ~15 fps with the same torus drawing code.

[heliosdev]

Thanks! I'm getting an average of 18fps. (but 25% more pixels)
You should increase the distance of the near/far planes to avoid geometry clipping.

[NuShrike]

Thanks! I'm getting an average of 18fps. (but 25% more pixels)
You should increase the distance of the near/far planes to avoid geometry clipping.I tried setting projection mode to exactly the way you had it, and it dropped FPS a bit. I tried another version with projection mode in perspective mode and best I can get is about ~17fps.

Have to say what you have in .NET is about the best it gets, and are we hitting the maximum theoretical of about 3-4 million triangles per second as advertised?

So anyways, I'm done exploring the chip/driver limits. New apps is my focus now.

[NeoS]

I tried setting projection mode to exactly the way you had it, and it dropped FPS a bit. I tried another version with projection mode in perspective mode and best I can get is about ~17fps.

Have to say what you have in .NET is about the best it gets, and are we hitting the maximum theoretical of about 3-4 million triangles per second as advertised?

So anyways, I'm done exploring the chip/driver limits. New apps is my focus now.
That's a bummer to hear that NuShrike. You've done some great stuff in this area. Good luck on your next projects!

[heliosdev]

This is an update with an OpenVG (http://www.khronos.org/openvg/) test. This test shows that using vector graphics the visual appearance of the geometric objects is independent of the scaling. The circles in the screenshots are rendered at different scale.
The test checks if OpenVG is available (/windows/openvg.dll) and the user can choose to run the OpenVG or the 'normal' OpenGLES test.
Dirvers for MSM720X devices can be found here (http://forum.xda-developers.com/showthread.php?t=496528).
I'm getting ~16500 kPixels/s.

edit: added new test showing some animated images in frames. Texture loading takes some seconds so be patient.:D

[LordDeath]

Hi,

in new new version I see little black lines in the torus test. It doesn't matter if I try it with the neo driver pack or without. I am using a Touch HD with the Stock ROM.

Here the results without a driver:
alpha texture (128x256 rgba)
32830 kPixels/s

torus (3k triangles) one light
5663 kPixels/s

color blending
34179 kPixels/s

sub texture (128x128 15x15)
10845 kPixels/s

Here with the neo driver pack:

alpha texture (128x256 rgba)
33572 kPixels/s

torus (3k triangles) one light
5482 kPixels/s

color blending
34671 kPixels/s

sub texture (128x128 15x15)
11016 kPixels/s

But there is an error in the OpenGLVG test with the neo driver pack:

TestOpenGL.exe
MissingMethodException
Der Einstiegspunkt 'eglBindAPI' in der PInvoke-DLL 'libgles_cm.dll' wurde nicht gefunden.

bei TestOpenGL.MainForm..ctor()
bei TestOpenGL.Program.Main()

[NeoS]

Ah. We're currently working on that. The team is currently testing a new driverpack. Consider that a Beta 2.

[heliosdev]

The libgles_cm.dll is missing a function for switching to OpenVG! Was the installation successful? What is the size of this dll in Bytes?

[yozgatg]

here's some results from my side (using a test build of Gen.Y D2 R1.5), this is on a GSM Touch Diamond (VGA)
i'm using Rhodium (stock ROM) video drivers, combined with the ATI D3D driver

Results
--------------------------------
alpha texture (128x256 rgba)
29390 kPixels/s

torus (3k triangles) one light
5029 kPixels/s

color blending
29587 kPixels/s

sub texture (128x128 15x15)
10337 kPixels/s

OpenVG
10240 kPixels/s

[LordDeath]

The libgles_cm.dll is missing a function for switching to OpenVG! Was the installation successful? What is the size of this dll in Bytes?

The installation was successful. I have a \Windows\libgles_cm.dll with the size of 330488 bytes from the 6th May 09.
But I have replaced the htcapi.dll with a 4KiB version to prevent Opera Mobile freezes.

[heliosdev]

Yes, this dll doesn't support the functionality to switch to OpenVG. Just a little patience, Neo will update the installation cab 'soon'!

[andrew-in-woking]

Hi

OpenVG seems to work "out of the box" on my TP2.

Cheers

andrew-in-woking

[Art1sta]

alpha texture
111.000


torus 3k triangles
85.000

color blending
153.000

sub texture
103.000

Cool no? :d

[heliosdev]

ok, currently 1st place! (by far)
But as we know the TG01 has the new snapdragon, but it's promising to see that the device is fast. Have you tried the OpenVG test? See post #33. (it will ask to perform the OpenVG tests if it is available)
For both tests: what does the report show for vendor/renderer/version?
Thanks!

[SinSilla]

More Reults of the TG01


Test OpenGL ES by heliosdev@hotmail.com
Results
--------------------------------
alpha texture (128x256 rgba)
106283 kPixels/s

torus (3k triangles) one light
85221 kPixels/s

color blending
72826 kPixels/s

sub texture (128x128 15x15)
92279 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Renderer
AMD Z430

Version
OpenGL ES-CM 1.1

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

OpenVG Test not possible in the Moment

[lunakirb]

it should be run using the stock rom?
im using a cooked rom for blackstone can i test it:confused:

[heliosdev]

I'm using it on the latest stock rom. You have to install the latest driver pack for better performance and OpenVG support (check first post of discussions thread link in first post)! Some cooked roms have these drivers included (not sure which version).

[lunakirb]

I'm using it on the latest stock rom. You have to install the latest driver pack for better performance and OpenVG support (check first post of discussions thread link in first post)! Some cooked roms have these drivers included (not sure which version).

i hve the driver pack installed already thank u.

[heliosdev]

Runs the OpenGLES tests in different resolutions.
Issue: By changing the rendering area I also wanted to restrict the clearing of the screen to this area. Check 'region update (scissor)' to enable this. On my current setup this gives weird results (part of manila and render time is rendered in the area outside the viewport). Anyone else having this behavior?

P.S.: The stress test was just a quick viewport (rendering area) test, but I decided to let it in.:D

update: Both versions will print information about available hardware acceleration in OpenVG into the results file (only if there is any).

[Howell]

My TG01 compared with others devices posted here :
http://www.xtupload.com/new/image-397C_4AC60C56.jpg (http://www.xtupload.com/new/share-397C_4AC60C56.html)

[Jarc112]

can somebody help me i not much of a pro in this, i was trying to run the video test but i get an error libgles_cm.dll, i found the one on my device, i try place the file in the open gl folder but i get the same error, by the way i have a treo pro

[heliosdev]

You don't have to copy any dll. You have the libgles_cm.dll in the /windows directory, right? Does an error message come up? And if so, what does it say when you click on 'Details'?

[Jarc112]

it says

TestOpenGL.exe
MissingMethodException
Cant find PInvoke DDL "libgles_cm.dll",

at
TestOpenGL.OpenGL.egl.GetDisplay()
at TestOpenGl.MainForm.ctor()
at TestOpenGL.Program.Main()

[deniur]

Hi,

I installed the two cabs, TestOpenGL and LightsOn.
However, both of them give error when I run.
The error is just a blank white page with the word Error on top left corner, nothing else.
Does this means Omnia II doesn't have OpenGL support?
By the way, I'm on WM6.1.

best regards

[heliosdev]

Don't know about the OpenGL support of Omnia II. Have a look at OpenGL ES 3D drivers, v1 compatibility layer (http://www.modaco.com/content/i8000-omnia-ii-http-i8000-modaco-com/291735/opengl-es-3d-drivers-v1-compatibility-layer/)! Maybe you'll get some information there.

[NuShrike]

The S3C6410 GPU in the Omnia II and Acer M900, among others, have still demonstrated sub-par performance so far for OGLES v1.1 emulation in the current Modaco builds.

You can probably conclude the Omnia II has much weaker video hardware/drivers than the original Omnia. This is probably because Samsung is Gen 1 (ONE) with their new hardware whereas ATI/Qualcomm is in Gen 12 and started off with Bitboys magic. Who knows what gen Tegra is, but it's from a known name in the graphics field, unlike Samsung/FIMG/HUONE.

[leoangel1267]

Results
--------------------------------
alpha texture (128x256 rgba)
142400 kPixels/s

torus (3k triangles) one light
106944 kPixels/s

color blending
149996 kPixels/s

sub texture (128x128 15x15)
142785 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Version
OpenGL ES-CM 1.1

Renderer
AMD Z430

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

[RDT]

Heh, I made a chart on which is compared results in Test OpenGL of older Qualcomm MSM720X with Snapdragon.

http://stahovat.okamzite.eu/novinky/Snapdragon/Snapdragon_vs_MSM720X.png

It's quite funny :D.

The whole article about comparison is here (http://stahovat.okamzite.eu/news.php?readmore=947).

[donbing]

TG01 on WM6.5

Results
--------------------------------
alpha texture (128x256 rgba)
103166 kPixels/s

torus (3k triangles) one light
95698 kPixels/s

color blending
132292 kPixels/s

sub texture (128x128 15x15)
92703 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Version
OpenGL ES-CM 1.1

Renderer
AMD Z430

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

[AceXNew]

my ultimate 6150 gave ~1200 kPixels/s for the first two tests and the 4th one !
~900 for 3rd,
it is really slow ? isn't it ?


i must upgrade my device....which device should i get ? with which CPU ?
thanks !

Edit:
and just for being documented:

Results
--------------------------------
alpha texture (128x256 rgba)
1237 kPixels/s

torus (3k triangles) one light
1197 kPixels/s

color blending
918 kPixels/s

sub texture (128x128 15x15)
1062 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Hans-Martin Will

Version
OpenGL ES-CL 1.1

Renderer
Software

Extensions
GL_OES_fixed_point
GL_OES_single_precision
GL_OES_read_format
GL_OES_query_matrix
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_compressed_paletted_texture

is there any settings i can tune ?
i mean it says renderer is Software , can i switch to hardware ? (will it be faster ?)

[heliosdev]

Well, it's slower than the other reports because your device is using software rendering (http://www.vincent3d.com/Vincent3D/software/ogles1/ogles1.html)! Can't tell you if there is any OpenGLES driver for you device. AFAIK this device uses an 'Intel Bulverde 520 MHz processor' and you can search if there is any hardware acceleration available.
A new device? Well, have a look at posts #55-57 or wait for a device using an NVidia Tegra chip. (I'm still not convinced that the Tegra chip (http://www.youtube.com/watch?v=rQa9nP4yyms) will/can be build in a thin device! Any pictures? Any phones?)

[AceXNew]

Well, it's slower than the other reports because your device is using software rendering (http://www.vincent3d.com/Vincent3D/software/ogles1/ogles1.html)! Can't tell you if there is any OpenGLES driver for you device. AFAIK this device uses an 'Intel Bulverde 520 MHz processor' and you can search if there is any hardware acceleration available.
A new device? Well, have a look at posts #55-57 or wait for a device using an NVidia Tegra chip. (I'm still not convinced that the Tegra chip (http://www.youtube.com/watch?v=rQa9nP4yyms) will/can be build in a thin device! Any pictures? Any phones?)
oh i forgot,
imate 6150 carries a "NVIDIA GoForce 5500 graphics accelerator." but AFAIK it has no good driver for using it, or in other words, it is not tweaked to make use of it as it should be.

P.S: today i saw brouchure of the Xperia X10 device. with a 1Ghz ! Snapdragon cpu. it is quite a techi.

[heliosdev]

Updated first post with a link to a video running this test on a HD and HD2 (Snapdragon).

[bteljuice]

so this is only for testing. It does not improve performance nor graphics like the DriverPack3.

[heliosdev]

Yes, just for testing. Has been extended with OpenVG tests as this is available in the driverpack.

[tplewa]

Touch HD (Stock ROM + Driverpack 3)


Results
--------------------------------
alpha texture (128x256 rgba)
33752 kPixels/s

torus (3k triangles) one light
6390 kPixels/s

color blending
37981 kPixels/s

sub texture (128x128 15x15)
10163 kPixels/s

OpengGL ES info
--------------------------------
Vendor
QUALCOMM, Inc.

Version
OpenGL ES-CM 1.0

Renderer
Q3Dimension MSM7500W 01.02.03 0 5.1.2

Extensions
GL_OES_byte_coordinates
GL_OES_read_format
GL_OES_fixed_point
GL_OES_single_precision_commands
GL_OES_compressed_paletted_texture
GL_OES_matrix_palette
GL_OES_point_sprite
GL_OES_point_size_array
GL_ARB_vertex_buffer_object
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirrored_repeat
GL_EXT_stencil_wrap
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_blend_func_separate
GL_EXT_blend_equation_separate
GL_ATI_texture_compression_atitc
GL_ATI_extended_texture_coordinate_data_formats
GL_OES_draw_texture

[NeoS]

Here's my benchmark on a 28008 WM 6.5 custom ROM.

Test OpenGL ES by heliosdev@hotmail.com
Results (480x764 )
--------------------------------
alpha texture (128x256 rgba)
21536 kPixels/s

torus (3k triangles) one light
6576 kPixels/s

color blending
18801 kPixels/s

sub texture (128x128 15x15)
7607 kPixels/s

OpengGL ES info
--------------------------------
Vendor
QUALCOMM, Inc.

Renderer
Q3Dimension MSM7500W 01.02.03 0 5.1.2

Version
OpenGL ES-CM 1.0

Extensions
GL_OES_byte_coordinates
GL_OES_read_format
GL_OES_fixed_point
GL_OES_single_precision_commands
GL_OES_compressed_paletted_texture
GL_OES_matrix_palette
GL_OES_point_sprite
GL_OES_point_size_array
GL_ARB_vertex_buffer_object
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirrored_repeat
GL_EXT_stencil_wrap
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_blend_func_separate
GL_EXT_blend_equation_separate
GL_ATI_texture_compression_atitc
GL_ATI_extended_texture_coordinate_data_formats
GL_OES_draw_texture


See that Windows Mobile 6.5's driverbase is REALLY bad? If you want performance: Stick with stock roms. That's all I can say for now.

[stylez]

Touch Pro \ RSW ROM

ATI & Driver Pack 3

Results
--------------------------------
alpha texture (128x256 rgba)
18708 kPixels/s

torus (3k triangles) one light
4744 kPixels/s

color blending
20234 kPixels/s

sub texture (128x128 15x15)
11544 kPixels/s

OpengGL ES info
--------------------------------
Vendor
QUALCOMM, Inc.

Version
OpenGL ES-CM 1.0

Renderer
Q3Dimension MSM7500W 01.02.03 0 5.1.2

Extensions
GL_OES_byte_coordinates
GL_OES_read_format
GL_OES_fixed_point
GL_OES_single_precision_commands
GL_OES_compressed_paletted_texture
GL_OES_matrix_palette
GL_OES_point_sprite
GL_OES_point_size_array
GL_ARB_vertex_buffer_object
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_env_crossbar
GL_ARB_texture_mirrored_repeat
GL_EXT_stencil_wrap
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_blend_func_separate
GL_EXT_blend_equation_separate
GL_ATI_texture_compression_atitc
GL_ATI_extended_texture_coordinate_data_formats
GL_OES_draw_texture

[tsamolotoff]

Hmm, there're some questions:

1. I happened to be too dumb to find out where is the results.txt file located. So please help the n00b :)

2. there's libopenvg.dll present in the \Windows directory, however OpenVG-enabled test is not enabled in the program.
Well, pressing "<" on my neoTouch helped me... Seems that title-bar tool closed the program incorrectly.

My results are attached below. Tested it without chainfire's hd2 patch. Seems to be a little bit stabler.

[heliosdev]

Added a new version which should enable the OpenVG test on newer devices like the HD2. Can you try it as I don't have such a device?

[saeros]

Tested the OpenVG part with HTC HD 2 and here are the results.

When running the tests I got a small window after clicking run. EGL init 12301. After clicking OK the tests started running.

[heliosdev]

Thanks for testing! The egl error 12301 is EGL_BAD_SURFACE. Can't debug without an HD2 device...
Your results are too low (~factor 4) for an HD2! Have you tried HD2/Leo 3D Driver Patch (http://forum.xda-developers.com/showthread.php?t=592663)?

[jug6ernaut]

Results (480x748 )
--------------------------------
alpha texture (128x256 rgba)
95837 kPixels/s

torus (3k triangles) one light
55426 kPixels/s

color blending
68766 kPixels/s

sub texture (128x128 15x15)
48736 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Renderer
AMD Z430

Version
OpenGL ES-CM 1.1

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

[fards]

Results (480x748 )
--------------------------------
alpha texture (128x256 rgba)
95837 kPixels/s

torus (3k triangles) one light
55426 kPixels/s

color blending
68766 kPixels/s

sub texture (128x128 15x15)
48736 kPixels/s


what drivers are you using with that? chainfires patch?
any chance of sharing them if they are unique so I can take a look?

and off topic(ish) what glbench 1.1 score do you get?

[fards]

Thanks for testing! The egl error 12301 is EGL_BAD_SURFACE. Can't debug without an HD2 device...
Your results are too low (~factor 4) for an HD2! Have you tried HD2/Leo 3D Driver Patch (http://forum.xda-developers.com/showthread.php?t=592663)?

I'm getting pretty much the same scores on my HD2 using chainfires patch.. :eek:
something odd going on, as I get one of the top glbench scores..

[tsamolotoff]

I've tried the openvg test, it works, however the message 'egl init' is in here... Works good

The results (Acer S200):
Image Frame
23713 kPixel/s

Circles
84686 kPixel/s

OpenVG info

Vendor
Hooked Wireless

Renderer
OpenVG 1.1 GLES Accelerated implementation

Extensions
none

[tsamolotoff]

It seems that OpenGLES test works better if started after a soft-reset... 110k scores if the device is 'fresh', and around 60-80 if it's used for quite a while

[heliosdev]

Thanks! Manila/Sense needs some OpenGL resources, too. (You can try to disable these home elements and after a soft reset the drop should disappear.)
Oh well, just ran the OpenVG test on my HD and I'm getting (percent of your results)
Image Frame
4436 kPixel/s (19%)

Circles
11035 kPixel/s (13%)

What's the time it takes to start the OpenVG test? (on HD ~5sec)
I should add some additional driver queries to see if anything is really gpu accelerated or if it's just the stronger cpu.

[tsamolotoff]

Thanks! Manila/Sense needs some OpenGL resources, too. (You can try to disable these home elements and after a soft reset the drop should disappear.)
I don't have any of shells (just the default acer homescreen, it's really simple)

What's the time it takes to start the OpenVG test? (on HD ~5sec)
Hmm, I can say that it's instant. maybe 1 second or less :)

[heliosdev]

Ah, Acer S200. Just updated the tests (http://forum.xda-developers.com/showpost.php?p=4305005&postcount=47) to show information about hardware acceleration in OpenVG.
Any other Snapdragon(HD2, TG01, LG Expo) users around to test OpenVG and see if there is any output in the results regarding hardware acceleration?

[tsamolotoff]

Here you are:
Test OpenVG by heliosdev@hotmail.com
Results (480x748 )
--------------------------------
Image Frame
27888 kPixels/s

Circles
94258 kPixels/s

OpenVG info
--------------------------------
Vendor
Hooked Wireless

Renderer
OpenVG 1.1 GLES Accelerated Implementation Dec 12 2007

Version
1.1

Extensions
none
Nothing has changed, apparently.

As it's mentioned there Hooked Wireless (http://hookedwireless.com/OpenVG.html), it seems that OpenVG is done through GLES layer... So is it bad or good?

[heliosdev]

Thanks! Yes, OpenVG is using OpenGLES for its rendering! And this is also the same dll info I get on my HD. So, the OpenVG hardware queries aren't telling anything about the 'real' performance.
And the performance difference is in factors of 5-7 which is also the case for the OpenGLES test (comparing MSM720X to Snapdragon).
I had to keep the OpenVG tests simple as the performance on the HD isn't good. (espacially all the image processing functions are very slow) Maybe now I can add some more OpenVG tests for Snapdragon devices?
A test for OpenGLES 2.0 would be interesting, too... (only for Snapdragon, of course)

[fards]

I'm getting the same, for OpenVg except quite a lot lower scores on Hd2. I still haven't recooked without chainfires patch to see if that's effecting my scores.

[tsamolotoff]

I can test it with chainfire's fix, if you wish. However, this breaks landscape 2d acceleration on Acer (dunno why, but it gradually gets slower and slower until choppyness become unbearable)

[fards]

I can test it with chainfire's fix, if you wish. However, this breaks landscape 2d acceleration on Acer (dunno why, but it gradually gets slower and slower until choppyness become unbearable)

it's ok I've just recooked and flashed.. testing now

with Chainfires patch;

Test OpenVG by heliosdev
--------------------------------
Image Frame
23138 kPixels/s

Circles
23495 kPixels/s

OpenVG info
--------------------------------
Vendor
Hooked Wireless

Renderer
OpenVG 1.1 GLES Accelerated Implementation Dec 12 2007

Version
1.1

Extensions
none


without

Test OpenVG by heliosdev
Results (480x764 )
--------------------------------
Image Frame
43103 kPixels/s

Circles
147999 kPixels/s

OpenVG info
--------------------------------
Vendor
Hooked Wireless

Renderer
OpenVG 1.1 GLES Accelerated Implementation Dec 12 2007

Version
1.1

Extensions
none

[fards]

and here's the opengl test on my HD2 rom without the patch (and some minor tweaking)

Test OpenGL ES
Results (480x764 )
--------------------------------
alpha texture (128x256 rgba)
162321 kPixels/s

torus (3k triangles) one light
156538 kPixels/s

color blending
154982 kPixels/s

sub texture (128x128 15x15)
123294 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Renderer
AMD Z430

Version
OpenGL ES-CM 1.1

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat


Helios if you want me to get involved in testing or debugging stuff let me know.

[heliosdev]

fards, you took the lead in the OpenGLES and OpenVG tests!
So, you get these fast results with your own rom without any patches/drivers (only some 'minor' tweaks), right? You don't work for Qualcomm, do you?:D
Are the glbenchmark tests still high?

Thanks again to all for testing. More tests to come, demands/requests are welcomed!

[fards]

fards, you took the lead in the OpenGLES and OpenVG tests!
So, you get these fast results with your own rom without any patches/drivers (only some 'minor' tweaks), right? You don't work for Qualcomm, do you?:D
Are the glbenchmark tests still high?

Thanks again to all for testing. More tests to come, demands/requests are welcomed!

lol I wish.. maybe I'd get paid some proper money then, and we'd have proper drivers available :p

the glbench drops quite a bit without chainfires patch (but still beats the other devices).

I'm using the leaked 202 t-mob as a base.
so SPL XIP, Drivers, some tweaks, and all the debug and phone home crap ripped out from that rom..
It has a new kernel (oemxip is different) there seems to be some mild changes with SPL (ok not so mild) and the drivers are pretty decent..
I've not released my rom as there's a few too many steps to go through to get it up for most people.. till we get a 200Hardspl anyway.

Can't wait to get another leak on this build to see whether it's much faster .

You really do need to get a HD2 mate, it's a great device, when used properly :D


I should add here that I'm still trying to piece together which driver effects what and how changing one can effect the other, these snapdragon drivers make even less sense now I'm using them to what we thought before when we had the first leaked roms..

for example I dont understand why the subtexture is lower in mine, is that likely to be a 2d thing? is it the same EGL ?
what dl is that test likely to be calling ? or maybe what is the function? I may be able to find the dll..

[tsamolotoff]

2heliosdev:
I think he tested right after a SR, and it's the results of the first loop... (they decline gradually after each loop ) I can disable 2d blt, see if I get closer results

Any GLES 2.0 tests would be great... or something kewl for the OpenVG (such as animation-rich or video-rich benchmark)

Well, pwnage in 2 of 4 tests from Russia, riding an Acer S200 :p
Test OpenGL ES by heliosdev@hotmail.com
Results (480x748 )
--------------------------------
alpha texture (128x256 rgba)
142226 kPixels/s

torus (3k triangles) one light
139306 kPixels/s

color blending
167847 kPixels/s

sub texture (128x128 15x15)
178945 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Renderer
AMD Z430

Version
OpenGL ES-CM 1.1

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

[tsamolotoff]

P.S. I must say this is the default settings for Acer S200 (the previous tests were conducted with 2D.HwBlt=1 in yamato_panel.txt (maybe it's an acer-specific configuration file, I urge the HD2 and Expo owners to check this thing)), so no cheating whatsoever. However, I must say that the test, despite the higher score, was choppy (it was smooth on previous attempts with hw blt enabled)

[fards]

2heliosdev:
I think he tested right after a SR, and it's the results of the first loop... (they decline gradually after each loop ) I can disable 2d blt, see if I get closer results



I didn't on either account..

that was a retest this morning after swapping a file around, device had been on standby all night.

and the test ran for 3 minutes..

I do agree the results settle down, but I'm just reporting the file that was saved..

[fards]

P.S. I must say this is the default settings for Acer S200 (the previous tests were conducted with 2D.HwBlt=1 in yamato_panel.txt (maybe it's an acer-specific configuration file, I urge the HD2 and Expo owners to check this thing)), so no cheating whatsoever. However, I must say that the test, despite the higher score, was choppy (it was smooth on previous attempts with hw blt enabled)

I remember seeing that file in very early leaks of the HD2 rom.

havent seen it for a while.. mmm
interesting

WinGrep is searching for me :D ps you really need WinGrep it's awesome ! http://code.google.com/p/grepwin/

haven't found that file but there's definitely references to it in libpanel (found while searching for it with wingrep)

having grabbed one ( I think yours and dropped into root) it's interesting that I don't get the bitrate at the bottom when running heliosdev app, the same as when I tested the expo EGL mm

I also get 0 Kpixels in results across the board.. very very mmmmm

[jug6ernaut]

Heliodev an idea, right now these results really dont mean shit...as they are only showing the maximum obtained */sec.

The results would mean 100000x times more if there were the average of */second for the entire duration of each test.

just my opinion :)

[fards]

more experimenting with the panel.txt
deleting the swapmode makes no difference.

setting 2D.HWblt=0 causes a non booting rom here..

I need to find one of the really early HD2 beta leaks to see what the panel.txt was for that..

[tsamolotoff]

setting 2D.HWblt=0 causes a non booting rom here..
Hmm, it's the default setting for Acer S200, and this is that thing that made landscape mode very choppy and slow. Setting it to "1" made things definitely better, albeit with incompatibilty with chainfire's 3d patch and some slowdowns in 3d (electopia runs not as smooth as it was with blt=0)

[heliosdev]

...
You really do need to get a HD2 mate, it's a great device, when used properly :D

for example I dont understand why the subtexture is lower in mine, is that likely to be a 2d thing? is it the same EGL ?
what dl is that test likely to be calling ? or maybe what is the function? I may be able to find the dll..


I'm still hoping for a device like the HD(2) with an Nvidia Tegra(2) chip. But I don't think that this will be available in the next 1-2 years...

Nothing really special (from a developer's point of view) in the subtexture test, just calling glTexSubImage2D. Will check the internal texture formats (bit formats used for rgb like 565)


...
Any GLES 2.0 tests would be great... or something kewl for the OpenVG (such as animation-rich or video-rich benchmark)


noted :D

Heliodev an idea, right now these results really dont mean shit...as they are only showing the maximum obtained */sec.

The results would mean 100000x times more if there were the average of */second for the entire duration of each test.

just my opinion :)

Yes, that would be better, I've thought about that, too.;)
I'm not sure when to do such changes as there are now some reported results around.
In an upcoming GLES2.0 test this will be changed.

[jug6ernaut]

Yes, that would be better, I've thought about that, too.;)
I'm not sure when to do such changes as there are now some reported results around.
In an upcoming GLES2.0 test this will be changed.[/QUOTE]

woot? ogles2 test u say? that would be awesome :D. plus the results change.. Cant wait :D.

[tsamolotoff]

I'm still hoping for a device like the HD(2) with an Nvidia Tegra(2) chip. But I don't think that this will be available in the next 1-2 years...
As a former owner of a nvidia-'blessed' device (Toshiba G900, ftw), I may say that if it were possible (with Tegra announced for at least 2 years, only 2 devices are on the market, and zero PPC/Commie stuff), this thing will underperform badly, and will have issues with suspend :D

[fards]

As a former owner of a nvidia-'blessed' device (Toshiba G900, ftw), I may say that if it were possible (with Tegra announced for at least 2 years, only 2 devices are on the market, and zero PPC/Commie stuff), this thing will underperform badly, and will have issues with suspend :D


shhh you'll make people believe that tegra isn't the holy grail if you carry on like that :D
it does look like the early chipsets were rush-jobs from nvidia. either the tegra is a bear to get right or they are really taking their time getting it right..

I want that lovely looking tablet that people keep reusing the image of when referring to winmo7
http://wmpoweruser.com/wp-content/uploads/2009/05/wmconcept-thumb.png


anyway could you do some kind of trigger test in there as well helios?
I mean so we can measure if there's any framerate drop by applying a drag motion to the screen. quite a few of the MS7xxx had it, would be interest to see if it's measurable with the snapdragons as well.. I've not noticed any, but it would be interesting to see.
maybe a drag a rotating ball around a maze type thing, so we could monitor the rate?

[fards]

anyway tsamolotoff

can you let me have your libpanel.dll?

there's one here at 10kb, but there's also a
dcINONE_HTCPANEL.dll which is odd

[tsamolotoff]

Well, I'm leaving the city in two hours, so I suggest you to ask rafyvitto (he's a rom cooker for S200) to provide you with this library. Still, if you need specifically mine, I will be able to do it in two days. Sorry for the inconvenience. As of dc*** files, there're two of them - one apparently for AU Optotronics H380VL01 panel (seems that's the real panel), and the other one for some Sharp display.

[silvaros]

Looking at some of the other posts mine seems pretty slow. Is there something I can do to speed it up? I'm using the original rom on a Omnia 2 which is said to have a 800MHz processor and a dedicated graphics accelerator. I found a libGLESv1_CM.dll that was "modified" in Oct 09. I might try to change the source to use that file and see if it works.


Test OpenGL ES by heliosdev@hotmail.com
Results (480x748 )
--------------------------------
alpha texture (128x256 rgba)
1715 kPixels/s

torus (3k triangles) one light
1715 kPixels/s

color blending
1209 kPixels/s

sub texture (128x128 15x15)
1518 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Hans-Martin Will

Renderer
Software

Version
OpenGL ES-CL 1.1

Extensions
GL_OES_fixed_point
GL_OES_single_precision
GL_OES_read_format
GL_OES_query_matrix
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_compressed_paletted_texture

[heliosdev]

The 'Renderer' is 'Software' and that's the reason why you have low numbers. I'm binding to libGLES_CM.dll and not libGLESv1_CM.dll. You can change the code (even though it isn't the latest version) by adding the v1 in all DllImport declarations in gl.cs. In egl.cs change it to libEgl.dll. (check that this file exists in /windows directory!)
Also check the MainForm constructor if there is a check for libGLES_CM.dll.

[Chainfire]

ehm just a comment...

you cannot use any results from this program when the HD2 patch is installed, as it limits the number of frames that can be drawn per second, doh ;)

However I did experiment with the version of this that was current when I first wrote the test, and I can do all sorts of magical things with it... the numbers just dont mean anything, as without the patch, frames are still being skipped etc.

[Chainfire]

Hmm, it's the default setting for Acer S200, and this is that thing that made landscape mode very choppy and slow. Setting it to "1" made things definitely better, albeit with incompatibilty with chainfire's 3d patch and some slowdowns in 3d (electopia runs not as smooth as it was with blt=0)

Can you explain what happens, incompatibility?

[tsamolotoff]

Can you explain what happens, incompatibility?
Well, it gradually slows down with your patch, especially if display mode was changed several times while doing something else (e.g. honeycomb menu slowly, but inevitably become jerky, landscape tearing becomes visible, and so on...). After pressing suspend button, and some waiting, everything reverts to somewhat normal state, but then it goes as described above. Without the patch, the only apps that utilise 3d and suffer horrible performance are Opera 9.7 and GLBenchmark. All the other things I've tested (NFS UC, Xtrakt, Helios'es Benchmark, Tower Defence etc ) run smoothly and should I say, faster than with your patch. And I think that acer's d3dm_ati is identical to TG01s (at least size is roughly 1.03 M (I've not checked it) for both of them

[Chainfire]

Well, it gradually slows down with your patch, especially if display mode was changed several times while doing something else (e.g. honeycomb menu slowly, but inevitably become jerky, landscape tearing becomes visible, and so on...). After pressing suspend button, and some waiting, everything reverts to somewhat normal state, but then it goes as described above. Without the patch, the only apps that utilise 3d and suffer horrible performance are Opera 9.7 and GLBenchmark. All the other things I've tested (NFS UC, Xtrakt, Helios'es Benchmark, Tower Defence etc ) run smoothly and should I say, faster than with your patch. And I think that acer's d3dm_ati is identical to TG01s (at least size is roughly 1.03 M (I've not checked it) for both of them

Interesting. Perhaps for your device you need only the config patch and not the other parts of my patch ... which are I admit still a bit shaky (haven't found a truely good way to do the fix, so it's all "kind-of works" pieces)

[june57]

Results
--------------------------------
alpha texture (128x256 rgba)
9607 kPixels/s

torus (3k triangles) one light
11186 kPixels/s

color blending
12024 kPixels/s

sub texture (128x128 15x15)
8467 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Samsung Electronics

Version
2.0 (lib version 20.02.24)

Renderer
FIMG

Extensions
GL_OES_matrix_palette
GL_OES_framebuffer_object
OES_read_format
OES_compressed_paletted_texture
OES_framebuffer_object
OES_stencil8
OES_texture_compression_S3TC
OES_mapbuffer
OES_texture_npot
GL_EXT_texture_compression_s3tc

[sc00basteve]

Did you ever figure out what Sprint had implemented in their drivers to make their TouchPro score so high? (back on page 3)
Just curious... I know you guys are in the middle of something

[tsamolotoff]

So, nothing new as of yet?

[heliosdev]

Did you ever figure out what Sprint had implemented in their drivers to make their TouchPro score so high? (back on page 3)
Just curious... I know you guys are in the middle of something

Well, in the meantime the results are in that region (e.g. results HD with v3 drivers (http://forum.xda-developers.com/showpost.php?p=5029165&postcount=64))

So, nothing new as of yet?

I've started with a gles2 test but stuck right now as I don't have such a device (not yet) and had to send the app to user fards for early alpha testing (thanks again!). Now I'm at a point where it 'should' work but the shader doesn't compile on the device.:confused:
Trying to set up a desktop project which uses a gles2 emulator dll.
I'm not sure what to prioritize more. Maybe still some more OpenVG tests and doing gles2 when I have such a device.

Any idea what to show in an 'animation-rich' benchmark?
A video benchmark would be very cool having gles2 with different image filters implemented in the shaders.:cool:

[tsamolotoff]

Well, I personally believe that GL2 benchmark is more of importance (and of use) than VG because there's at least one GL2-enabled game on the market, and zero non-OEM apps with openvg support...

As for 'animation-rich' I may suggest you to implement some effects (e.g. 'plasma', photoshop-like filters) which are applied to different regions of canvas, maybe with some video blending effects if it's possible (I don't know for sure, what mobile DShow does allow you to do with rendering). Anyway, thanks for the benchmark :)

[xos84292]

Looking at some of the other posts mine seems pretty slow. Is there something I can do to speed it up? I'm using the original rom on a Omnia 2 which is said to have a 800MHz processor and a dedicated graphics accelerator. I found a libGLESv1_CM.dll that was "modified" in Oct 09. I might try to change the source to use that file and see if it works.


Test OpenGL ES by heliosdev@hotmail.com
Results (480x748 )
--------------------------------
alpha texture (128x256 rgba)
1715 kPixels/s

torus (3k triangles) one light
1715 kPixels/s

color blending
1209 kPixels/s

sub texture (128x128 15x15)
1518 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Hans-Martin Will

Renderer
Software

Version
OpenGL ES-CL 1.1

Extensions
GL_OES_fixed_point
GL_OES_single_precision
GL_OES_read_format
GL_OES_query_matrix
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_compressed_paletted_texture



do your setup the opengl es hack 1.0/1.1?

i test at hack, it will have 8xxx/10xxx / 9xxx / 8xxx

[xos84292]

Results
--------------------------------
alpha texture (128x256 rgba)
9518 kPixels/s

torus (3k triangles) one light
10979 kPixels/s

color blending
12026 kPixels/s

sub texture (128x128 15x15)
8440 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Samsung Electronics

Version
2.0 (lib version 20.02.24)

Renderer
FIMG

Extensions
GL_IMG_texture_compression_pvrtc
OES_matrix_palette
GL_OES_framebuffer_object
GL_OES_read_format
GL_OES_compressed_paletted_texture
GL_OES_framebuffer_object
GL_OES_stencil8
GL_OES_texture_compression_S3TC
GL_OES_mapbuffer
GL_OES_texture_npot
GL_EXT_texture_compression_s3tc

I8000 with OpenGL ES 3D drivers, v1 compatibility layer, [2009-12-01] v0.258

[heliosdev]

Have you changed the code to use the v1 dll or are you using the latest version?
The version shows gles2.0 but the numbers are still too low! As gles2.0 is not backward compatible it implements the gles1.x using shaders (which isn't always very performant).

[Acctoj]

Hi guyz! Just started to play with my HD2, before own mostly Symbian devices. I installed your drivers, and when I run the test, it looks really weird. After installing Chainfire's driver, I've got

alpha texture - 20405
Torus -33222
color blending 20360
sub texture -19947

numbers decreased like five times from original results w/o drivers!!!

original ROM 1.43.411.2 gives me folowing:

alpha texture - 138405
Torus -123541
color blending 145300
sub texture -88305

but most weird thing that WITH driver, tests and games flying like lightning, showing low score, and without drivers I see hi score(not that much, comparing to others;)) but tests show some stutters. Please help me to understand why is that?!!!

[heliosdev]

That's the expected behavior like Chainfire explained in his post (http://forum.xda-developers.com/showpost.php?p=5422334&postcount=102)!

[streamg]

have installed chainfire patch 1.1

but i see 2 posts above that he hace much better TORUS results


Results
--------------------------------
alpha texture (128x256 rgba)
19590 kPixels/s

torus (3k triangles) one light
22044 kPixels/s

color blending
20079 kPixels/s

sub texture (128x128 15x15)
19600 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Version
OpenGL ES-CM 1.1

Renderer
AMD Z430

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

[hariprasad96]

Any source code available?

is there any c source code or java jsr 239 code is available for the same ?

[optical10]

After installing Throttle Launcher 1.1 yesterday on my HTC Advantage I was reminded that it doesn't have any of the above drivers for 3-D acceleration.

It's a fast phone at 624MHz CPU but because I've been using SPB Mobile Shell 3.5 I haven't noticed or needed 3-D acceleration because that program worked extremely well without it.

However as this phone is coming to the end of its life and it's out of warranty I don't mind tinkering to get OpenGL working on it.

I just can't find any sites or information specifically for HTC WM6 phones supporting an official or unofficial driver download, can anyone point me in the right direction for such a driver download?

Any help appreciated, sorry for hijacking this thread but it seemed the closest fit for my issue.

[heliosdev]

Just did a quick search in the HTC Advantage forum (http://forum.xda-developers.com/forumdisplay.php?f=351) (use 'search this forum' with the keyword opengl and you'll get three hits). You should try to find another device which has the 'same' chipset (HTC Diamond?) and try to use its drivers. (I'll take no responsibility for any damages or hard resets ;))

P.S.:You've posted the same question in the Test OpenGLES2 thread. I'm answering here as you'll get OpenGLES1.x support if any! And do not cross post, especially when the threads are from the same user as now you can delete the other post and point to this thread/answer...

[heliosdev]

is there any c source code or java jsr 239 code is available for the same ?

Well, there are examples/tutorials for opengles development just use your favorite search engine.
jsr239 opengles 1.0.1 bindings (http://java.sun.com/javame/reference/apis/jsr239/index.html)

[optical10]

Just did a quick search in the HTC Advantage forum (http://forum.xda-developers.com/forumdisplay.php?f=351) (use 'search this forum' with the keyword opengl and you'll get three hits). You should try to find another device which has the 'same' chipset (HTC Diamond?) and try to use its drivers. (I'll take no responsibility for any damages or hard resets ;))

P.S.:You've posted the same question in the Test OpenGLES2 thread. I'm answering here as you'll get OpenGLES1.x support if any! And do not cross post, especially when the threads are from the same user as now you can delete the other post and point to this thread/answer...

oops sorry about that, I got a little bit carried away/ desperate & posted in the two threads, not sure which one was still active. As you suggest I will have to delete the other and continue here.

"@optical10 The software renderer is your last resort. The Vincent3d is the only one I know and it's OpenGL ES 1.x. You can copy it to /windows (don't overwrite existing file). Don't expect too high performance."

Quoting from your answer in the other post I just wanted to clarify which file/Folder I should place in the root of Windows, just the "libgles_cm.dll " file or the whole folder "ogles-bin-1.0.0" including its numerous subfolders? Any help appreciated.

[heliosdev]

Just the libgles_cm.dll file to the windows folder.
Check results with software renderer (http://forum.xda-developers.com/showpost.php?p=3766699&postcount=13) to "compare" your results.

[Stefano_1986_]

Results on my Acer NeoTouch:
Results
--------------------------------
alpha texture (128x256 rgba)
126818 kPixels/s

torus (3k triangles) one light
93885 kPixels/s

color blending
119385 kPixels/s

sub texture (128x128 15x15)
130367 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Version
OpenGL ES-CM 1.1

Renderer
AMD Z430

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

[mishok_006]

NeoTouch p400
Results
--------------------------------
alpha texture (128x256 rgba)
8674 kPixels/s

torus (3k triangles) one light
7928 kPixels/s

color blending
8462 kPixels/s

sub texture (128x128 15x15)
7857 kPixels/s

OpengGL ES info
--------------------------------
Vendor
Advanced Micro Devices, Inc.

Version
OpenGL ES-CM 1.1

Renderer
AMD Z430

Extensions
GL_AMD_compressed_3DC_texture
GL_AMD_compressed_ATC_texture
GL_ARB_texture_env_combine
GL_ARB_texture_env_dot3
GL_ARB_texture_mirrored_repeat
GL_ARB_vertex_buffer_object
GL_ATI_compressed_texture_atitc
GL_ATI_texture_compression_atitc
GL_EXT_blend_equation_separate
GL_EXT_blend_func_separate
GL_EXT_blend_minmax
GL_EXT_blend_subtract
GL_EXT_stencil_wrap
GL_OES_blend_equation_separate
GL_OES_blend_func_separate
GL_OES_blend_subtract
GL_OES_compressed_ETC1_RGB8_texture
GL_OES_compressed_paletted_texture
GL_OES_draw_texture
GL_OES_extended_matrix_palette
GL_OES_framebuffer_object
GL_OES_matrix_palette
GL_OES_point_size_array
GL_OES_point_sprite
GL_OES_read_format
GL_OES_stencil_wrap
GL_OES_texture_cube_map
GL_OES_texture_env_crossbar
GL_OES_texture_mirrored_repeat

[edwardar06]

thanks a lot

[iRiKi]

Please post your frame rates and the model/ROM running this test.
More tests to come.


Hello heliosdev,

I've done some benchmarks using your app, you can check the results here --> http://forum.xda-developers.com/showthread.php?p=11736290