Next Week We Should Hear All About OpenGL 5.0

[flim]

OpenGL 4.5 spec released: (http://www.opengl.org/registry/) http://www.opengl.org/registry/doc/glspec45.core.pdf

new features:

• GL_ARB_clip_control (https://www.opengl.org/registry/spec...ip_control.txt)
• GL_ARB_cull_distance (https://www.opengl.org/registry/spec...l_distance.txt)
• GL_ARB_ES3_1_compatibility (https://www.opengl.org/registry/spec...patibility.txt)
• GL_ARB_conditional_render_inverted (https://www.opengl.org/registry/spec...r_inverted.txt)
• GL_KHR_context_flush_control (https://www.opengl.org/registry/spec...sh_control.txt)
• GL_ARB_derivative_control (OpenGL Shading Language Only) (https://www.opengl.org/registry/spec...ve_control.txt)
• GL_ARB_direct_state_access (https://www.opengl.org/registry/spec...ate_access.txt)
• GL_ARB_get_texture_sub_image (https://www.opengl.org/registry/spec..._sub_image.txt)
• GL_KHR_robustness (https://www.opengl.org/registry/spec...robustness.txt)
• GL_ARB_shader_texture_image_samples (OpenGL Shading Language Only) (https://www.opengl.org/registry/spec...ge_samples.txt)
• GL_ARB_texture_barrier (https://www.opengl.org/registry/spec...re_barrier.txt)

[przemoli]

OS X 10.9 is OpenGL 4.1.

iOS OpenGL ES will be 3.x, but more importantly, Metal API.

Thats separate issue.

And I agree. Biggest enemy of OpenGL is OpenGL on OSX, and for OGL ES its Metal. On linux. On Windows Mantle is NOT accessible to wide audience. DX12 is not. Metla do not support it. OpenGL "AZDO" is supported by all 3 vendors right now. Thats where I see biggest task for Khronos.

[Marc Driftmeyer]

5 or 4.5 makes no difference.

As things stand now, we lack anything worthwhile on the horizon in terms of new hardware.

AMD say that first iteration of Mantle will be about current hw.
MS says that DX12 will be just happy with DX FL_11.x.
Nvidia do not have any hidden hw either.
Intel is perfectly happy to keep to their GEN architecture too
Apple even served LESS then OpenGL ES 3.1



The who will propose anything meaningful for next iteration?
Nobody.

Small things? Plenty of them. Just read AMD/Nvidia extensions. Bits and pieces here and there added to this and that existing functionality.


So in the end, number wont matter.

What is important is how far Khronos will push "AZDO", and how well they PR it.
(Anyone see serious comparison of OpenGL "AZDO" in the article about DX12 and Mantle on Windows gaming site?)

OS X 10.9 is OpenGL 4.1.

iOS OpenGL ES will be 3.x, but more importantly, Metal API.

[Kivada]

You're seriously wrong here. A lot of 4.X features benefit low-end GPUs a lot more, than high-end one. Most of current-gen games unable to utilize power of high-end GPU so people who buy them don't actually need performance tweaks or AZDO because for most of them 50FPS vs 70FPS doesn't make sense.

Obviously I not talking about eye-candy things like tessellation, but about things like compute shaders, texture/buffer storage and other extensions that make possible to do things in more optimal way.

Note the last part you quoted from me. Intel is still only producing pitiful GPUs at insane prices for the level of performance.

[_SXX_]

Even now though with the Iris Pro 5200 they aren't fast enough to actually USE the OpenGL4.* feature set for anything even remotely demanding.

You're seriously wrong here. A lot of 4.X features benefit low-end GPUs a lot more, than high-end one. Most of current-gen games unable to utilize power of high-end GPU so people who buy them don't actually need performance tweaks or AZDO because for most of them 50FPS vs 70FPS doesn't make sense.

Obviously I not talking about eye-candy things like tessellation, but about things like compute shaders, texture/buffer storage and other extensions that make possible to do things in more optimal way.

[przemoli]

5 or 4.5 makes no difference.

As things stand now, we lack anything worthwhile on the horizon in terms of new hardware.

AMD say that first iteration of Mantle will be about current hw.
MS says that DX12 will be just happy with DX FL_11.x.
Nvidia do not have any hidden hw either.
Intel is perfectly happy to keep to their GEN architecture too
Apple even served LESS then OpenGL ES 3.1



The who will propose anything meaningful for next iteration?
Nobody.

Small things? Plenty of them. Just read AMD/Nvidia extensions. Bits and pieces here and there added to this and that existing functionality.


So in the end, number wont matter.

What is important is how far Khronos will push "AZDO", and how well they PR it.
(Anyone see serious comparison of OpenGL "AZDO" in the article about DX12 and Mantle on Windows gaming site?)

[shmerl]

I hope OpenGL 5 would redesign and improve multithreading approach. DX and Mantle are seriously ahead in this aspect.

[smitty3268]

That's exactly the same document I read.
4 years and 1/2 since OpenGL 4.x and still working on it, this pushes 5.0 around 2020.

We're seriously within 12 months of full GL 4.4 support in Mesa, now. GL5 won't take until 2020.

[Kivada]

Intel is very close to OpenGL 4.x support in Mesa 11 according to this http://www.phoronix.com/scan.php?pag...tem&px=MTc0NzY

Up until this current generation Intel hardware was useless for everything but grandma's email, go back and look how many of the GPU features where offloaded to the CPU even just a few years ago while Nvidia and AMD IGPs put a full GPU on the mobo.

Even now though with the Iris Pro 5200 they aren't fast enough to actually USE the OpenGL4.* feature set for anything even remotely demanding. For that you need high end GPUs, with the advent of OSS drivers for high end GPUs came a real reason to develop those features in Mesa.

I say this with a 4Mb Intel i740 2xAGP card GPU sitting on the shelf.

[losko]

OpenGL support advances more quickly than you think.

That's exactly the same document I read.
4 years and 1/2 since OpenGL 4.x and still working on it, this pushes 5.0 around 2020.

Remember, Intel hardware just barely supported this stuff, but was way too slow to actually make use of it and still is generally is too slow to make use of it. When the new features would just bring the hardware to an unplayable crawl why even implement it till theres hardware capable of making use of it?

Intel is very close to OpenGL 4.x support in Mesa 11 according to this http://www.phoronix.com/scan.php?pag...tem&px=MTc0NzY