The OpenGL ES 2.0 thread

Posted on: 2019/3/28 17:15 #1

Just can't stay away

Edited by Capehill on 2019/3/28 17:49:06
Edited by Capehill on 2019/3/28 18:59:10
Edited by Capehill on 2019/4/1 19:14:35
Edited by Capehill on 2019/4/3 17:36:00
Edited by Capehill on 2019/4/7 16:32:46
Edited by Capehill on 2019/7/4 8:47:52
Edited by Capehill on 2019/7/9 8:52:41
Edited by Capehill on 2019/9/14 18:53:06
Edited by Capehill on 2019/12/29 12:00:58
Edited by Capehill on 2020/1/5 17:30:16
Edited by Capehill on 2020/5/13 8:07:15
Edited by Capehill on 2021/3/28 18:31:55
Edited by Capehill on 2021/4/21 14:45:36
Edited by Capehill on 2022/8/29 8:09:07
Edited by Capehill on 2022/10/2 19:22:18
Edited by Capehill on 2022/10/3 16:45:14
Edited by Capehill on 2022/10/3 16:56:24

Re: The OpenGL ES 2.0 thread

Posted on: 2019/3/29 12:14 #2

Not too shy to talk

OpenGL ES 2 version 2.3 for Warp3D Nova / AmigaOS4 is on my FTP for testers to test now!

- Support for GL_MIN and GL_MAX / GL_MIN_EXT and GL_MAX_EXT blend equations.

- Consequently added GL_EXT_blend_minmax to the extensions string.

- GL_QUAD_STRIP is now sort of supported, by simply switching to triangle strips internally.

- by default the lib's internal compatibility patcher kicks in if your GLSL lacks a #version tag. In rare cases this may fail and if it does it will now try again using the unpatched GLSL instead. This is usually more desirable than to globally disable the patcher on context creation.

- the glslangvalidator_redux tool has been updated to reflect that new behaviour.

- Critical fix: one of the 32bit-step-hashers had a problem with small data-packets; it would eventually produce the same hash for similar data, which would result in geometry becoming distorted or being falsely positioned, stuff like that. Actually the symptoms were pretty much like those of the bug fixed in the previous version 2.2.
Thanks again to Capehill for reporting and for providing a very nice tiny test program for it! Luckily this was also the reason for a certain sky-box-issue reported by kas1e and which I failed to pinpoint until now, thanks to him too!

- Fix: if the library was told to automatically handle the output window, eventual window-border overdraw was possible. Thanks again to Capehill for reporting and providing a test-prog and to Hans for additional information.

- !don't forget to download the new include-folder too!

- version set to 2.3 (30.3.2019)

Cheers,
Daniel

Edited by Daytona675x on 2019/3/30 9:29:57

[Facebook] [YouTube Channel] [ko-fi]

Re: The OpenGL ES 2.0 thread

Posted on: 2019/3/29 15:19 #3

Just can't stay away

@Daytona675x

Yay, thanks!

Fixed in 2.3:
- No more quad mysteries.
- Blitting over window borders on window minimize.

Fixed in 2.4:
- Diamonds and Dust title texts work again.

Edited by Capehill on 2019/3/29 19:20:04
Edited by Capehill on 2019/3/31 17:35:15
Edited by Capehill on 2019/4/1 19:11:41

Re: The OpenGL ES 2.0 thread

Posted on: 2019/3/30 9:32 #4

Not too shy to talk

@Capehill
Quote:

Blitting over window borders on window minimize. Mantis ticket opened

Fixed. Since 2.3 is still "hot" and Matthew didn't update it yet, I put the fix in there without increasing the version number (lib date has been updated though).

[Facebook] [YouTube Channel] [ko-fi]

Re: The OpenGL ES 2.0 thread

Posted on: 2019/3/30 10:33 #5

Home away from home

@Daniel
Quote:

Fixed.

Confirmed !

@Capehill
Now all SDL apps which use gl4es do not have anymore that overlaping of window's borders about which i told in the SDL2 thread. Foobillard, lf3 and all the games now reacts correctly once you resize it. So those reports for SDL2 about can be forgotten then.

Join us to improve dopus5!
AmigaOS4 on youtube

Re: The OpenGL ES 2.0 thread

Posted on: 2019/3/30 13:42 #6

Just can't stay away

@kas1e

Good to hear. Hope to test it soon...

EDIT: tested and works. Thanks!

Regarding OpenGL ES 2.0 test programs: they seem to use the original version aglCreateContextTags, maybe could be updated to use aglCreateContextTags2?

Edited by Capehill on 2019/3/31 17:37:13

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/1 12:01 #7

Not too shy to talk

@Capehill
Quote:

Diamonds and Dust title texts (the blue ones) are still broken.

Fixed as of version 2.4:

OpenGL ES 2 version 2.4 for Warp3D Nova / AmigaOS4 is on my FTP for testers to test now!
No April fool hoax, we got a fresh version for real again

- Fix: I forgot to enforce a Nova submission if a user-supplied VBO was modified and if this VBO was already marked as being queued for rendering. The effects could vary. Capehill reported it as missing letters in one of his games (in fact those letters weren't really missing, but had mutated to others which were now being drawn twice

), thanks again!

- Fix: glMapBufferOES now issues a Nova submission flush in such a case too, if the access-mode is not GL_READ_ONLY.

- performance: some more branch hints, one redundant Nova VBOSetArray call removed.

- version set to 2.4 (1.4.2019)

EDIT: 2.4 has been updated again, I had added a fresh bug but luckily Capehill immediately smelled it

Fixed.

Cheers,
Daniel

Edited by Daytona675x on 2019/4/1 21:28:50

[Facebook] [YouTube Channel] [ko-fi]

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/1 13:00 #8

Home away from home

@Daniel
Quote:

- Fix: I forgot to enforce a Nova submission if a user-supplied VBO was modified and if this VBO was already marked as being queued for rendering. The effects could vary. Capehill reported it as missing letters in one of his games (in fact those letters weren't really missing, but had mutated to others which were now being drawn twice ), thanks again!

That one confirmed for sure.

Also tested all games,etc : all works as expected.

@Capehill

Quote:

Regarding OpenGL ES 2.0 test programs: they seem to use the original version aglCreateContextTags, maybe could be updated to use aglCreateContextTags2?

Once Daniel made those "2" versions of functions, i start to use them when adding gl4es into SDL1/SDL2, so all my tests already with it.

Btw, you may also add to the list of apps which use ogles2 (via gl4es as well) these:

Huno's RCTW Reborn:

http://hunoppc.amiga-projects.net/con ... return-castle-wolfenstein

and his EGL wrapper:

http://hunoppc.amiga-projects.net/con ... ry-eglwrapper-opengles-20

And AmiBoing's "Spencer" also use ogles2.library directly for sure:

http://www.amiboing.de/gameDetail.php?id=25

Join us to improve dopus5!
AmigaOS4 on youtube

RIBDEVIL

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/1 13:29 #9

Not too shy to talk

@Daytona675x

Well done.

When we will have it ?

Virginio

AmigaOne X5000 OS4.1 FEU1 And Lubuntu 10.04

1200 towered with Blizzard PPC - BVision and Mediator
And a new fantastic Chameleon64

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/1 19:17 #10

Just can't stay away

@kas1e

Confirmed 2.4 fix and updated links in the first post too.

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/1 21:28 #11

Not too shy to talk

@RIBDEVIL
Thanks! AFAIK it won't take long for the next official update.

@Capehill
@all
2.4 has just been updated, I had added a fresh bug but luckily Capehill immediately smelled it 🙂 Fixed.

[Facebook] [YouTube Channel] [ko-fi]

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/2 6:29 #12

Just can't stay away

@Daytona675x

2.4 II works for me. Excellent.

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/7 11:33 #13

Not too shy to talk

OpenGL ES 2 version 2.5 for Warp3D Nova / AmigaOS4 is on my FTP for testers to test now!

- Fixed a very subtle OpenGL specs violation in glVertexAttrib which could result in crashs in certain use-cases because it would make the library to try to copy from a nullptr. This fixes the crash in the game Fricking Shark reported by kas1e.
Fun fact, no kidding: back in 2016 I placed this comment behind the critical line of code which had to be removed now:

// FIXME: is that correct?! I don't think so

EDIT, little addon:
- Further speedup of the lib's internal ubyte8->float32 copy-conversion. kas1e reported 2+ fps in his Q3 tests (84.4 vs. 86.6). Not too shabby.

- version set to 2.5 (7.4.2019)

Cheers,
Daniel

Edited by Daytona675x on 2019/4/7 13:05:05

[Facebook] [YouTube Channel] [ko-fi]

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/7 11:42 #14

Home away from home

@Daniel
Confirmed ! Thanks a lot !:) At least and because of that bug and of some others, we have glSnoop from Capehill which can trace/debug ogles2.library as well as warp3dnova.

@Capehill
You probabaly can add glSnoop to the list of ogles2 apps too :)

Join us to improve dopus5!
AmigaOS4 on youtube

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/7 12:44 #15

Not too shy to talk

@kas1e
Please check out the FTP. I made an alternative lib version which probably speeds up the lib's internal ubyte8-VA-workaround a bit more. A bit. Don't expect too much, but maybe it's good enough for a few fps here and there.

[Facebook] [YouTube Channel] [ko-fi]

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/7 12:55 #16

Home away from home

@Daniel
Yeah, it is ! Speed up in compare with pure 2.5 on ~2.5 fps, and in compare with mgl on ~3fps.

There is q3 results 1600x1200:


maybe faster: 1st run: 86.5,  2st run: 86.7 , 3st run: 86.6

pure 2.5:     1st run: 83.9,  2st run: 84.4 , 3st run: 84.4

mgl:          1st run: 83.3,  2st run: 83.6 , 3st run: 83.6

Cool :)

Join us to improve dopus5!
AmigaOS4 on youtube

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/9 18:00 #17

Home away from home

@Daniel
May i suggest if you doenst mind with every ogles2 release, include not only ogles2.library and libogles2.a , but also ogles2.library.debug , which will be the same ogles2.library just contain debug symbols.

Currently Hans with warp3dnova also do it like this (and Mattew add it to Enhancer as well), will be handy to have the same with ogles2 as default too. For you it will be imho better if any of us will create bug reports with proper stack traces, etc

Join us to improve dopus5!
AmigaOS4 on youtube

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/10 7:19 #18

Not too shy to talk

@kas1e
Yes, will do. Don't get hyped and don't overestimate the value of this though:

for example 99% of all invalid vertex-pointers will end up crashing (if crashing) at the same location, namely inside one of the hashers when drawing, because that's where they are touched first, although the root of evil is maybe a pointer-assignment hundreds of frames ago or a forgotten VA disable or whatever. Same thing with most other stuff, most is "delayed".

Therefore pretty much the only real good thing is and remains to provide me with small isolated test programs, like those Capehill came up with recently. Without those the crash-site info alone (even incl. function params or whatever) would have left me in the dark.

Edited by Daytona675x on 2019/4/10 7:44:32

[Facebook] [YouTube Channel] [ko-fi]

Caras

Re: The OpenGL ES 2.0 thread

Posted on: 2019/4/11 6:28 #19

Just popping in

I know it's not part of the ES 2.0 API but is there any chance of adding support for GL_TEXTURE_BASE_LEVEL and GL_TEXTURE_MAX_LEVEL? I could then add mipmapping to the texture atlas in AmiCraft.