GL4ES: another OpenGL over OpenGLES2 emulation - some tech. info and porting progress

@Hans
Trying to port latest cube2 (SDL2 for now), and seems cube2 now use shaders in any case (before it was possible to switch usage of shaders off), so we need firstly to fix some issues in warp3dnova shader's compiler before i can go futher with it

There are 4 of them which is necessary to deal with:

Unsupported opCode OpImageSampleProjImplicitLod

Warp3DNova didn't accept empty fragment shaders

Vertex shader with arrays support issue 1

Vertex shader with arrays support issue 2

@All
Also build SDL2 version (Capehil's one) of quake3 via gl4es, but it has some strange issue at first run which disappear once i start actual game, see screenshot:

(press to open in new tab for fullsize):


Very well can be because of the way i added hackingly gl4es support to SDL2 (like forgot something which wasn't need with SDL1?). Once i start levels itself, that trashing gone.

@kas1e

Please test https://github.com/capehill/ioq3/commi ... 8ee74db6bfc9ad0934054a64f

@Capehill
Yes, recompiled sdl2 version with that change over gl4es and vertex arrays works, thanks !

@all

Do OpenGL and OpenGLES2 exclude themselves?
If one is used the other can't be used?

Or is OpenGLES just an addition to OpenGL and without OpenGL it wouldn't work for itself?

Trying to understand some basics...

@Raziel

On OS4 there should be no exclusion in a sense that you should be able to run both OGLES2 and MiniGL apps at the same time, if you have the necessary drivers. Programmer makes a decision which one (MiniGL or OGLES2) is opened and used.

There are different versions of OpenGL ES: https://en.wikipedia.org/wiki/OpenGL_ES

If you compare OpenGL ES and OpenGL, ES version has less functionality as some legacy features have been removed.

@Raziel
Quote:

Its 2 different realisations (different libraries, different fucntions). OpenGLES2 is limited set of OpenGL2. I.e. there is full OpenGL2 and OpenGLES2, its reduced versions of OpenGL2 with less functions, no fixed pipiline, some more legacy removed (for easy work on mobile devices, etc)

On os4 we have 1 realisation of full opengl : minigl. Which is also limited, but not because of standard as opengles2, but just some fucntions not implemnted, and minigl is just opengl 1.5.


Quote:

As 2 different programms at the same time: yes.
As one programm which use both: no (as context different).

You can open minigl.library, create context of minigl (opengl) and use opengl functions. Or, you can open ogles2.library, and use its context (which by logic alos opengl context, but different anyway, as different library , etc). Some functions of opengl are the same as in ogles2 (of course, as ogles2 its opengl, just with less fucntions).


Quote:

Its 2 differetn realisation of opengl standard. Full opengl are full, and opengles are opengl with less fucntions.

In terms of SDL, you choice from SDL app what one to use.

In the net, you can find sources which use opengl, or opengles, and they will be different. Of course, it possible to write opengl code , which will works "as it" over opengles , but most time it will not, as full opengl will use one or another function which not present in ogles2.


Because of that, we worry about that gl4es : to have full opengl. I.e. again making jumbo mambo, to have full opengl via something else (in our case it is gl4es->ogles2.library->warp3dnova.library).

Of course , if we have in first place just some mesa opengl driver for full opengl, there wasn't needs not for ogles2.library, not for gl4es. But as writing full opengl driver is much harder than ogles2 driver, we have what we have.

@kas1e

Thank you for the explanation.

The opengles2 implementation is "complete", isn't it?
Nothing left out from the "official" opengles2?

Because then it's a matter of supporting it correctly or dig a little deeper as to why it kind of work, but not completely (e.g. wrong colors on moveable objects etc.).

Maybe there are functions (ab)used that we don't have?

EDIT:
I have a project where it tries to compile with "-lGLESv2" instead of "-lGLES2"...so what is the correct way?
Nevermind...seems i need to use "-logles2" to make it link correctly

@kas1e

Do the opengles2 examples from the Enhancer SDK directory crash on exit for you too?

They run fine, but whenever i close the window, it's nirvana for me, that bad, that i have to unplug the X1000 for 10 seconds to make it come back to life.

@kas1e

Problem:



Help?

Thank you

@Raziel
Simply change
Bool flipy;
To
int flipy
In the shaders and in C sources (and .h headers if necessary)

@thellier

Hi,

As it expects any boolean result, maybe you need to define flipy to 0 or 1.

Regards,

@Raziel
Quote:

As far as i know : yes, it is fully complete, and even some functionality added from opengles3 (a little, but still).

In the Warp3dnova itself missing at moment (taken from Daniel's ogles2.library readme): sample coverage, dithering, compressed textures and render-to-texture, which mean it also missing in ogles2 because of that. I.e all functions of opengles2 standard are in, just when this kind of functionality used, it probabaly return "true" without doing anything. But for example in gl4es (which is real big thing), none of that feature used, so my bet they at moment missing as really unimportant (except probabaly those 2 about textures).

Hans and Daniel know things better about of course, but i think i not mistaken anything there.

Quote:

I can tell from all my tests with gl4es and bug reports to ogles2 and warp3dnova : that at moment it's all complete enough and all works as should. Some bugs still left in warp3dnova, but Hans works on them.


Quote:

In terms of realisation of opengles2 our Daniel's ogles2.library is "full", with even some bits from opengles3 standard.

Quote:

Yeah, its just matter of naming. I doing tests on all oses where i have opengles2 support, and -lGLESv2 happens to be on win32 for me as well. On os4 we just use -logles2, yep. But name of fucntions, headers, etc, all of course about the same.

Quote:

Can you upload that shader somewhere ?

Usually, when i want to know, if shader works at all, or if it works on our warp3dnova, i do that:

1. I use glslangvalidator (latest version from Daniel), placed there:

http://www.amiga.org/developer/bugrep ... .php?file_id=230&type=bug

So, via that glslangvalidator i check, if shader is working at all (glslanvvalidator check shader for be "normal one" at all, does not matter on what OS and on which drivers). If it compiles fine, then it mean that shader by itself ok.

Command:

ram:> glslangvalidator -G -o test.vert.spv test.vert

2. Then i use Warp3DNovaShaderInfo utility which come with warp3dnova sdk, and take the converted shader from glslangvalidator as input for warp3dnovashaderinfo.

Then, if there is no error : then shader also works over warp3dnova. But if there is error, then its bug in the warp3dnova, which need to be reported, and Hans will fix it.

@kas1e

I wouldn't know where to get that shader, should it be part of the source?
Can i extract it somehow?

Sorry, noob here, help needed

EDIT:

Ok, i see, this seems to be part of the source

No promises that i got the correct one here

@Raziel
I also do not know shader's language, but i tested your one: and glslangvalidator errors on it with errors 'in for stage inputs': no supported for this version or enabled extensions.

I do not know what it all mean, but i can only say from test that its issue in the shader itself, and not related to ogles2.library or warp3dnova.

glslangvalidator its kind of "de facto" tool from the those ones who made standard, and if it didn't works over it, then something wrong in shader.

@kas1e

That would be starnge as that shader works on dozens of other platforms, just not on AmigaOS4...i highly doubt that it's an issue with the shader.

The folks over at the app say that it needs boolean uniforms, which seem to be not supported in our version.

Maybe Hans or Daytona could comment on that issue?

@Raziel
Then why it didn't works over glslangvalidator ? If only it is something very new, and even new glslangvalidator from Daniel are too old for :) But yep, there need Daniel or Hans.

@kas1e

Well, first of all, i don't even know if that is the correct shader (or if it even *IS* a shader), i just found it under the same name in the source, but that doesn't really mean anything

Maybe it's just some kind of handshake code for the real shader?

Anyway, thank you for time and help so far

@Raziel
It is for sure shader, and are vertex one (see at end gl_position word). Also structured like usual shader: init stuff at top, then main with gl_position. It just seems something at top can be missing, or our new glslangvalidator too old (code of which also in ogles2.library)

@kas1e

Hmm, or maybe the shader is "optimized" for more modern versions of opengles2, leaving out stuff that is now missing, just like you assumed?

Too bad, that glslangvalidator doesn't tell what is wrong.

Could you maybe post the output here, so i can pass it on to the app devs?