gl2shim for core/gles2+ contexts #1446

mittorn · 2023-10-13T12:17:25Z

Limited FFP emulation with streaming buffers for gles/core contexts
Please check on nvidia/amd/intel proprietary drivers before merging, maybe need change some feature detection/blacklisting logic

…gles context support (SLOW, still needs ffp for matrix operations)

…, wrap gles functions in gl2wrap when XASH_GLES enabled

…g dumb array object

…d functions

…on on core context

…in GLES contexts

…o/world array render work

…s generates GL Errors, but renders fine on mesa

…ngs, Draw QUADS with TRIFAN when possible

…fer works now on core contexts

…port

…ent quering extensions on core profile

a1batross · 2023-10-14T02:46:08Z

engine/platform/sdl/vid_sdl.c

@@ -812,8 +812,13 @@ qboolean VID_CreateWindow( int width, int height, window_mode_t window_mode )
 	{
 		if( !glw_state.initialized )
 		{
-			if( !GL_CreateContext( ))
-				return false;
+			while( !GL_CreateContext( ))


I don't get it. There is already function VID_CreateWindowWithSafeGL that iterates over safegl options.

Why there is a second loop?

a1batross · 2023-10-14T04:41:11Z

ref/gl/gl_image.c

@@ -1070,7 +1070,7 @@ static void GL_TextureImageRAW( gl_texture_t *tex, GLint side, GLint level, GLin
 	}
 	else if( tex->target == GL_TEXTURE_2D_MULTISAMPLE )
 	{
-#if !defined( XASH_GLES ) && !defined( XASH_GL4ES )
+#if !defined( XASH_GL_STATIC ) || (!defined( XASH_GLES ) && !defined( XASH_GL4ES ))


What's the point for checking for XASH_GL_STATIC here? It doesn't seem to change anything for MSAA.

If you're updating macros, don't forget to update comments after #else and #endif

a1batross · 2023-10-14T04:41:56Z

ref/gl/gl_local.h

@@ -40,6 +40,9 @@ int VGL_ShimInit( void );
 void VGL_ShimShutdown( void );
 void VGL_ShimEndFrame( void );
 #endif
+#if !defined(XASH_GL_STATIC)
+#include "gl2_shim/gl2_shim.h"


I don't think #if !defined(XASH_GL_STATIC) really needed here. Including gl2shim header won't break anything.

a1batross · 2023-10-14T04:44:51Z

ref/gl/gl_opengl.c

+
+static dllfunc_t shaderobjectsfuncs_gles[] =
+{
+	{ "glDeleteShader"             , (void **)&pglDeleteObjectARB },


Let's use GL_CALL macro here, it's very easy to pass wrong pointer here. You can create a second macro that will include ARB postfix.

a1batross · 2023-10-14T04:45:07Z

ref/gl/gl_opengl.c

+
+static dllfunc_t vaofuncs[] =
+{
+{ "glBindVertexArray"    , (void **)&pglBindVertexArray },


Use GL_CALL here.

a1batross · 2023-10-14T04:45:58Z

ref/gl/gl_opengl.c

-			// one or more functions are invalid, extension will be disabled
-			GL_SetExtension( r_ext, false );
+			// HACK: fix ARB names
+			char *str = Q_strstr(func->name, "ARB");


Add spaces in parentheses.

a1batross · 2023-10-14T04:46:46Z

ref/gl/gl_opengl.c

-			GL_SetExtension( extid, true ); // required to be supported by wrapper
+			if(!GL_CheckExtension( "multitexture", multitexturefuncs, "gl_arb_multitexture", GL_ARB_MULTITEXTURE, 1.0) && glConfig.wrapper == GLES_WRAPPER_NONE  )
+#ifndef XASH_GL_STATIC
+				if( !GL_CheckExtension( "multitexture_es1", multitexturefuncs_es, "gl_arb_multitexture", GL_ARB_MULTITEXTURE, 1.0 ) )


Can you just use && here instead of nested ifs?

a1batross · 2023-10-14T04:48:57Z

ref/gl/gl_opengl.c

+			break;
+		case GL_ARB_VERTEX_ARRAY_OBJECT_EXT:
+			if(!GL_CheckExtension( "GL_OES_vertex_array_object", vaofuncs, "gl_vertex_array_object", extid, 3.0 ))
+				!GL_CheckExtension( "GL_EXT_vertex_array_object", vaofuncs, "gl_vertex_array_object", extid, 3.0 );


Unused NOT operator here.

a1batross · 2023-10-14T04:50:43Z

ref/gl/gl_opengl.c


 	// get our various GL strings
 	glConfig.vendor_string = (const char *)pglGetString( GL_VENDOR );
 	glConfig.renderer_string = (const char *)pglGetString( GL_RENDERER );
 	glConfig.version_string = (const char *)pglGetString( GL_VERSION );
 	glConfig.extensions_string = (const char *)pglGetString( GL_EXTENSIONS );
+
+	pglGetIntegerv(GL_MAJOR_VERSION, &major);


Spaces in parentheses please.

a1batross · 2023-10-14T04:51:21Z

ref/gl/gl_opengl.c

+
+	pglGetIntegerv(GL_MAJOR_VERSION, &major);
+	pglGetIntegerv(GL_MINOR_VERSION, &minor);
+	if( !major && glConfig.version_string )


Can this even happen? O_o?

Yes. GL_MAJOR_VERSION was intoduced in gl 3.0

a1batross · 2023-10-14T05:12:58Z

ref/gl/gl2_shim/gl2_shim.h

+#pragma once
+
+// max verts in a single frame
+#define GL2_MAX_VERTS 32768


It's pretty low for our frame.

net_graph for example, easily exceeds this

…ernary ops, remove singleline multiple assignments, use bitset macros

mittorn · 2023-10-14T23:35:34Z

Because it is not called when GL_CreateContext failed. It is only called when it failed create window, but does not work in case window created, but context not. It just runs engine without valid context and it crashes on bad gl function calls. 14 октября 2023 г. 5:46:20 GMT+03:00, Alibek Omarov ***@***.***> пишет:

…

@a1batross commented on this pull request. > @@ -812,8 +812,13 @@ qboolean VID_CreateWindow( int width, int height, window_mode_t window_mode ) { if( !glw_state.initialized ) { - if( !GL_CreateContext( )) - return false; + while( !GL_CreateContext( )) I don't get it. There is already function `VID_CreateWindowWithSafeGL` that iterates over safegl options. Why there is a second loop? -- Reply to this email directly or view it on GitHub: #1446 (review) You are receiving this because you authored the thread. Message ID: ***@***.***>

mittorn · 2023-10-14T23:41:12Z

I do not want to have declaration of missing functtion present. It wont be present if gl2shim part not compiled at all (it have macro in source file), so why include this header? 14 октября 2023 г. 7:42:07 GMT+03:00, Alibek Omarov ***@***.***> пишет:

…

@a1batross commented on this pull request. > @@ -40,6 +40,9 @@ int VGL_ShimInit( void ); void VGL_ShimShutdown( void ); void VGL_ShimEndFrame( void ); #endif +#if !defined(XASH_GL_STATIC) +#include "gl2_shim/gl2_shim.h" I don't think #if !defined(XASH_GL_STATIC) really needed here. Including gl2shim header won't break anything. -- Reply to this email directly or view it on GitHub: #1446 (review) You are receiving this because you authored the thread. Message ID: ***@***.***>

mittorn · 2023-10-14T23:42:59Z

gl changed function names, so we need macro including two names in this case, ARB does not call all cases 14 октября 2023 г. 7:45:03 GMT+03:00, Alibek Omarov ***@***.***> пишет:

…

@a1batross commented on this pull request. > + { GL_CALL( glDisableVertexAttribArrayARB ) }, + { GL_CALL( glBindAttribLocationARB ) }, + { GL_CALL( glGetActiveAttribARB ) }, + { GL_CALL( glGetAttribLocationARB ) }, + { GL_CALL( glVertexAttrib2fARB ) }, + { GL_CALL( glVertexAttrib2fvARB ) }, +// { GL_CALL( glVertexAttrib3fv ) }, +// { GL_CALL( glVertexAttrib4f ) }, +// { GL_CALL( glVertexAttrib4fv ) }, +// { GL_CALL( glVertexAttrib4ubv ) }, +{ NULL, NULL } +}; + +static dllfunc_t shaderobjectsfuncs_gles[] = +{ + { "glDeleteShader" , (void **)&pglDeleteObjectARB }, Let's use GL_CALL macro here, it's very easy to pass wrong pointer here. You can create a second macro that will include ARB postfix. -- Reply to this email directly or view it on GitHub: #1446 (review) You are receiving this because you authored the thread. Message ID: ***@***.***>

mittorn · 2023-10-14T23:44:57Z

&& inside macro? I not sure if it would be better 14 октября 2023 г. 7:46:57 GMT+03:00, Alibek Omarov ***@***.***> пишет:

…

@a1batross commented on this pull request. > break; case GL_ARB_MULTITEXTURE: - GL_SetExtension( extid, true ); // required to be supported by wrapper + if(!GL_CheckExtension( "multitexture", multitexturefuncs, "gl_arb_multitexture", GL_ARB_MULTITEXTURE, 1.0) && glConfig.wrapper == GLES_WRAPPER_NONE ) +#ifndef XASH_GL_STATIC + if( !GL_CheckExtension( "multitexture_es1", multitexturefuncs_es, "gl_arb_multitexture", GL_ARB_MULTITEXTURE, 1.0 ) ) Can you just use && here instead of nested ifs? -- Reply to this email directly or view it on GitHub: #1446 (review) You are receiving this because you authored the thread. Message ID: ***@***.***>

mittorn · 2023-10-14T23:47:47Z

It seems to be some bug inside net_graph, it should not draw to many quads, but i'll add workaround for big GL_QUADS cases. 14 октября 2023 г. 8:13:09 GMT+03:00, Alibek Omarov ***@***.***> пишет:

…

@a1batross commented on this pull request. > + +This program is free software: you can redistribute it and/or modify +it under the terms of the GNU General Public License as published by +the Free Software Foundation, either version 3 of the License, or +(at your option) any later version. + +This program is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. +*/ + +#pragma once + +// max verts in a single frame +#define GL2_MAX_VERTS 32768 It's pretty low for our frame. net_graph for example, easily exceeds this -- Reply to this email directly or view it on GitHub: #1446 (review) You are receiving this because you authored the thread. Message ID: ***@***.***>

… overflow

a1batross · 2023-10-15T00:41:30Z

Because it is not called when GL_CreateContext failed. It is only called when it failed create window, but does not work in case window created, but context not. It just runs engine without valid context and it crashes on bad gl function calls.

It's better to be moved out of the scope of this patchset then.

I do not want to have declaration of missing functtion present. It wont be present if gl2shim part not compiled at all (it have macro in source file), so why include this header?

It will be detected at the link time. I don't see a problem here.

gl changed function names

Yeah, we discussed this already. It's very unfortunate, so let's keep it as is then.

&& inside macro? I not sure if it would be better

I think this code is not an example of a good code. It's not very obvious at which level break statement goes. The absence of code blocks even make it so much worse, that you even made a mistake in the checks. Try to take both macro paths, the one with defined XASH_GL_STATIC and the one where it's not defined.

mittorn · 2023-10-15T01:35:15Z

This might be availiable in gles context without static gl. I just rechecked all XASH_GLES leftovers because many of them refered to XASH_GL_STATIC really, not to GLES context itself. But GLES without wrappers was not supported before. This commit first time add launching on pure gles not completely wrapped by nanogl/wes/gl4es, only having ffp implemented with gl2shim 14 октября 2023 г. 7:41:22 GMT+03:00, Alibek Omarov ***@***.***> пишет:

…

@a1batross commented on this pull request. > @@ -1070,7 +1070,7 @@ static void GL_TextureImageRAW( gl_texture_t *tex, GLint side, GLint level, GLin } else if( tex->target == GL_TEXTURE_2D_MULTISAMPLE ) { -#if !defined( XASH_GLES ) && !defined( XASH_GL4ES ) +#if !defined( XASH_GL_STATIC ) || (!defined( XASH_GLES ) && !defined( XASH_GL4ES )) What's the point for checking for XASH_GL_STATIC here? It doesn't seem to change anything for MSAA. -- Reply to this email directly or view it on GitHub: #1446 (review) You are receiving this because you authored the thread. Message ID: ***@***.***>

mittorn · 2023-10-15T01:49:20Z

Alibek Omarov писал(а) 2023-10-15 02:41:

> Because it is not called when GL_CreateContext failed. It is only > called when it failed create window, but does not work in case window > created, but context not. It just runs engine without valid context > and it crashes on bad gl function calls. It's better to be moved out of the scope of this patchset then. Message ID: ***@***.***>

I need to test if i did not break older GL versions and found that i cannot even create context before 3.x with compatibility flag enabled and safegl was not set automaticly. I searched other places where it might be set, but it was unrelated: this function only calls SDL_CreateWindow. Even when this safegl level enough to create window, it might not enough to create context, so we need to increment it later --=_f98e57ced455552daa412d819c95ca55 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=UTF-8 <html><head><meta http-equiv=3D"Content-Type" content=3D"text/html; charset= =3DUTF-8" /></head><body style=3D'font-size: 10pt; font-family: Verdana,Gen= eva,sans-serif'> Alibek Omarov =D0=BF=D0=B8=D1=81=D0=B0=D0=BB(=D0=B0) = 2023-10-15 02:41: <blockquote type=3D"cite" style=3D"padding: 0 0.4em; border-left: #1010ff 2= px solid; margin: 0"> <div id=3D"replybody1"> <blockquote type=3D"cite" style=3D"padding: 0 0.4em; border-left: #1010ff 2= px solid; margin: 0"> Because it is not called when GL_CreateContext failed. It i= s only called when it failed create window, but does not work in case windo= w created, but context not. It just runs engine without valid context and i= t crashes on bad gl function calls. </blockquote> It's better to be moved out of the scope of this patchset t= hen. <p style=3D"font-size: small; -webkit-text-size-adjust: none; color: #666;"=

<span style=3D"color: transparent; font-size: 0; display: none; visibility=

: hidden; overflow: hidden; opacity: 0; width: 0px; height: 0px; max-width:= 0; max-height: 0; mso-hide: all;">Message ID: <FWGS/xash3d-fwgs/p= ull/1446/c1763221270@github.<= span>com> </div> </blockquote> I need to test if i did not break older GL versions and found that i can= not even create context before 3.x with compatibility flag enabled and safe= gl was not set automaticly. I searched other places where it might be set, = but it was unrelated: this function only calls SDL_CreateWindow. Even when = this safegl level enough to create window, it might not enough to create co= ntext, so we need to increment it later </body></html>

…

--=_f98e57ced455552daa412d819c95ca55--

--=_4bdf773d7ffda6467a4f2456ec038f46 Content-Transfer-Encoding: base64 Content-Type: image/gif; name=blocked.gif Content-Disposition: attachment; filename=blocked.gif; size=118 R0lGODlhZAAyAIAAAPrOzgAAACH5BAAAAAAALAAAAABkADIAAAJNhI+py+0Po5y02ouz3rz7D4bi SJbmiabqyrbuC8fyTNf2jef6zvf+DwwKh8Si8YhMKpfMpvMJjUqn1Kr1is1qt9yu9wsOi8fksvls KwAAOw== --=_4bdf773d7ffda6467a4f2456ec038f46--

a1batross · 2023-10-16T03:48:36Z

GitHub didn't close it automatically AGAIN.

Closing manually, merged.

mittorn added 30 commits October 13, 2023 15:05

ref_gl: introduce gl2_shim based on vgl_shim for future limited core/…

16e5f37

…gles context support (SLOW, still needs ffp for matrix operations)

gl2shim: limited matrix support

d779e8b

ref_gl: allow enabling XASH_GLES without wrapper, add ref_gles3compat…

bb3c038

…, wrap gles functions in gl2wrap when XASH_GLES enabled

ref_gl: fix creating core context, support core in gl2shim by creatin…

a609ed1

…g dumb array object

Workaround core context issues

e0ce0e6

gl2shim: Cleanup warnings, use APIENTRY declarations, stub unsupporte…

271e025

…d functions

gl2shim: support more GLSL versions

2a47120

ref_gl: fix build with XASH_GL_STATIC, avoid using ARB shader extensi…

d71ed26

…on on core context

ref_gl: remove complete useless macro, enable VBO under option

0687a52

ref_gl: fix late studio cvars registration, enable DrawRangeElements …

159c241

…in GLES contexts

gl2shim: implement client arrays (not in glcore/vao mode), make studi…

91248ab

…o/world array render work

gl2shim: support low (<130) shaders

c11ef08

ref_gl: fix enabling multitexturing on GLES

eaf4004

gl2shim: bind dummy VAO on DrawElements implementation on glcore, thi…

4f7880e

…s generates GL Errors, but renders fine on mesa

gl2shim: simple matrix invalidation, do not calculate MVP every drawcall

012a563

gl2shim: batcher experiments (UGLY, DO NOT USE), fix bad shader bindi…

7376d93

…ngs, Draw QUADS with TRIFAN when possible

gl2shim: implement fallback vbo, glDrawRangeElements with non-vbo buf…

59b9cb7

…fer works now on core contexts

gl2shim: persistent incremental streaming buffers, buffer cycling sup…

67545cc

…port

gl2shim: support incremental non-persistent buffers for arrays

de4dbb1

gl2shim: improve shader version handling, fix particle issues

ba27e43

gl2shim: limit begin-end chain length, cycle incremental buffers

3fa226d

gl2shim: use IBO for quads

7a1fa45

gl2shim: refactoring

db374cb

gl2shim: make shaders code more portable

9f4b832

ref_gl: try fix psvita

c6418c3

gl2shim: more refactoring

9bf7168

gl2shim: fix wrong color vertex

5411933

ref_gl: implement minimum inclusion version for GL extensions, implem…

fc34fce

…ent quering extensions on core profile

ref_gl: add missing non-arb shader functions

a4d307d

ref_gl: integrate gl2shim extensions

4f6cc92

mittorn added 2 commits October 13, 2023 22:40

ref_gl: do not trust REFAPI context version, get it from OpenGL anyway

911a7ee

ref_gl: fix XASH_GL_STATIC build

20a4e23

a1batross reviewed Oct 14, 2023

View reviewed changes

ref_gl: gl2shim: refactoring, add missing spaces in parentheses and t…

8c46357

…ernary ops, remove singleline multiple assignments, use bitset macros

a1batross force-pushed the gl2shim-rebase branch from e0d66c1 to 8c46357 Compare October 14, 2023 14:38

gl2shim: Allow drawing huge QUADS sequences by splitting drawcalls on…

6f7e1f3

… overflow

ref_gl: fix codestyle

ab23ae0

a1batross closed this Oct 16, 2023

a1batross deleted the gl2shim-rebase branch October 16, 2023 03:48

a1batross restored the gl2shim-rebase branch October 16, 2023 03:49

a1batross deleted the gl2shim-rebase branch October 31, 2023 23:42

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

gl2shim for core/gles2+ contexts #1446

gl2shim for core/gles2+ contexts #1446

mittorn commented Oct 13, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

a1batross Oct 14, 2023

lewa-j Oct 14, 2023

a1batross Oct 14, 2023

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

a1batross commented Oct 15, 2023

mittorn commented Oct 15, 2023 via email

mittorn commented Oct 15, 2023 via email

a1batross commented Oct 16, 2023

gl2shim for core/gles2+ contexts #1446

gl2shim for core/gles2+ contexts #1446

Conversation

mittorn commented Oct 13, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

mittorn commented Oct 14, 2023 via email

a1batross commented Oct 15, 2023

mittorn commented Oct 15, 2023 via email

mittorn commented Oct 15, 2023 via email

a1batross commented Oct 16, 2023