efl/src/modules/evas/engines/gl_x11/evas_engine.c

3065 lines
88 KiB
C
Raw Normal View History

#include "evas_common.h" /* Also includes international specific stuff */
#include "evas_engine.h"
#include "evas_gl_core_private.h"
#ifdef HAVE_DLSYM
# include <dlfcn.h> /* dlopen,dlclose,etc */
#else
# error gl_x11 should not get compiled if dlsym is not found on the system!
#endif
#define EVAS_GL_NO_GL_H_CHECK 1
#include "Evas_GL.h"
enum {
MODE_FULL,
MODE_COPY,
MODE_DOUBLE,
MODE_TRIPLE
};
typedef struct _Render_Engine Render_Engine;
struct _Render_Engine
{
Tilebuf_Rect *rects;
Tilebuf_Rect *rects_prev[3];
Eina_Inlist *cur_rect;
Evas_GL_X11_Window *win;
Evas_Engine_Info_GL_X11 *info;
Evas *evas;
Tilebuf *tb;
int end;
int mode;
int w, h;
int vsync;
int lost_back;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
EVGL_Engine *evgl_engine;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
};
static int initted = 0;
static int gl_wins = 0;
static int _ext_initted = 0;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
typedef void (*_eng_fn) (void);
typedef _eng_fn (*glsym_func_eng_fn) ();
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
typedef void (*glsym_func_void) ();
typedef void *(*glsym_func_void_ptr) ();
typedef int (*glsym_func_int) ();
typedef unsigned int (*glsym_func_uint) ();
typedef const char *(*glsym_func_const_char_ptr) ();
#ifdef GL_GLES
#ifndef EGL_NATIVE_PIXMAP_KHR
# define EGL_NATIVE_PIXMAP_KHR 0x30b0
#endif
_eng_fn (*glsym_eglGetProcAddress) (const char *a) = NULL;
void *(*glsym_eglCreateImage) (EGLDisplay a, EGLContext b, EGLenum c, EGLClientBuffer d, const int *e) = NULL;
void (*glsym_eglDestroyImage) (EGLDisplay a, void *b) = NULL;
void (*glsym_glEGLImageTargetTexture2DOES) (int a, void *b) = NULL;
void *(*glsym_eglMapImageSEC) (void *a, void *b) = NULL;
unsigned int (*glsym_eglUnmapImageSEC) (void *a, void *b) = NULL;
const char *(*glsym_eglQueryString) (EGLDisplay a, int name) = NULL;
#else
typedef XID (*glsym_func_xid) ();
_eng_fn (*glsym_glXGetProcAddress) (const char *a) = NULL;
void (*glsym_glXBindTexImage) (Display *a, GLXDrawable b, int c, int *d) = NULL;
void (*glsym_glXReleaseTexImage) (Display *a, GLXDrawable b, int c) = NULL;
int (*glsym_glXGetVideoSync) (unsigned int *a) = NULL;
int (*glsym_glXWaitVideoSync) (int a, int b, unsigned int *c) = NULL;
XID (*glsym_glXCreatePixmap) (Display *a, void *b, Pixmap c, const int *d) = NULL;
void (*glsym_glXDestroyPixmap) (Display *a, XID b) = NULL;
void (*glsym_glXQueryDrawable) (Display *a, XID b, int c, unsigned int *d) = NULL;
int (*glsym_glXSwapIntervalSGI) (int a) = NULL;
void (*glsym_glXSwapIntervalEXT) (Display *s, GLXDrawable b, int c) = NULL;
#endif
//----------------------------------------------------------//
// NEW_EVAS_GL Engine Functions
static void *
evgl_eng_display_get(void *data)
{
Render_Engine *re = (Render_Engine *)data;
if (!re)
{
ERR("Invalid Render Engine Data!");
return 0;
}
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
#ifdef GL_GLES
if (re->win)
return (void*)re->win->egl_disp;
#else
if (re->info)
return (void*)re->info->info.display;
#endif
else
return NULL;
}
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
static void *
evgl_eng_evas_surface_get(void *data)
{
Render_Engine *re = (Render_Engine *)data;
if (!re)
{
ERR("Invalid Render Engine Data!");
return 0;
}
#ifdef GL_GLES
if (re->win)
return (void*)re->win->egl_surface[0];
#else
if (re->win)
return (void*)re->win->win;
#endif
else
return NULL;
}
static int
evgl_eng_make_current(void *data, void *surface, void *context, int flush)
{
Render_Engine *re = (Render_Engine *)data;
int ret = 0;
if (!re)
{
ERR("Invalid Render Engine Data!");
return 0;
}
#ifdef GL_GLES
EGLContext ctx = (EGLContext)context;
EGLSurface sfc = (EGLSurface)surface;
EGLDisplay dpy = re->win->egl_disp; //eglGetCurrentDisplay();
if ((context==NULL) && (surface==NULL))
{
ret = eglMakeCurrent(dpy, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
if (!ret)
{
ERR("eglMakeCurrent() failed! Error Code=%#x", eglGetError());
return 0;
}
return 1;
}
if ((eglGetCurrentContext() != ctx) ||
(eglGetCurrentSurface(EGL_READ) != sfc) ||
(eglGetCurrentSurface(EGL_DRAW) != sfc) )
{
//!!!! Does it need to be flushed with it's set to NULL above??
// Flush remainder of what's in Evas' pipeline
if (flush) eng_window_use(NULL);
// Do a make current
ret = eglMakeCurrent(dpy, sfc, sfc, ctx);
if (!ret)
{
ERR("eglMakeCurrent() failed! Error Code=%#x", eglGetError());
return 0;
}
}
return 1;
#else
GLXContext ctx = (GLXContext)context;
Window sfc = (Window)surface;
if ((context==NULL) && (surface==NULL))
{
ret = glXMakeCurrent(re->info->info.display, None, NULL);
if (!ret)
{
ERR("glXMakeCurrent() failed!");
return 0;
}
return 1;
}
if ((glXGetCurrentContext() != ctx))
{
//!!!! Does it need to be flushed with it's set to NULL above??
// Flush remainder of what's in Evas' pipeline
if (flush) eng_window_use(NULL);
// Do a make current
ret = glXMakeCurrent(re->info->info.display, sfc, ctx);
if (!ret)
{
ERR("glXMakeCurrent() failed. Ret: %d! Context: %p Surface: %p", ret, (void*)ctx, (void*)sfc);
return 0;
}
}
return 1;
#endif
}
static void *
evgl_eng_native_window_create(void *data)
{
Render_Engine *re = (Render_Engine *)data;
if (!re)
{
ERR("Invalid Render Engine Data!");
return NULL;
}
XSetWindowAttributes attr;
Window win;
attr.backing_store = NotUseful;
attr.override_redirect = True;
attr.border_pixel = 0;
attr.background_pixmap = None;
attr.bit_gravity = NorthWestGravity;
attr.win_gravity = NorthWestGravity;
attr.save_under = False;
attr.do_not_propagate_mask = NoEventMask;
attr.event_mask = 0;
win = XCreateWindow(re->info->info.display,
DefaultRootWindow(re->info->info.display),
0, 0, 2, 2, 0,
CopyFromParent, InputOutput, CopyFromParent,
CWBackingStore | CWOverrideRedirect |
CWBorderPixel | CWBackPixmap |
CWSaveUnder | CWDontPropagate |
CWEventMask | CWBitGravity |
CWWinGravity, &attr);
if (!win)
{
ERR("Creating native X window failed.");
return NULL;
}
return (void*)win;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
static int
evgl_eng_native_window_destroy(void *data, void *native_window)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
Render_Engine *re = (Render_Engine *)data;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
if (!re)
{
ERR("Invalid Render Engine Data!");
return 0;
}
if (!native_window)
{
ERR("Inavlid native surface.");
return 0;
}
XDestroyWindow(re->info->info.display, (Window)native_window);
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
native_window = NULL;
return 1;
}
// Theoretically, we wouldn't need this functoin if the surfaceless context
// is supported. But, until then...
static void *
evgl_eng_window_surface_create(void *data, void *native_window)
{
Render_Engine *re = (Render_Engine *)data;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
if (!re)
{
ERR("Invalid Render Engine Data!");
return NULL;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
#ifdef GL_GLES
EGLSurface surface = EGL_NO_SURFACE;
// Create resource surface for EGL
surface = eglCreateWindowSurface(re->win->egl_disp,
re->win->egl_config,
(EGLNativeWindowType)native_window,
NULL);
if (!surface)
{
ERR("Creating window surface failed. Error: %#x.", eglGetError());
return NULL;
}
return (void*)surface;
#else
/*
// We don't need to create new one for GLX
Window surface;
surface = re->win->win;
return (void *)surface;
*/
return (void*)native_window;
#endif
2002-11-08 00:02:15 -08:00
}
static int
evgl_eng_window_surface_destroy(void *data, void *surface)
{
Render_Engine *re = (Render_Engine *)data;
if (!re)
{
ERR("Invalid Render Engine Data!");
return 0;
}
#ifdef GL_GLES
if (!surface)
{
ERR("Invalid surface.");
return 0;
}
eglDestroySurface(re->win->egl_disp, (EGLSurface)surface);
#endif
return 1;
if (surface) return 0;
}
static void *
evgl_eng_context_create(void *data, void *share_ctx)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
Render_Engine *re = (Render_Engine *)data;
if (!re)
{
ERR("Invalid Render Engine Data!");
return NULL;
}
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
#ifdef GL_GLES
EGLContext context = EGL_NO_CONTEXT;
int context_attrs[3];
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
context_attrs[0] = EGL_CONTEXT_CLIENT_VERSION;
context_attrs[1] = 2;
context_attrs[2] = EGL_NONE;
// Share context already assumes that it's sharing with evas' context
if (share_ctx)
{
context = eglCreateContext(re->win->egl_disp,
re->win->egl_config,
(EGLContext)share_ctx,
context_attrs);
}
else
{
context = eglCreateContext(re->win->egl_disp,
re->win->egl_config,
re->win->egl_context[0], // Evas' GL Context
context_attrs);
}
if (!context)
{
ERR("Engine Context Creations Failed. Error: %#x.", eglGetError());
return NULL;
}
return (void*)context;
#else
GLXContext context = NULL;
// Share context already assumes that it's sharing with evas' context
if (share_ctx)
{
context = glXCreateContext(re->info->info.display,
re->win->visualinfo,
(GLXContext)share_ctx,
1);
}
else
{
context = glXCreateContext(re->info->info.display,
re->win->visualinfo,
re->win->context, // Evas' GL Context
1);
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
if (!context)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
ERR("Internal Resource Context Creations Failed.");
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
return NULL;
}
return (void*)context;
#endif
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
static int
evgl_eng_context_destroy(void *data, void *context)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
Render_Engine *re = (Render_Engine *)data;
if ((!re) || (!context))
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
ERR("Invalid Render Input Data. Engine: %p, Context: %p", data, context);
return 0;
}
#ifdef GL_GLES
eglDestroyContext(re->win->egl_disp, (EGLContext)context);
#else
glXDestroyContext(re->info->info.display, (GLXContext)context);
#endif
return 1;
}
static const char *
evgl_eng_string_get(void *data)
{
Render_Engine *re = (Render_Engine *)data;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
if (!re)
{
ERR("Invalid Render Engine Data!");
return NULL;
}
#ifdef GL_GLES
// EGL Extensions
if (glsym_eglQueryString)
return glsym_eglQueryString(re->win->egl_disp, EGL_EXTENSIONS);
else
return "";
#else
return glXQueryExtensionsString(re->info->info.display,
re->info->info.screen);
#endif
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
static void *
evgl_eng_proc_address_get(const char *name)
{
#ifdef GL_GLES
if (glsym_eglGetProcAddress) return glsym_eglGetProcAddress(name);
return dlsym(RTLD_DEFAULT, name);
#else
if (glsym_glXGetProcAddress) return glsym_glXGetProcAddress(name);
return dlsym(RTLD_DEFAULT, name);
#endif
}
static int
evgl_eng_rotation_angle_get(void *data)
{
Render_Engine *re = (Render_Engine *)data;
if (!re)
{
ERR("Invalid Render Engine Data!");
return 0;
}
if ((re->win) && (re->win->gl_context))
return re->win->gl_context->rot;
else
{
ERR("Unable to retrieve rotation angle.");
return 0;
}
}
static EVGL_Interface evgl_funcs =
{
evgl_eng_display_get,
evgl_eng_evas_surface_get,
evgl_eng_native_window_create,
evgl_eng_native_window_destroy,
evgl_eng_window_surface_create,
evgl_eng_window_surface_destroy,
evgl_eng_context_create,
evgl_eng_context_destroy,
evgl_eng_make_current,
evgl_eng_proc_address_get,
evgl_eng_string_get,
evgl_eng_rotation_angle_get
};
//----------------------------------------------------------//
static void
_ext_sym_init(void)
{
static int done = 0;
if (done) return;
#ifdef GL_GLES
#define FINDSYM(dst, sym, typ) \
if ((!dst) && (glsym_eglGetProcAddress)) dst = (typ)glsym_eglGetProcAddress(sym); \
if (!dst) dst = (typ)dlsym(RTLD_DEFAULT, sym)
FINDSYM(glsym_eglGetProcAddress, "eglGetProcAddress", glsym_func_eng_fn);
FINDSYM(glsym_eglGetProcAddress, "eglGetProcAddressEXT", glsym_func_eng_fn);
FINDSYM(glsym_eglGetProcAddress, "eglGetProcAddressARB", glsym_func_eng_fn);
FINDSYM(glsym_eglGetProcAddress, "eglGetProcAddressKHR", glsym_func_eng_fn);
FINDSYM(glsym_eglCreateImage, "eglCreateImage", glsym_func_void_ptr);
FINDSYM(glsym_eglCreateImage, "eglCreateImageEXT", glsym_func_void_ptr);
FINDSYM(glsym_eglCreateImage, "eglCreateImageARB", glsym_func_void_ptr);
FINDSYM(glsym_eglCreateImage, "eglCreateImageKHR", glsym_func_void_ptr);
FINDSYM(glsym_eglDestroyImage, "eglDestroyImage", glsym_func_void);
FINDSYM(glsym_eglDestroyImage, "eglDestroyImageEXT", glsym_func_void);
FINDSYM(glsym_eglDestroyImage, "eglDestroyImageARB", glsym_func_void);
FINDSYM(glsym_eglDestroyImage, "eglDestroyImageKHR", glsym_func_void);
FINDSYM(glsym_glEGLImageTargetTexture2DOES, "glEGLImageTargetTexture2DOES", glsym_func_void);
FINDSYM(glsym_eglMapImageSEC, "eglMapImageSEC", glsym_func_void_ptr);
FINDSYM(glsym_eglUnmapImageSEC, "eglUnmapImageSEC", glsym_func_uint);
FINDSYM(glsym_eglQueryString, "eglQueryString", glsym_func_const_char_ptr);
#else
#define FINDSYM(dst, sym, typ) \
if ((!dst) && (glsym_glXGetProcAddress)) dst = (typ)glsym_glXGetProcAddress(sym); \
if (!dst) dst = (typ)dlsym(RTLD_DEFAULT, sym)
FINDSYM(glsym_glXGetProcAddress, "glXGetProcAddress", glsym_func_eng_fn);
FINDSYM(glsym_glXGetProcAddress, "glXGetProcAddressEXT", glsym_func_eng_fn);
FINDSYM(glsym_glXGetProcAddress, "glXGetProcAddressARB", glsym_func_eng_fn);
FINDSYM(glsym_glXBindTexImage, "glXBindTexImage", glsym_func_void);
FINDSYM(glsym_glXBindTexImage, "glXBindTexImageEXT", glsym_func_void);
FINDSYM(glsym_glXBindTexImage, "glXBindTexImageARB", glsym_func_void);
FINDSYM(glsym_glXReleaseTexImage, "glXReleaseTexImage", glsym_func_void);
FINDSYM(glsym_glXReleaseTexImage, "glXReleaseTexImageEXT", glsym_func_void);
FINDSYM(glsym_glXReleaseTexImage, "glXReleaseTexImageARB", glsym_func_void);
FINDSYM(glsym_glXGetVideoSync, "glXGetVideoSyncSGI", glsym_func_int);
FINDSYM(glsym_glXWaitVideoSync, "glXWaitVideoSyncSGI", glsym_func_int);
FINDSYM(glsym_glXCreatePixmap, "glXCreatePixmap", glsym_func_xid);
FINDSYM(glsym_glXCreatePixmap, "glXCreatePixmapEXT", glsym_func_xid);
FINDSYM(glsym_glXCreatePixmap, "glXCreatePixmapARB", glsym_func_xid);
FINDSYM(glsym_glXDestroyPixmap, "glXDestroyPixmap", glsym_func_void);
FINDSYM(glsym_glXDestroyPixmap, "glXDestroyPixmapEXT", glsym_func_void);
FINDSYM(glsym_glXDestroyPixmap, "glXDestroyPixmapARB", glsym_func_void);
FINDSYM(glsym_glXQueryDrawable, "glXQueryDrawable", glsym_func_void);
FINDSYM(glsym_glXQueryDrawable, "glXQueryDrawableEXT", glsym_func_void);
FINDSYM(glsym_glXQueryDrawable, "glXQueryDrawableARB", glsym_func_void);
FINDSYM(glsym_glXSwapIntervalSGI, "glXSwapIntervalMESA", glsym_func_int);
FINDSYM(glsym_glXSwapIntervalSGI, "glXSwapIntervalSGI", glsym_func_int);
FINDSYM(glsym_glXSwapIntervalEXT, "glXSwapIntervalEXT", glsym_func_void);
#endif
}
int _evas_engine_GL_X11_log_dom = -1;
/* function tables - filled in later (func and parent func) */
static Evas_Func func, pfunc;
static void *
eng_info(Evas *eo_e EINA_UNUSED)
{
Evas_Engine_Info_GL_X11 *info;
info = calloc(1, sizeof(Evas_Engine_Info_GL_X11));
info->magic.magic = rand();
info->func.best_visual_get = eng_best_visual_get;
info->func.best_colormap_get = eng_best_colormap_get;
info->func.best_depth_get = eng_best_depth_get;
info->render_mode = EVAS_RENDER_MODE_BLOCKING;
return info;
}
static void
eng_info_free(Evas *eo_e EINA_UNUSED, void *info)
{
Evas_Engine_Info_GL_X11 *in;
// dont free! why bother? its not worth it
// eina_log_domain_unregister(_evas_engine_GL_X11_log_dom);
in = (Evas_Engine_Info_GL_X11 *)info;
free(in);
}
static int
_re_wincheck(Render_Engine *re)
{
if (re->win->surf) return 1;
eng_window_resurf(re->win);
re->lost_back = 1;
if (!re->win->surf)
{
ERR("GL engine can't re-create window surface!");
}
return 0;
}
static void
_re_winfree(Render_Engine *re)
{
if (!re->win->surf) return;
eng_window_unsurf(re->win);
}
static int
eng_setup(Evas *eo_e, void *in)
2002-11-08 00:02:15 -08:00
{
Evas_Public_Data *e = eo_data_get(eo_e, EVAS_CLASS);
2002-11-08 00:02:15 -08:00
Render_Engine *re;
Evas_Engine_Info_GL_X11 *info;
const char *s;
2005-05-21 19:49:50 -07:00
info = (Evas_Engine_Info_GL_X11 *)in;
2002-11-08 00:02:15 -08:00
if (!e->engine.data.output)
{
#ifdef GL_GLES
#else
int eb, evb;
if (!glXQueryExtension(info->info.display, &eb, &evb)) return 0;
#endif
re = calloc(1, sizeof(Render_Engine));
if (!re) return 0;
re->info = info;
re->evas = (Evas *)e;
re->w = e->output.w;
re->h = e->output.h;
re->win = eng_window_new(re->info->info.display,
re->info->info.drawable,
re->info->info.screen,
re->info->info.visual,
re->info->info.colormap,
re->info->info.depth,
re->w,
re->h,
re->info->indirect,
re->info->info.destination_alpha,
re->info->info.rotation);
if (!re->win)
{
free(re);
return 0;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
e->engine.data.output = re;
gl_wins++;
re->evgl_engine = evgl_engine_create(&evgl_funcs, (void*)re);
if (!re->evgl_engine)
ERR("Error Creating Evas_GL Engine. Evas GL will not be supported!");
if (!initted)
{
evas_common_cpu_init();
evas_common_blend_init();
evas_common_image_init();
evas_common_convert_init();
evas_common_scale_init();
evas_common_rectangle_init();
evas_common_polygon_init();
evas_common_line_init();
evas_common_font_init();
evas_common_draw_init();
evas_common_tilebuf_init();
initted = 1;
}
}
else
{
re = e->engine.data.output;
if (_re_wincheck(re))
2010-01-25 06:02:14 -08:00
{
if ((re->info->info.display != re->win->disp) ||
(re->info->info.drawable != re->win->win) ||
(re->info->info.screen != re->win->screen) ||
(re->info->info.visual != re->win->visual) ||
(re->info->info.colormap != re->win->colormap) ||
(re->info->info.depth != re->win->depth) ||
(re->info->info.destination_alpha != re->win->alpha) ||
(re->info->info.rotation != re->win->rot))
2010-07-28 01:11:30 -07:00
{
int inc = 0;
if (re->win)
{
re->win->gl_context->references++;
eng_window_free(re->win);
inc = 1;
gl_wins--;
}
re->w = e->output.w;
re->h = e->output.h;
re->win = eng_window_new(re->info->info.display,
re->info->info.drawable,
re->info->info.screen,
re->info->info.visual,
re->info->info.colormap,
re->info->info.depth,
re->w,
re->h,
re->info->indirect,
re->info->info.destination_alpha,
re->info->info.rotation);
eng_window_use(re->win);
if (re->win) gl_wins++;
if ((re->win) && (inc))
re->win->gl_context->references--;
}
else if ((re->win->w != e->output.w) ||
(re->win->h != e->output.h))
{
re->w = e->output.w;
re->h = e->output.h;
re->win->w = e->output.w;
re->win->h = e->output.h;
eng_window_use(re->win);
evas_gl_common_context_resize(re->win->gl_context, re->win->w, re->win->h, re->win->rot);
2010-07-28 01:11:30 -07:00
}
2010-01-25 06:02:14 -08:00
}
}
if ((s = getenv("EVAS_GL_SWAP_MODE")))
{
if ((!strcasecmp(s, "full")) ||
(!strcasecmp(s, "f")))
re->mode = MODE_FULL;
else if ((!strcasecmp(s, "copy")) ||
(!strcasecmp(s, "c")))
re->mode = MODE_COPY;
else if ((!strcasecmp(s, "double")) ||
(!strcasecmp(s, "d")) ||
(!strcasecmp(s, "2")))
re->mode = MODE_DOUBLE;
else if ((!strcasecmp(s, "triple")) ||
(!strcasecmp(s, "t")) ||
(!strcasecmp(s, "3")))
re->mode = MODE_TRIPLE;
}
else
{
// in most gl implementations - egl and glx here that we care about the TEND
// to either swap or copy backbuffer and front buffer, but strictly that is
// not true. technically backbuffer content is totally undefined after a swap
// and thus you MUST re-render all of it, thus MODE_FULL
re->mode = MODE_FULL;
// BUT... reality is that lmost every implementation copies or swaps so
// triple buffer mode can be used as it is a superset of double buffer and
// copy (though using those explicitly is more efficient). so let's play with
// triple buffer mdoe as a default and see.
// re->mode = MODE_TRIPLE;
// XXX: note - the above seems to break on some older intel chipsets and
// drivers. it seems we CANT depend on backbuffer staying around. bugger!
switch (info->swap_mode)
{
case EVAS_ENGINE_GL_X11_SWAP_MODE_FULL:
re->mode = MODE_FULL;
break;
case EVAS_ENGINE_GL_X11_SWAP_MODE_COPY:
re->mode = MODE_COPY;
break;
case EVAS_ENGINE_GL_X11_SWAP_MODE_DOUBLE:
re->mode = MODE_DOUBLE;
break;
case EVAS_ENGINE_GL_X11_SWAP_MODE_TRIPLE:
re->mode = MODE_TRIPLE;
break;
default:
break;
}
}
if (!re->win)
{
free(re);
return 0;
}
if (!e->engine.data.output)
{
if (re->win)
{
eng_window_free(re->win);
gl_wins--;
}
free(re);
return 0;
}
re->tb = evas_common_tilebuf_new(re->win->w, re->win->h);
if (!re->tb)
{
if (re->win)
{
eng_window_free(re->win);
gl_wins--;
}
free(re);
return 0;
}
evas_common_tilebuf_set_tile_size(re->tb, TILESIZE, TILESIZE);
2002-11-08 00:02:15 -08:00
if (!e->engine.data.context)
e->engine.data.context =
e->engine.func->context_new(e->engine.data.output);
eng_window_use(re->win);
re->vsync = 0;
if (!_ext_initted)
{
_ext_sym_init();
_ext_initted = 1 ;
}
return 1;
2002-11-08 00:02:15 -08:00
}
static void
eng_output_free(void *data)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
if (re)
{
#if 0
#ifdef GL_GLES
// Destroy the resource surface
// Only required for EGL case
if (re->surface)
eglDestroySurface(re->win->egl_disp, re->surface);
#endif
// Destroy the resource context
_destroy_internal_context(re, context);
#endif
if (re->win)
{
eng_window_free(re->win);
gl_wins--;
// NEW_EVAS_GL
if (gl_wins==0)
evgl_engine_destroy(re->evgl_engine);
}
evas_common_tilebuf_free(re->tb);
if (re->rects) evas_common_tilebuf_free_render_rects(re->rects);
if (re->rects_prev[0]) evas_common_tilebuf_free_render_rects(re->rects_prev[0]);
if (re->rects_prev[1]) evas_common_tilebuf_free_render_rects(re->rects_prev[1]);
if (re->rects_prev[2]) evas_common_tilebuf_free_render_rects(re->rects_prev[2]);
free(re);
}
if ((initted == 1) && (gl_wins == 0))
{
evas_common_image_shutdown();
evas_common_font_shutdown();
initted = 0;
}
2002-11-08 00:02:15 -08:00
}
static void
eng_output_resize(void *data, int w, int h)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
re->win->w = w;
re->win->h = h;
eng_window_use(re->win);
evas_gl_common_context_resize(re->win->gl_context, w, h, re->win->rot);
evas_common_tilebuf_free(re->tb);
re->tb = evas_common_tilebuf_new(w, h);
if (re->tb)
evas_common_tilebuf_set_tile_size(re->tb, TILESIZE, TILESIZE);
2002-11-08 00:02:15 -08:00
}
static void
eng_output_tile_size_set(void *data, int w, int h)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
re = (Render_Engine *)data;
evas_common_tilebuf_set_tile_size(re->tb, w, h);
2002-11-08 00:02:15 -08:00
}
static void
eng_output_redraws_rect_add(void *data, int x, int y, int w, int h)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
eng_window_use(re->win);
evas_gl_common_context_resize(re->win->gl_context, re->win->w, re->win->h, re->win->rot);
evas_common_tilebuf_add_redraw(re->tb, x, y, w, h);
2002-11-08 00:02:15 -08:00
}
static void
eng_output_redraws_rect_del(void *data, int x, int y, int w, int h)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
re = (Render_Engine *)data;
evas_common_tilebuf_del_redraw(re->tb, x, y, w, h);
2002-11-08 00:02:15 -08:00
}
static void
eng_output_redraws_clear(void *data)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
evas_common_tilebuf_clear(re->tb);
// INF("GL: finish update cycle!");
2002-11-08 00:02:15 -08:00
}
static Tilebuf_Rect *
_merge_rects(Tilebuf *tb, Tilebuf_Rect *r1, Tilebuf_Rect *r2, Tilebuf_Rect *r3)
{
Tilebuf_Rect *r, *rects;
Evas_Point p1, p2;
if (r1)
{
EINA_INLIST_FOREACH(EINA_INLIST_GET(r1), r)
{
evas_common_tilebuf_add_redraw(tb, r->x, r->y, r->w, r->h);
}
}
if (r2)
{
EINA_INLIST_FOREACH(EINA_INLIST_GET(r2), r)
{
evas_common_tilebuf_add_redraw(tb, r->x, r->y, r->w, r->h);
}
}
if (r2)
{
EINA_INLIST_FOREACH(EINA_INLIST_GET(r3), r)
{
evas_common_tilebuf_add_redraw(tb, r->x, r->y, r->w, r->h);
}
}
rects = evas_common_tilebuf_get_render_rects(tb);
// bounding box -> make a bounding box single region update of all regions.
// yes we could try and be smart and figure out size of regions, how far
// apart etc. etc. to try and figure out an optimal "set". this is a tradeoff
// between multiple update regions to render and total pixels to render.
if (rects)
{
p1.x = rects->x; p1.y = rects->y;
p2.x = rects->x + rects->w; p2.y = rects->y + rects->h;
EINA_INLIST_FOREACH(EINA_INLIST_GET(rects), r)
{
if (r->x < p1.x) p1.x = r->x;
if (r->y < p1.y) p1.y = r->y;
if ((r->x + r->w) > p2.x) p2.x = r->x + r->w;
if ((r->y + r->h) > p2.y) p2.y = r->y + r->h;
}
evas_common_tilebuf_free_render_rects(rects);
rects = calloc(1, sizeof(Tilebuf_Rect));
if (rects)
{
rects->x = p1.x;
rects->y = p1.y;
rects->w = p2.x - p1.x;
rects->h = p2.y - p1.y;
}
}
evas_common_tilebuf_clear(tb);
return rects;
}
/* vsync games - not for now though */
#define VSYNC_TO_SCREEN 1
2002-11-08 00:02:15 -08:00
static void *
eng_output_redraws_next_update_get(void *data, int *x, int *y, int *w, int *h, int *cx, int *cy, int *cw, int *ch)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
Tilebuf_Rect *rect;
int i;
Eina_Bool first_rect = EINA_FALSE;
#define CLEAR_PREV_RECTS(x) \
do { \
if (re->rects_prev[i]) \
evas_common_tilebuf_free_render_rects(re->rects_prev[i]); \
re->rects_prev[i] = NULL; \
} while (0)
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
/* get the upate rect surface - return engine data as dummy */
if (re->end)
{
re->end = 0;
return NULL;
}
if (!re->rects)
{
re->rects = evas_common_tilebuf_get_render_rects(re->tb);
if (re->rects)
{
if (re->lost_back)
{
/* if we lost our backbuffer since the last frame redraw all */
re->lost_back = 0;
evas_common_tilebuf_add_redraw(re->tb, 0, 0, re->win->w, re->win->h);
evas_common_tilebuf_free_render_rects(re->rects);
re->rects = evas_common_tilebuf_get_render_rects(re->tb);
}
/* ensure we get rid of previous rect lists we dont need if mode
* changed/is appropriate */
switch (re->mode)
{
case MODE_FULL:
case MODE_COPY: // no prev rects needed
for (i = 0; i < 3; i++) CLEAR_PREV_RECTS(i);
break;
case MODE_DOUBLE: // double mode - only 1 level of prev rect
for (i = 1; i < 3; i++) CLEAR_PREV_RECTS(i);
re->rects_prev[1] = re->rects_prev[0];
re->rects_prev[0] = re->rects;
// merge prev[1] + prev[0] -> rects
re->rects = _merge_rects(re->tb, re->rects_prev[0], re->rects_prev[1], NULL);
break;
case MODE_TRIPLE: // keep all
for (i = 2; i < 3; i++) CLEAR_PREV_RECTS(i);
re->rects_prev[2] = re->rects_prev[1];
re->rects_prev[1] = re->rects_prev[0];
re->rects_prev[0] = re->rects;
re->rects = NULL;
// merge prev[2] + prev[1] + prev[0] -> rects
re->rects = _merge_rects(re->tb, re->rects_prev[0], re->rects_prev[1], re->rects_prev[2]);
break;
default:
break;
}
re->cur_rect = EINA_INLIST_GET(re->rects);
first_rect = EINA_TRUE;
}
evas_common_tilebuf_clear(re->tb);
}
if (!re->cur_rect) return NULL;
rect = (Tilebuf_Rect *)re->cur_rect;
if (re->rects)
{
switch (re->mode)
{
case MODE_COPY:
case MODE_DOUBLE:
case MODE_TRIPLE:
rect = (Tilebuf_Rect *)re->cur_rect;
*x = rect->x;
*y = rect->y;
*w = rect->w;
*h = rect->h;
*cx = rect->x;
*cy = rect->y;
*cw = rect->w;
*ch = rect->h;
re->cur_rect = re->cur_rect->next;
break;
case MODE_FULL:
re->cur_rect = NULL;
if (x) *x = 0;
if (y) *y = 0;
if (w) *w = re->win->w;
if (h) *h = re->win->h;
if (cx) *cx = 0;
if (cy) *cy = 0;
if (cw) *cw = re->win->w;
if (ch) *ch = re->win->h;
break;
default:
break;
}
if (first_rect)
{
#ifdef GL_GLES
// dont need to for egl - eng_window_use() can check for other ctxt's
#else
eng_window_use(NULL);
#endif
eng_window_use(re->win);
if (!_re_wincheck(re)) return NULL;
evas_gl_common_context_flush(re->win->gl_context);
evas_gl_common_context_newframe(re->win->gl_context);
//// debug partial updates :)
// glClearColor(1.0, 0.5, 0.2, 1.0);
// glClear(GL_COLOR_BUFFER_BIT);
}
if (!re->cur_rect)
{
re->end = 1;
}
return re->win->gl_context->def_surface;
}
return NULL;
2002-11-08 00:02:15 -08:00
}
static int safe_native = -1;
2002-11-08 00:02:15 -08:00
static void
eng_output_redraws_next_update_push(void *data, void *surface EINA_UNUSED, int x EINA_UNUSED, int y EINA_UNUSED, int w EINA_UNUSED, int h EINA_UNUSED)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
/* put back update surface.. in this case just unflag redraw */
if (!_re_wincheck(re)) return;
re->win->draw.drew = 1;
evas_gl_common_context_flush(re->win->gl_context);
if (safe_native == -1)
{
const char *s = getenv("EVAS_GL_SAFE_NATIVE");
safe_native = 0;
if (s) safe_native = atoi(s);
else
{
s = (const char *)glGetString(GL_RENDERER);
if (s)
{
if (strstr(s, "PowerVR SGX 540") ||
strstr(s, "Mali-400 MP"))
safe_native = 1;
}
}
}
#ifdef GL_GLES
// this is needed to make sure all previous rendering is flushed to
// buffers/surfaces
// previous rendering should be done and swapped
if (!safe_native) eglWaitNative(EGL_CORE_NATIVE_ENGINE);
// if (eglGetError() != EGL_SUCCESS)
// {
// printf("Error: eglWaitNative(EGL_CORE_NATIVE_ENGINE) fail.\n");
// }
#else
// previous rendering should be done and swapped
if (!safe_native) glXWaitX();
#endif
2002-11-08 00:02:15 -08:00
}
static void
eng_output_flush(void *data)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
if (!_re_wincheck(re)) return;
if (!re->win->draw.drew) return;
re->win->draw.drew = 0;
eng_window_use(re->win);
#ifdef GL_GLES
if (!re->vsync)
{
if (re->info->vsync) eglSwapInterval(re->win->egl_disp, 1);
else eglSwapInterval(re->win->egl_disp, 0);
re->vsync = 1;
}
if (re->info->callback.pre_swap)
{
re->info->callback.pre_swap(re->info->callback.data, re->evas);
}
// XXX: if partial swaps can be done use re->rects
eglSwapBuffers(re->win->egl_disp, re->win->egl_surface[0]);
if (!safe_native) eglWaitGL();
if (re->info->callback.post_swap)
{
re->info->callback.post_swap(re->info->callback.data, re->evas);
}
// if (eglGetError() != EGL_SUCCESS)
// {
// printf("Error: eglSwapBuffers() fail.\n");
// }
#else
#ifdef VSYNC_TO_SCREEN
if (re->info->vsync)
{
if (glsym_glXSwapIntervalEXT)
{
if (!re->vsync)
{
if (re->info->vsync) glsym_glXSwapIntervalEXT(re->win->disp, re->win->win, 1);
else glsym_glXSwapIntervalEXT(re->win->disp, re->win->win, 0);
re->vsync = 1;
}
}
if (glsym_glXSwapIntervalSGI)
{
if (!re->vsync)
{
if (re->info->vsync) glsym_glXSwapIntervalSGI(1);
else glsym_glXSwapIntervalSGI(0);
re->vsync = 1;
}
}
else
{
if ((glsym_glXGetVideoSync) && (glsym_glXWaitVideoSync))
{
unsigned int rc;
glsym_glXGetVideoSync(&rc);
glsym_glXWaitVideoSync(1, 0, &rc);
}
}
}
#endif
if (re->info->callback.pre_swap)
{
re->info->callback.pre_swap(re->info->callback.data, re->evas);
}
// XXX: if partial swaps can be done use re->rects
glXSwapBuffers(re->win->disp, re->win->win);
if (re->info->callback.post_swap)
{
re->info->callback.post_swap(re->info->callback.data, re->evas);
}
#endif
// clear out rects after swap as we may use them during swap
if (re->rects)
{
evas_common_tilebuf_free_render_rects(re->rects);
re->rects = NULL;
}
2002-11-08 00:02:15 -08:00
}
static void
eng_output_idle_flush(void *data)
{
Render_Engine *re;
re = (Render_Engine *)data;
(void) re;
}
static void
eng_output_dump(void *data)
{
Render_Engine *re;
re = (Render_Engine *)data;
evas_common_image_image_all_unload();
evas_common_font_font_all_unload();
evas_gl_common_image_all_unload(re->win->gl_context);
_re_winfree(re);
}
2002-11-08 00:02:15 -08:00
static void
eng_context_cutout_add(void *data EINA_UNUSED, void *context, int x, int y, int w, int h)
2002-11-08 00:02:15 -08:00
{
// Render_Engine *re;
//
// re = (Render_Engine *)data;
// re->win->gl_context->dc = context;
evas_common_draw_context_add_cutout(context, x, y, w, h);
2002-11-08 00:02:15 -08:00
}
static void
eng_context_cutout_clear(void *data EINA_UNUSED, void *context)
2002-11-08 00:02:15 -08:00
{
// Render_Engine *re;
//
// re = (Render_Engine *)data;
// re->win->gl_context->dc = context;
evas_common_draw_context_clear_cutouts(context);
2002-11-08 00:02:15 -08:00
}
static void
eng_rectangle_draw(void *data, void *context, void *surface, int x, int y, int w, int h)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
eng_window_use(re->win);
evas_gl_common_context_target_surface_set(re->win->gl_context, surface);
re->win->gl_context->dc = context;
evas_gl_common_rect_draw(re->win->gl_context, x, y, w, h);
2002-11-08 00:02:15 -08:00
}
static void
eng_line_draw(void *data, void *context, void *surface, int p1x, int p1y, int p2x, int p2y)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
eng_window_use(re->win);
evas_gl_common_context_target_surface_set(re->win->gl_context, surface);
re->win->gl_context->dc = context;
evas_gl_common_line_draw(re->win->gl_context, p1x, p1y, p2x, p2y);
2002-11-08 00:02:15 -08:00
}
static void *
eng_polygon_point_add(void *data, void *context EINA_UNUSED, void *polygon, int x, int y)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
(void) re;
return evas_gl_common_poly_point_add(polygon, x, y);
2002-11-08 00:02:15 -08:00
}
static void *
eng_polygon_points_clear(void *data, void *context EINA_UNUSED, void *polygon)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
(void) re;
return evas_gl_common_poly_points_clear(polygon);
2002-11-08 00:02:15 -08:00
}
static void
eng_polygon_draw(void *data, void *context, void *surface EINA_UNUSED, void *polygon, int x, int y)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
eng_window_use(re->win);
evas_gl_common_context_target_surface_set(re->win->gl_context, surface);
re->win->gl_context->dc = context;
evas_gl_common_poly_draw(re->win->gl_context, polygon, x, y);
2002-11-08 00:02:15 -08:00
}
static int
eng_image_alpha_get(void *data EINA_UNUSED, void *image)
{
// Render_Engine *re;
Evas_GL_Image *im;
// re = (Render_Engine *)data;
2006-12-17 08:46:30 -08:00
if (!image) return 1;
im = image;
return im->alpha;
}
static int
eng_image_colorspace_get(void *data EINA_UNUSED, void *image)
{
// Render_Engine *re;
Evas_GL_Image *im;
// re = (Render_Engine *)data;
2006-12-17 08:46:30 -08:00
if (!image) return EVAS_COLORSPACE_ARGB8888;
im = image;
return im->cs.space;
}
static void
eng_image_mask_create(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *im;
if (!image) return;
im = image;
if (!im->im->image.data)
evas_cache_image_load_data(&im->im->cache_entry);
if (!im->tex)
im->tex = evas_gl_common_texture_new(im->gc, im->im);
}
static void *
eng_image_alpha_set(void *data, void *image, int has_alpha)
{
Render_Engine *re;
Evas_GL_Image *im;
re = (Render_Engine *)data;
if (!image) return NULL;
im = image;
if (im->alpha == has_alpha) return image;
if (im->native.data)
{
im->alpha = has_alpha;
return image;
}
eng_window_use(re->win);
if ((im->tex) && (im->tex->pt->dyn.img))
{
im->alpha = has_alpha;
im->tex->alpha = im->alpha;
return image;
}
/* FIXME: can move to gl_common */
if (im->cs.space != EVAS_COLORSPACE_ARGB8888) return im;
if ((has_alpha) && (im->im->cache_entry.flags.alpha)) return image;
else if ((!has_alpha) && (!im->im->cache_entry.flags.alpha)) return image;
if (im->references > 1)
{
Evas_GL_Image *im_new;
if (!im->im->image.data)
evas_cache_image_load_data(&im->im->cache_entry);
im_new = evas_gl_common_image_new_from_copied_data
(im->gc, im->im->cache_entry.w, im->im->cache_entry.h,
im->im->image.data,
eng_image_alpha_get(data, image),
eng_image_colorspace_get(data, image));
if (!im_new) return im;
evas_gl_common_image_free(im);
im = im_new;
}
else
evas_gl_common_image_dirty(im, 0, 0, 0, 0);
return evas_gl_common_image_alpha_set(im, has_alpha ? 1 : 0);
// im->im->cache_entry.flags.alpha = has_alpha ? 1 : 0;
// return image;
}
static void *
eng_image_border_set(void *data EINA_UNUSED, void *image, int l EINA_UNUSED, int r EINA_UNUSED, int t EINA_UNUSED, int b EINA_UNUSED)
{
// Render_Engine *re;
//
// re = (Render_Engine *)data;
return image;
}
static void
eng_image_border_get(void *data EINA_UNUSED, void *image EINA_UNUSED, int *l EINA_UNUSED, int *r EINA_UNUSED, int *t EINA_UNUSED, int *b EINA_UNUSED)
{
// Render_Engine *re;
//
// re = (Render_Engine *)data;
}
static char *
eng_image_comment_get(void *data EINA_UNUSED, void *image, char *key EINA_UNUSED)
{
// Render_Engine *re;
Evas_GL_Image *im;
// re = (Render_Engine *)data;
if (!image) return NULL;
im = image;
if (!im->im) return NULL;
return im->im->info.comment;
}
static char *
eng_image_format_get(void *data EINA_UNUSED, void *image)
{
// Render_Engine *re;
Evas_GL_Image *im;
// re = (Render_Engine *)data;
im = image;
(void) im;
return NULL;
}
static void
eng_image_colorspace_set(void *data, void *image, int cspace)
{
Render_Engine *re;
Evas_GL_Image *im;
re = (Render_Engine *)data;
if (!image) return;
im = image;
if (im->native.data) return;
/* FIXME: can move to gl_common */
if (im->cs.space == cspace) return;
eng_window_use(re->win);
evas_cache_image_colorspace(&im->im->cache_entry, cspace);
switch (cspace)
{
case EVAS_COLORSPACE_ARGB8888:
if (im->cs.data)
{
if (!im->cs.no_free) free(im->cs.data);
im->cs.data = NULL;
im->cs.no_free = 0;
}
break;
case EVAS_COLORSPACE_YCBCR422P601_PL:
case EVAS_COLORSPACE_YCBCR422P709_PL:
case EVAS_COLORSPACE_YCBCR422601_PL:
case EVAS_COLORSPACE_YCBCR420NV12601_PL:
case EVAS_COLORSPACE_YCBCR420TM12601_PL:
if (im->tex) evas_gl_common_texture_free(im->tex);
im->tex = NULL;
if (im->cs.data)
{
if (!im->cs.no_free) free(im->cs.data);
}
if (im->im->cache_entry.h > 0)
im->cs.data =
calloc(1, im->im->cache_entry.h * sizeof(unsigned char *) * 2);
else
im->cs.data = NULL;
im->cs.no_free = 0;
break;
default:
abort();
break;
}
im->cs.space = cspace;
}
/////////////////////////////////////////////////////////////////////////
//
//
typedef struct _Native Native;
struct _Native
{
Evas_Native_Surface ns;
Pixmap pixmap;
Visual *visual;
#ifdef GL_GLES
void *egl_surface;
#else
void *fbc;
XID glx_pixmap;
#endif
};
// FIXME: this is enabled so updates happen - but its SLOOOOOOOOOOOOOOOW
// (i am sure this is the reason) not to mention seemingly superfluous. but
// i need to enable it for it to work on fglrx at least. havent tried nvidia.
//
// why is this the case? does anyone know? has anyone tried it on other gfx
// drivers?
//
//#define GLX_TEX_PIXMAP_RECREATE 1
static void
_native_bind_cb(void *data, void *image)
{
Evas_GL_Image *im = image;
Native *n = im->native.data;
if (n->ns.type == EVAS_NATIVE_SURFACE_X11)
{
#ifdef GL_GLES
if (n->egl_surface)
{
if (glsym_glEGLImageTargetTexture2DOES)
{
glsym_glEGLImageTargetTexture2DOES(GL_TEXTURE_2D, n->egl_surface);
if (eglGetError() != EGL_SUCCESS)
ERR("glEGLImageTargetTexture2DOES() failed.");
}
else
ERR("Try glEGLImageTargetTexture2DOES on EGL with no support");
}
#else
# ifdef GLX_BIND_TO_TEXTURE_TARGETS_EXT
Render_Engine *re = data;
if (glsym_glXBindTexImage)
{
glsym_glXBindTexImage(re->win->disp, n->glx_pixmap,
GLX_FRONT_LEFT_EXT, NULL);
GLERR(__FUNCTION__, __FILE__, __LINE__, "");
}
else
ERR("Try glXBindTexImage on GLX with no support");
# endif
#endif
}
else if (n->ns.type == EVAS_NATIVE_SURFACE_OPENGL)
{
glBindTexture(GL_TEXTURE_2D, n->ns.data.opengl.texture_id);
GLERR(__FUNCTION__, __FILE__, __LINE__, "");
}
return;
data = NULL;
}
static void
_native_unbind_cb(void *data, void *image)
{
Evas_GL_Image *im = image;
Native *n = im->native.data;
if (n->ns.type == EVAS_NATIVE_SURFACE_X11)
{
#ifdef GL_GLES
// nothing
#else
# ifdef GLX_BIND_TO_TEXTURE_TARGETS_EXT
Render_Engine *re = data;
if (glsym_glXReleaseTexImage)
{
glsym_glXReleaseTexImage(re->win->disp, n->glx_pixmap,
GLX_FRONT_LEFT_EXT);
GLERR(__FUNCTION__, __FILE__, __LINE__, "");
}
else
ERR("Try glXReleaseTexImage on GLX with no support");
# endif
#endif
}
else if (n->ns.type == EVAS_NATIVE_SURFACE_OPENGL)
{
glBindTexture(GL_TEXTURE_2D, 0);
GLERR(__FUNCTION__, __FILE__, __LINE__, "");
}
return;
data = NULL;
}
static void
_native_free_cb(void *data, void *image)
{
Render_Engine *re = data;
Evas_GL_Image *im = image;
Native *n = im->native.data;
uint32_t pmid, texid;
if (n->ns.type == EVAS_NATIVE_SURFACE_X11)
{
pmid = n->pixmap;
eina_hash_del(re->win->gl_context->shared->native_pm_hash, &pmid, im);
#ifdef GL_GLES
if (n->egl_surface)
{
if (glsym_eglDestroyImage)
{
glsym_eglDestroyImage(re->win->egl_disp,
n->egl_surface);
if (eglGetError() != EGL_SUCCESS)
ERR("eglDestroyImage() failed.");
}
else
ERR("Try eglDestroyImage on EGL with no support");
}
#else
# ifdef GLX_BIND_TO_TEXTURE_TARGETS_EXT
if (n->glx_pixmap)
{
if (im->native.loose)
{
if (glsym_glXReleaseTexImage)
{
glsym_glXReleaseTexImage(re->win->disp, n->glx_pixmap,
GLX_FRONT_LEFT_EXT);
GLERR(__FUNCTION__, __FILE__, __LINE__, "");
}
else
ERR("Try glXReleaseTexImage on GLX with no support");
}
if (glsym_glXDestroyPixmap)
{
glsym_glXDestroyPixmap(re->win->disp, n->glx_pixmap);
GLERR(__FUNCTION__, __FILE__, __LINE__, "");
}
else
ERR("Try glXDestroyPixmap on GLX with no support");
n->glx_pixmap = 0;
}
# endif
#endif
}
else if (n->ns.type == EVAS_NATIVE_SURFACE_OPENGL)
{
texid = n->ns.data.opengl.texture_id;
eina_hash_del(re->win->gl_context->shared->native_tex_hash, &texid, im);
}
im->native.data = NULL;
im->native.func.data = NULL;
im->native.func.bind = NULL;
im->native.func.unbind = NULL;
im->native.func.free = NULL;
free(n);
}
static void *
eng_image_native_set(void *data, void *image, void *native)
{
Render_Engine *re = (Render_Engine *)data;
Evas_Native_Surface *ns = native;
Evas_GL_Image *im = image, *im2 = NULL;
Visual *vis = NULL;
Pixmap pm = 0;
Native *n = NULL;
uint32_t pmid, texid;
unsigned int tex = 0;
unsigned int fbo = 0;
if (!im)
{
if ((ns) && (ns->type == EVAS_NATIVE_SURFACE_OPENGL))
{
im = evas_gl_common_image_new_from_data(re->win->gl_context,
ns->data.opengl.w,
ns->data.opengl.h,
NULL, 1,
EVAS_COLORSPACE_ARGB8888);
}
else
return NULL;
}
if (ns)
{
if (ns->type == EVAS_NATIVE_SURFACE_X11)
{
vis = ns->data.x11.visual;
pm = ns->data.x11.pixmap;
if (im->native.data)
{
Evas_Native_Surface *ens = im->native.data;
if ((ens->data.x11.visual == vis) &&
(ens->data.x11.pixmap == pm))
return im;
}
}
else if (ns->type == EVAS_NATIVE_SURFACE_OPENGL)
{
tex = ns->data.opengl.texture_id;
fbo = ns->data.opengl.framebuffer_id;
if (im->native.data)
{
Evas_Native_Surface *ens = im->native.data;
if ((ens->data.opengl.texture_id == tex) &&
(ens->data.opengl.framebuffer_id == fbo))
return im;
}
}
}
if ((!ns) && (!im->native.data)) return im;
eng_window_use(re->win);
if (im->native.data)
{
if (im->native.func.free)
im->native.func.free(im->native.func.data, im);
evas_gl_common_image_native_disable(im);
}
if (!ns) return im;
if (ns->type == EVAS_NATIVE_SURFACE_X11)
{
pmid = pm;
im2 = eina_hash_find(re->win->gl_context->shared->native_pm_hash, &pmid);
if (im2 == im) return im;
if (im2)
{
n = im2->native.data;
if (n)
{
evas_gl_common_image_ref(im2);
evas_gl_common_image_free(im);
return im2;
}
}
}
else if (ns->type == EVAS_NATIVE_SURFACE_OPENGL)
{
texid = tex;
im2 = eina_hash_find(re->win->gl_context->shared->native_tex_hash, &texid);
if (im2 == im) return im;
if (im2)
{
n = im2->native.data;
if (n)
{
evas_gl_common_image_ref(im2);
evas_gl_common_image_free(im);
return im2;
}
}
}
im2 = evas_gl_common_image_new_from_data(re->win->gl_context,
im->w, im->h, NULL, im->alpha,
EVAS_COLORSPACE_ARGB8888);
evas_gl_common_image_free(im);
im = im2;
if (!im) return NULL;
if (ns->type == EVAS_NATIVE_SURFACE_X11)
{
#ifdef GL_GLES
if (native)
{
n = calloc(1, sizeof(Native));
if (n)
{
EGLConfig egl_config;
int config_attrs[20];
int num_config, i = 0;
eina_hash_add(re->win->gl_context->shared->native_pm_hash, &pmid, im);
// assume 32bit pixmap! :)
config_attrs[i++] = EGL_RED_SIZE;
config_attrs[i++] = 8;
config_attrs[i++] = EGL_GREEN_SIZE;
config_attrs[i++] = 8;
config_attrs[i++] = EGL_BLUE_SIZE;
config_attrs[i++] = 8;
config_attrs[i++] = EGL_ALPHA_SIZE;
config_attrs[i++] = 8;
config_attrs[i++] = EGL_DEPTH_SIZE;
config_attrs[i++] = 0;
config_attrs[i++] = EGL_STENCIL_SIZE;
config_attrs[i++] = 0;
config_attrs[i++] = EGL_RENDERABLE_TYPE;
config_attrs[i++] = EGL_OPENGL_ES2_BIT;
config_attrs[i++] = EGL_SURFACE_TYPE;
config_attrs[i++] = EGL_PIXMAP_BIT;
config_attrs[i++] = EGL_NONE;
if (!eglChooseConfig(re->win->egl_disp, config_attrs,
&egl_config, 1, &num_config))
ERR("eglChooseConfig() failed for pixmap 0x%x, num_config = %i", (unsigned int)pm, num_config);
memcpy(&(n->ns), ns, sizeof(Evas_Native_Surface));
n->pixmap = pm;
n->visual = vis;
if (glsym_eglCreateImage)
n->egl_surface = glsym_eglCreateImage(re->win->egl_disp,
EGL_NO_CONTEXT,
EGL_NATIVE_PIXMAP_KHR,
(void *)pm,
NULL);
else
ERR("Try eglCreateImage on EGL with no support");
if (!n->egl_surface)
ERR("eglCreatePixmapSurface() for 0x%x failed", (unsigned int)pm);
im->native.yinvert = 1;
im->native.loose = 0;
im->native.data = n;
im->native.func.data = re;
im->native.func.bind = _native_bind_cb;
im->native.func.unbind = _native_unbind_cb;
im->native.func.free = _native_free_cb;
im->native.target = GL_TEXTURE_2D;
im->native.mipmap = 0;
evas_gl_common_image_native_enable(im);
}
}
#else
# ifdef GLX_BIND_TO_TEXTURE_TARGETS_EXT
if (native)
{
int dummy;
unsigned int w, h, depth = 32, border;
Window wdummy;
// fixme: round trip :(
XGetGeometry(re->win->disp, pm, &wdummy, &dummy, &dummy,
&w, &h, &border, &depth);
if (depth <= 32)
{
n = calloc(1, sizeof(Native));
if (n)
{
int pixmap_att[20], i;
int config_attrs[40], num = 0;
int tex_format = 0, tex_target = 0, yinvert = 0, mipmap = 0;
unsigned int target;
GLXFBConfig *configs;
i = 0;
config_attrs[i++] = GLX_BUFFER_SIZE;
config_attrs[i++] = depth;
if (depth == 32)
{
config_attrs[i++] = GLX_BIND_TO_TEXTURE_RGBA_EXT;
config_attrs[i++] = 1;
}
else
{
config_attrs[i++] = GLX_BIND_TO_TEXTURE_RGB_EXT;
config_attrs[i++] = 1;
}
#ifndef GLX_VISUAL_ID
# define GLX_VISUAL_ID 0x800b
#endif
config_attrs[i++] = GLX_VISUAL_ID;
config_attrs[i++] = XVisualIDFromVisual(vis);
#ifndef GLX_SAMPLE_BUFFERS
# define GLX_SAMPLE_BUFFERS 0x186a0
#endif
config_attrs[i++] = GLX_SAMPLE_BUFFERS;
config_attrs[i++] = 0;
config_attrs[i++] = GLX_DEPTH_SIZE;
config_attrs[i++] = 0;
config_attrs[i++] = GLX_STENCIL_SIZE;
config_attrs[i++] = 0;
config_attrs[i++] = GLX_AUX_BUFFERS;
config_attrs[i++] = 0;
config_attrs[i++] = GLX_STEREO;
config_attrs[i++] = 0;
config_attrs[i++] = 0;
configs = glXChooseFBConfig(re->win->disp,
re->win->screen,
config_attrs,
&num);
if (configs)
{
int j = 0, val = 0;
tex_format = GLX_TEXTURE_FORMAT_RGB_EXT;
glXGetFBConfigAttrib(re->win->disp, configs[j],
GLX_ALPHA_SIZE, &val);
if (val > 0)
{
glXGetFBConfigAttrib(re->win->disp, configs[j],
GLX_BIND_TO_TEXTURE_RGBA_EXT, &val);
if (val) tex_format = GLX_TEXTURE_FORMAT_RGBA_EXT;
}
else
{
glXGetFBConfigAttrib(re->win->disp, configs[j],
GLX_BIND_TO_TEXTURE_RGB_EXT, &val);
if (val) tex_format = GLX_TEXTURE_FORMAT_RGB_EXT;
}
glXGetFBConfigAttrib(re->win->disp, configs[j],
GLX_Y_INVERTED_EXT, &val);
yinvert = val;
glXGetFBConfigAttrib(re->win->disp, configs[j],
GLX_BIND_TO_TEXTURE_TARGETS_EXT,
&val);
tex_target = val;
glXGetFBConfigAttrib(re->win->disp, configs[j],
GLX_BIND_TO_MIPMAP_TEXTURE_EXT, &val);
mipmap = val;
n->fbc = configs[j];
XFree(configs);
}
eina_hash_add(re->win->gl_context->shared->native_pm_hash, &pmid, im);
if ((tex_target & GLX_TEXTURE_2D_BIT_EXT))
target = GLX_TEXTURE_2D_EXT;
else if ((target & GLX_TEXTURE_RECTANGLE_BIT_EXT))
{
ERR("rect!!! (not handled)");
target = GLX_TEXTURE_RECTANGLE_EXT;
}
if (!target)
{
ERR("broken tex-from-pixmap");
if (!(tex_target & GLX_TEXTURE_2D_BIT_EXT))
target = GLX_TEXTURE_RECTANGLE_EXT;
else if (!(tex_target & GLX_TEXTURE_RECTANGLE_BIT_EXT))
target = GLX_TEXTURE_2D_EXT;
}
i = 0;
pixmap_att[i++] = GLX_TEXTURE_FORMAT_EXT;
pixmap_att[i++] = tex_format;
pixmap_att[i++] = GLX_MIPMAP_TEXTURE_EXT;
pixmap_att[i++] = mipmap;
if (target)
{
pixmap_att[i++] = GLX_TEXTURE_TARGET_EXT;
pixmap_att[i++] = target;
}
pixmap_att[i++] = 0;
memcpy(&(n->ns), ns, sizeof(Evas_Native_Surface));
n->pixmap = pm;
n->visual = vis;
if (glsym_glXCreatePixmap)
n->glx_pixmap = glsym_glXCreatePixmap(re->win->disp,
n->fbc,
n->pixmap,
pixmap_att);
else
ERR("Try glXCreatePixmap on GLX with no support");
if (n->glx_pixmap)
{
// printf("%p: new native texture for %x | %4i x %4i @ %2i = %p\n",
// n, pm, w, h, depth, n->glx_pixmap);
if (!target)
{
ERR("no target :(");
if (glsym_glXQueryDrawable)
glsym_glXQueryDrawable(re->win->disp,
n->pixmap,
GLX_TEXTURE_TARGET_EXT,
&target);
}
if (target == GLX_TEXTURE_2D_EXT)
{
im->native.target = GL_TEXTURE_2D;
im->native.mipmap = mipmap;
}
# ifdef GL_TEXTURE_RECTANGLE_ARB
else if (target == GLX_TEXTURE_RECTANGLE_EXT)
{
im->native.target = GL_TEXTURE_RECTANGLE_ARB;
im->native.mipmap = 0;
}
# endif
else
{
im->native.target = GL_TEXTURE_2D;
im->native.mipmap = 0;
ERR("still unknown target");
}
}
else
ERR("GLX Pixmap create fail");
im->native.yinvert = yinvert;
im->native.loose = re->win->detected.loose_binding;
im->native.data = n;
im->native.func.data = re;
im->native.func.bind = _native_bind_cb;
im->native.func.unbind = _native_unbind_cb;
im->native.func.free = _native_free_cb;
evas_gl_common_image_native_enable(im);
}
}
}
# endif
#endif
}
else if (ns->type == EVAS_NATIVE_SURFACE_OPENGL)
{
if (native)
{
n = calloc(1, sizeof(Native));
if (n)
{
memcpy(&(n->ns), ns, sizeof(Evas_Native_Surface));
eina_hash_add(re->win->gl_context->shared->native_tex_hash, &texid, im);
n->pixmap = 0;
n->visual = 0;
#ifdef GL_GLES
n->egl_surface = 0;
#else
n->fbc = 0;
n->glx_pixmap = 0;
#endif
im->native.yinvert = 0;
im->native.loose = 0;
im->native.data = n;
im->native.func.data = re;
im->native.func.bind = _native_bind_cb;
im->native.func.unbind = _native_unbind_cb;
im->native.func.free = _native_free_cb;
im->native.target = GL_TEXTURE_2D;
im->native.mipmap = 0;
// FIXME: need to implement mapping sub texture regions
// x, y, w, h for possible texture atlasing
evas_gl_common_image_native_enable(im);
}
}
}
return im;
}
static void *
eng_image_native_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *im = image;
Native *n;
if (!im) return NULL;
n = im->native.data;
if (!n) return NULL;
return &(n->ns);
}
#if 0 // filtering disabled
static void
eng_image_draw_filtered(void *data, void *context, void *surface,
void *image, Evas_Filter_Info *filter)
{
Render_Engine *re = data;
if (!image) return;
eng_window_use(re->win);
evas_gl_common_context_target_surface_set(re->win->gl_context, surface);
re->win->gl_context->dc = context;
evas_gl_common_filter_draw(re->win->gl_context, image, filter);
}
static Filtered_Image *
eng_image_filtered_get(void *im, uint8_t *key, size_t keylen)
{
return evas_gl_common_image_filtered_get(im, key, keylen);
}
static Filtered_Image *
eng_image_filtered_save(void *im, void *fim, uint8_t *key, size_t keylen)
{
return evas_gl_common_image_filtered_save(im, fim, key, keylen);
}
static void
eng_image_filtered_free(void *im, Filtered_Image *fim)
{
evas_gl_common_image_filtered_free(im, fim);
}
#endif
//
//
/////////////////////////////////////////////////////////////////////////
2002-11-08 00:02:15 -08:00
static void *
eng_image_load(void *data, const char *file, const char *key, int *error, Evas_Image_Load_Opts *lo)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
re = (Render_Engine *)data;
*error = EVAS_LOAD_ERROR_NONE;
eng_window_use(re->win);
return evas_gl_common_image_load(re->win->gl_context, file, key, lo, error);
2002-11-08 00:02:15 -08:00
}
static void *
eng_image_new_from_data(void *data, int w, int h, DATA32 *image_data, int alpha, int cspace)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
eng_window_use(re->win);
return evas_gl_common_image_new_from_data(re->win->gl_context, w, h, image_data, alpha, cspace);
2002-11-08 00:02:15 -08:00
}
static void *
eng_image_new_from_copied_data(void *data, int w, int h, DATA32 *image_data, int alpha, int cspace)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
eng_window_use(re->win);
return evas_gl_common_image_new_from_copied_data(re->win->gl_context, w, h, image_data, alpha, cspace);
2002-11-08 00:02:15 -08:00
}
static void
eng_image_free(void *data, void *image)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
re = (Render_Engine *)data;
if (!image) return;
eng_window_use(re->win);
evas_gl_common_image_free(image);
2002-11-08 00:02:15 -08:00
}
static void
eng_image_size_get(void *data EINA_UNUSED, void *image, int *w, int *h)
2002-11-08 00:02:15 -08:00
{
if (!image)
{
*w = 0;
*h = 0;
return;
}
if (w) *w = ((Evas_GL_Image *)image)->w;
if (h) *h = ((Evas_GL_Image *)image)->h;
2002-11-08 00:02:15 -08:00
}
static void *
eng_image_size_set(void *data, void *image, int w, int h)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
Evas_GL_Image *im = image;
Evas_GL_Image *im_old;
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
if (!im) return NULL;
if (im->native.data)
{
im->w = w;
im->h = h;
return image;
}
eng_window_use(re->win);
if ((im->tex) && (im->tex->pt->dyn.img))
{
evas_gl_common_texture_free(im->tex);
im->tex = NULL;
im->w = w;
im->h = h;
im->tex = evas_gl_common_texture_dynamic_new(im->gc, im);
return image;
}
im_old = image;
switch (eng_image_colorspace_get(data, image))
{
case EVAS_COLORSPACE_YCBCR422P601_PL:
case EVAS_COLORSPACE_YCBCR422P709_PL:
case EVAS_COLORSPACE_YCBCR422601_PL:
case EVAS_COLORSPACE_YCBCR420NV12601_PL:
case EVAS_COLORSPACE_YCBCR420TM12601_PL:
w &= ~0x1;
break;
}
if ((im_old->im) &&
((int)im_old->im->cache_entry.w == w) &&
((int)im_old->im->cache_entry.h == h))
return image;
if (im_old)
{
im = evas_gl_common_image_new(re->win->gl_context, w, h,
eng_image_alpha_get(data, image),
eng_image_colorspace_get(data, image));
/*
evas_common_load_image_data_from_file(im_old->im);
if (im_old->im->image->data)
{
evas_common_blit_rectangle(im_old->im, im->im, 0, 0, w, h, 0, 0);
evas_common_cpu_end_opt();
}
*/
evas_gl_common_image_free(im_old);
}
else
im = evas_gl_common_image_new(re->win->gl_context, w, h, 1, EVAS_COLORSPACE_ARGB8888);
return im;
2002-11-08 00:02:15 -08:00
}
static void *
eng_image_dirty_region(void *data, void *image, int x, int y, int w, int h)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
Evas_GL_Image *im = image;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
if (!image) return NULL;
if (im->native.data) return image;
eng_window_use(re->win);
evas_gl_common_image_dirty(image, x, y, w, h);
return image;
2002-11-08 00:02:15 -08:00
}
static void *
eng_image_data_get(void *data, void *image, int to_write, DATA32 **image_data, int *err)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
Evas_GL_Image *im;
int error;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
if (!image)
{
*image_data = NULL;
if (err) *err = EVAS_LOAD_ERROR_GENERIC;
return NULL;
}
im = image;
if (im->native.data)
{
*image_data = NULL;
if (err) *err = EVAS_LOAD_ERROR_NONE;
return im;
}
#ifdef GL_GLES
eng_window_use(re->win);
if ((im->tex) && (im->tex->pt) && (im->tex->pt->dyn.img) && (im->cs.space == EVAS_COLORSPACE_ARGB8888))
{
if (im->tex->pt->dyn.checked_out > 0)
{
im->tex->pt->dyn.checked_out++;
*image_data = im->tex->pt->dyn.data;
if (err) *err = EVAS_LOAD_ERROR_NONE;
return im;
}
*image_data = im->tex->pt->dyn.data = glsym_eglMapImageSEC(re->win->egl_disp, im->tex->pt->dyn.img);
if (!im->tex->pt->dyn.data)
{
if (err) *err = EVAS_LOAD_ERROR_RESOURCE_ALLOCATION_FAILED;
GLERR(__FUNCTION__, __FILE__, __LINE__, "");
return im;
}
im->tex->pt->dyn.checked_out++;
if (err) *err = EVAS_LOAD_ERROR_NONE;
return im;
}
#else
if ((im->tex) && (im->tex->pt) && (im->tex->pt->dyn.data))
{
*image_data = im->tex->pt->dyn.data;
if (err) *err = EVAS_LOAD_ERROR_NONE;
return im;
}
eng_window_use(re->win);
#endif
/* Engine can be fail to create texture after cache drop like eng_image_content_hint_set function,
so it is need to add code which check im->im's NULL value*/
if (!im->im)
{
*image_data = NULL;
if (err) *err = EVAS_LOAD_ERROR_RESOURCE_ALLOCATION_FAILED;
return NULL;
}
error = evas_cache_image_load_data(&im->im->cache_entry);
switch (im->cs.space)
{
case EVAS_COLORSPACE_ARGB8888:
if (to_write)
{
if (im->references > 1)
{
Evas_GL_Image *im_new;
im_new = evas_gl_common_image_new_from_copied_data
(im->gc, im->im->cache_entry.w, im->im->cache_entry.h,
im->im->image.data,
eng_image_alpha_get(data, image),
eng_image_colorspace_get(data, image));
if (!im_new)
{
*image_data = NULL;
if (err) *err = EVAS_LOAD_ERROR_RESOURCE_ALLOCATION_FAILED;
return NULL;
}
evas_gl_common_image_free(im);
im = im_new;
}
else
evas_gl_common_image_dirty(im, 0, 0, 0, 0);
}
*image_data = im->im->image.data;
break;
case EVAS_COLORSPACE_YCBCR422P601_PL:
case EVAS_COLORSPACE_YCBCR422P709_PL:
case EVAS_COLORSPACE_YCBCR422601_PL:
case EVAS_COLORSPACE_YCBCR420NV12601_PL:
case EVAS_COLORSPACE_YCBCR420TM12601_PL:
*image_data = im->cs.data;
break;
default:
abort();
break;
}
if (err) *err = error;
return im;
2002-11-08 00:02:15 -08:00
}
static void *
eng_image_data_put(void *data, void *image, DATA32 *image_data)
2002-11-08 00:02:15 -08:00
{
Render_Engine *re;
2006-12-17 08:46:30 -08:00
Evas_GL_Image *im, *im2;
2005-05-21 19:49:50 -07:00
2002-11-08 00:02:15 -08:00
re = (Render_Engine *)data;
if (!image) return NULL;
im = image;
if (im->native.data) return image;
eng_window_use(re->win);
if ((im->tex) && (im->tex->pt)
&& (im->tex->pt->dyn.data)
&& (im->cs.space == EVAS_COLORSPACE_ARGB8888))
{
if (im->tex->pt->dyn.data == image_data)
{
if (im->tex->pt->dyn.checked_out > 0)
{
im->tex->pt->dyn.checked_out--;
#ifdef GL_GLES
if (im->tex->pt->dyn.checked_out == 0)
glsym_eglUnmapImageSEC(re->win->egl_disp, im->tex->pt->dyn.img);
#endif
}
return image;
}
im2 = eng_image_new_from_data(data, im->w, im->h, image_data,
eng_image_alpha_get(data, image),
eng_image_colorspace_get(data, image));
if (!im2) return im;
evas_gl_common_image_free(im);
im = im2;
evas_gl_common_image_dirty(im, 0, 0, 0, 0);
return im;
}
switch (im->cs.space)
{
case EVAS_COLORSPACE_ARGB8888:
if ((!im->im) || (image_data != im->im->image.data))
{
im2 = eng_image_new_from_data(data, im->w, im->h, image_data,
eng_image_alpha_get(data, image),
eng_image_colorspace_get(data, image));
if (!im2) return im;
evas_gl_common_image_free(im);
im = im2;
}
break;
case EVAS_COLORSPACE_YCBCR422P601_PL:
case EVAS_COLORSPACE_YCBCR422P709_PL:
case EVAS_COLORSPACE_YCBCR422601_PL:
case EVAS_COLORSPACE_YCBCR420NV12601_PL:
case EVAS_COLORSPACE_YCBCR420TM12601_PL:
if (image_data != im->cs.data)
{
if (im->cs.data)
{
if (!im->cs.no_free) free(im->cs.data);
}
im->cs.data = image_data;
}
evas_gl_common_image_dirty(im, 0, 0, 0, 0);
break;
default:
abort();
break;
}
return im;
2002-11-08 00:02:15 -08:00
}
static void
eng_image_data_preload_request(void *data EINA_UNUSED, void *image, const void *target)
{
Evas_GL_Image *gim = image;
RGBA_Image *im;
if (!gim) return;
if (gim->native.data) return;
im = (RGBA_Image *)gim->im;
if (!im) return;
evas_cache_image_preload_data(&im->cache_entry, target);
}
static void
eng_image_data_preload_cancel(void *data EINA_UNUSED, void *image, const void *target)
{
Evas_GL_Image *gim = image;
RGBA_Image *im;
if (!gim) return;
if (gim->native.data) return;
im = (RGBA_Image *)gim->im;
if (!im) return;
evas_cache_image_preload_cancel(&im->cache_entry, target);
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
static void
eng_image_draw(void *data, void *context, void *surface, void *image, int src_x, int src_y, int src_w, int src_h, int dst_x, int dst_y, int dst_w, int dst_h, int smooth)
{
Render_Engine *re;
re = (Render_Engine *)data;
Evas_GL_Image *im = image;
Native *n;
if (!im) return;
n = im->native.data;
if ((n) && (n->ns.type == EVAS_NATIVE_SURFACE_OPENGL) && (evgl_direct_rendered(re->evgl_engine)))
{
DBG("Rendering Directly to the window");
evas_object_image_pixels_dirty_set(evgl_direct_img_obj_get(re->evgl_engine), EINA_TRUE);
}
else
{
eng_window_use(re->win);
evas_gl_common_context_target_surface_set(re->win->gl_context, surface);
re->win->gl_context->dc = context;
evas_gl_common_image_draw(re->win->gl_context, image,
src_x, src_y, src_w, src_h,
dst_x, dst_y, dst_w, dst_h,
smooth);
}
}
static void
eng_image_scale_hint_set(void *data EINA_UNUSED, void *image, int hint)
{
if (image) evas_gl_common_image_scale_hint_set(image, hint);
}
static int
eng_image_scale_hint_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *gim = image;
if (!gim) return EVAS_IMAGE_SCALE_HINT_NONE;
return gim->scale_hint;
}
static void
eng_image_map_draw(void *data, void *context, void *surface, void *image, RGBA_Map *m, int smooth, int level)
{
Evas_GL_Image *gim = image;
Render_Engine *re;
re = (Render_Engine *)data;
if (!image) return;
eng_window_use(re->win);
evas_gl_common_context_target_surface_set(re->win->gl_context, surface);
re->win->gl_context->dc = context;
if (m->count != 4)
{
// FIXME: nash - you didn't fix this
abort();
}
if ((m->pts[0].x == m->pts[3].x) &&
(m->pts[1].x == m->pts[2].x) &&
(m->pts[0].y == m->pts[1].y) &&
(m->pts[3].y == m->pts[2].y) &&
(m->pts[0].x <= m->pts[1].x) &&
(m->pts[0].y <= m->pts[2].y) &&
(m->pts[0].u == 0) &&
(m->pts[0].v == 0) &&
(m->pts[1].u == (gim->w << FP)) &&
(m->pts[1].v == 0) &&
(m->pts[2].u == (gim->w << FP)) &&
(m->pts[2].v == (gim->h << FP)) &&
(m->pts[3].u == 0) &&
(m->pts[3].v == (gim->h << FP)) &&
(m->pts[0].col == 0xffffffff) &&
(m->pts[1].col == 0xffffffff) &&
(m->pts[2].col == 0xffffffff) &&
(m->pts[3].col == 0xffffffff))
{
int dx, dy, dw, dh;
dx = m->pts[0].x >> FP;
dy = m->pts[0].y >> FP;
dw = (m->pts[2].x >> FP) - dx;
dh = (m->pts[2].y >> FP) - dy;
eng_image_draw(data, context, surface, image,
0, 0, gim->w, gim->h, dx, dy, dw, dh, smooth);
}
else
{
evas_gl_common_image_map_draw(re->win->gl_context, image, m->count, &m->pts[0],
smooth, level);
}
}
static void
eng_image_map_clean(void *data EINA_UNUSED, RGBA_Map *m EINA_UNUSED)
{
}
static void *
eng_image_map_surface_new(void *data, int w, int h, int alpha)
{
Render_Engine *re;
re = (Render_Engine *)data;
return evas_gl_common_image_surface_new(re->win->gl_context, w, h, alpha);
}
static void
eng_image_map_surface_free(void *data EINA_UNUSED, void *surface)
{
evas_gl_common_image_free(surface);
}
static void
eng_image_content_hint_set(void *data EINA_UNUSED, void *image, int hint)
{
if (image) evas_gl_common_image_content_hint_set(image, hint);
}
static int
eng_image_content_hint_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *gim = image;
if (!gim) return EVAS_IMAGE_CONTENT_HINT_NONE;
return gim->content_hint;
}
static void
eng_image_cache_flush(void *data)
{
Render_Engine *re;
int tmp_size;
re = (Render_Engine *)data;
tmp_size = evas_common_image_get_cache();
evas_common_image_set_cache(0);
evas_common_rgba_image_scalecache_flush();
evas_gl_common_image_cache_flush(re->win->gl_context);
evas_common_image_set_cache(tmp_size);
}
static void
eng_image_cache_set(void *data, int bytes)
{
Render_Engine *re;
re = (Render_Engine *)data;
evas_common_image_set_cache(bytes);
evas_common_rgba_image_scalecache_size_set(bytes);
evas_gl_common_image_cache_flush(re->win->gl_context);
}
static int
eng_image_cache_get(void *data EINA_UNUSED)
{
return evas_common_image_get_cache();
}
static void
eng_image_stride_get(void *data EINA_UNUSED, void *image, int *stride)
{
Evas_GL_Image *im = image;
if ((im->tex) && (im->tex->pt->dyn.img))
*stride = im->tex->pt->dyn.stride;
else
*stride = im->w * 4;
}
static void
eng_font_draw(void *data, void *context, void *surface, Evas_Font_Set *font EINA_UNUSED, int x, int y, int w EINA_UNUSED, int h EINA_UNUSED, int ow EINA_UNUSED, int oh EINA_UNUSED, Evas_Text_Props *intl_props)
{
Render_Engine *re;
re = (Render_Engine *)data;
eng_window_use(re->win);
evas_gl_common_context_target_surface_set(re->win->gl_context, surface);
re->win->gl_context->dc = context;
{
// FIXME: put im into context so we can free it
static RGBA_Image *im = NULL;
if (!im)
im = (RGBA_Image *)evas_cache_image_empty(evas_common_image_cache_get());
im->cache_entry.w = re->win->gl_context->shared->w;
im->cache_entry.h = re->win->gl_context->shared->h;
evas_common_draw_context_font_ext_set(context,
re->win->gl_context,
evas_gl_font_texture_new,
evas_gl_font_texture_free,
evas_gl_font_texture_draw);
evas_common_font_draw_prepare(intl_props);
evas_common_font_draw(im, context, x, y, intl_props);
evas_common_draw_context_font_ext_set(context,
NULL,
NULL,
NULL,
NULL);
}
}
static Eina_Bool
eng_canvas_alpha_get(void *data, void *info EINA_UNUSED)
{
Render_Engine *re = (Render_Engine *)data;
return re->win->alpha;
}
//--------------------------------//
// Evas GL Related Code
static void *
eng_gl_surface_create(void *data, void *config, int w, int h)
{
Render_Engine *re = (Render_Engine *)data;
Evas_GL_Config *cfg = (Evas_GL_Config *)config;
return evgl_surface_create(re->evgl_engine, cfg, w, h);
}
static int
eng_gl_surface_destroy(void *data, void *surface)
{
Render_Engine *re = (Render_Engine *)data;
EVGL_Surface *sfc = (EVGL_Surface *)surface;
return evgl_surface_destroy(re->evgl_engine, sfc);
}
static void *
eng_gl_context_create(void *data, void *share_context)
{
Render_Engine *re = (Render_Engine *)data;
EVGL_Context *sctx = (EVGL_Context *)share_context;
return evgl_context_create(re->evgl_engine, sctx);
}
static int
eng_gl_context_destroy(void *data, void *context)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
Render_Engine *re = (Render_Engine *)data;
EVGL_Context *ctx = (EVGL_Context *)context;
return evgl_context_destroy(re->evgl_engine, ctx);
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
static int
eng_gl_make_current(void *data EINA_UNUSED, void *surface, void *context)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
Render_Engine *re = (Render_Engine *)data;
EVGL_Surface *sfc = (EVGL_Surface *)surface;
EVGL_Context *ctx = (EVGL_Context *)context;
return evgl_make_current(re->evgl_engine, sfc, ctx);
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
static void *
eng_gl_string_query(void *data EINA_UNUSED, int name)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
Render_Engine *re = (Render_Engine *)data;
return (void *)evgl_string_query(re->evgl_engine, name);
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
// Need to deprecate this function..
static void *
eng_gl_proc_address_get(void *data EINA_UNUSED, const char *name EINA_UNUSED)
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
{
return NULL;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
}
static int
eng_gl_native_surface_get(void *data EINA_UNUSED, void *surface, void *native_surface)
{
Render_Engine *re = (Render_Engine *)data;
EVGL_Surface *sfc = (EVGL_Surface *)surface;
Evas_Native_Surface *ns = (Evas_Native_Surface *)native_surface;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
return evgl_native_surface_get(re->evgl_engine, sfc, ns);
}
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
static void *
eng_gl_api_get(void *data EINA_UNUSED)
{
Render_Engine *re = (Render_Engine *)data;
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
return evgl_api_get(re->evgl_engine);
}
static void
eng_gl_img_obj_set(void *data, void *image, int has_alpha)
{
Render_Engine *re = (Render_Engine *)data;
// Normally direct rendering isn't allowed if alpha is on and
// rotation is not 0. BUT, if override is on, allow it.
if ((has_alpha) || (re->win->gl_context->rot!=0))
{
if (re->evgl_engine->direct_override)
evgl_direct_img_obj_set(re->evgl_engine, image);
else
evgl_direct_img_obj_set(re->evgl_engine, NULL);
}
else
evgl_direct_img_obj_set(re->evgl_engine, image);
}
//--------------------------------//
static int
eng_image_load_error_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *im;
if (!image) return EVAS_LOAD_ERROR_NONE;
im = image;
return im->im->cache_entry.load_error;
}
static Eina_Bool
eng_image_animated_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *gim = image;
Image_Entry *im;
if (!gim) return EINA_FALSE;
im = (Image_Entry *)gim->im;
if (!im) return EINA_FALSE;
return im->flags.animated;
}
static int
eng_image_animated_frame_count_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *gim = image;
Image_Entry *im;
if (!gim) return -1;
im = (Image_Entry *)gim->im;
if (!im) return -1;
if (!im->flags.animated) return -1;
return im->frame_count;
}
static Evas_Image_Animated_Loop_Hint
eng_image_animated_loop_type_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *gim = image;
Image_Entry *im;
if (!gim) return EVAS_IMAGE_ANIMATED_HINT_NONE;
im = (Image_Entry *)gim->im;
if (!im) return EVAS_IMAGE_ANIMATED_HINT_NONE;
if (!im->flags.animated) return EVAS_IMAGE_ANIMATED_HINT_NONE;
return im->loop_hint;
}
static int
eng_image_animated_loop_count_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *gim = image;
Image_Entry *im;
if (!gim) return -1;
im = (Image_Entry *)gim->im;
if (!im) return -1;
if (!im->flags.animated) return -1;
return im->loop_count;
}
static double
eng_image_animated_frame_duration_get(void *data EINA_UNUSED, void *image, int start_frame, int frame_num)
{
Evas_GL_Image *gim = image;
Image_Entry *im;
if (!gim) return -1;
im = (Image_Entry *)gim->im;
if (!im) return -1;
if (!im->flags.animated) return -1;
return evas_common_load_rgba_image_frame_duration_from_file(im, start_frame, frame_num);
}
static Eina_Bool
eng_image_animated_frame_set(void *data EINA_UNUSED, void *image, int frame_index)
{
Evas_GL_Image *gim = image;
Image_Entry *im;
if (!gim) return EINA_FALSE;
im = (Image_Entry *)gim->im;
if (!im) return EINA_FALSE;
if (!im->flags.animated) return EINA_FALSE;
if (im->cur_frame == frame_index) return EINA_FALSE;
im->cur_frame = frame_index;
return EINA_TRUE;
}
static Eina_Bool
eng_image_can_region_get(void *data EINA_UNUSED, void *image)
{
Evas_GL_Image *gim = image;
Image_Entry *im;
if (!gim) return EINA_FALSE;
im = (Image_Entry *)gim->im;
if (!im) return EINA_FALSE;
return ((Evas_Image_Load_Func*) im->info.loader)->do_region;
}
static void
eng_image_max_size_get(void *data, int *maxw, int *maxh)
{
Render_Engine *re = (Render_Engine *)data;
if (maxw) *maxw = re->win->gl_context->shared->info.max_texture_size;
if (maxh) *maxh = re->win->gl_context->shared->info.max_texture_size;
}
static int
module_open(Evas_Module *em)
{
static Eina_Bool xrm_inited = EINA_FALSE;
if (!xrm_inited)
{
xrm_inited = EINA_TRUE;
XrmInitialize();
}
if (!em) return 0;
if (!evas_gl_common_module_open()) return 0;
/* get whatever engine module we inherit from */
if (!_evas_module_engine_inherit(&pfunc, "software_generic")) return 0;
if (_evas_engine_GL_X11_log_dom < 0)
_evas_engine_GL_X11_log_dom = eina_log_domain_register
("evas-gl_x11", EVAS_DEFAULT_LOG_COLOR);
if (_evas_engine_GL_X11_log_dom < 0)
{
EINA_LOG_ERR("Can not create a module log domain.");
return 0;
}
/* store it for later use */
func = pfunc;
/* now to override methods */
#define ORD(f) EVAS_API_OVERRIDE(f, &func, eng_)
ORD(info);
ORD(info_free);
ORD(setup);
ORD(canvas_alpha_get);
ORD(output_free);
ORD(output_resize);
ORD(output_tile_size_set);
ORD(output_redraws_rect_add);
ORD(output_redraws_rect_del);
ORD(output_redraws_clear);
ORD(output_redraws_next_update_get);
ORD(output_redraws_next_update_push);
ORD(context_cutout_add);
ORD(context_cutout_clear);
ORD(output_flush);
ORD(output_idle_flush);
ORD(output_dump);
ORD(rectangle_draw);
ORD(line_draw);
ORD(polygon_point_add);
ORD(polygon_points_clear);
ORD(polygon_draw);
ORD(image_load);
ORD(image_new_from_data);
ORD(image_new_from_copied_data);
ORD(image_free);
ORD(image_size_get);
ORD(image_size_set);
ORD(image_dirty_region);
ORD(image_data_get);
ORD(image_data_put);
ORD(image_data_preload_request);
ORD(image_data_preload_cancel);
ORD(image_alpha_set);
ORD(image_alpha_get);
ORD(image_border_set);
ORD(image_border_get);
ORD(image_draw);
ORD(image_comment_get);
ORD(image_format_get);
ORD(image_colorspace_set);
ORD(image_colorspace_get);
ORD(image_can_region_get);
ORD(image_mask_create);
ORD(image_native_set);
ORD(image_native_get);
#if 0 // filtering disabled
ORD(image_draw_filtered);
ORD(image_filtered_get);
ORD(image_filtered_save);
ORD(image_filtered_free);
#endif
ORD(font_draw);
ORD(image_scale_hint_set);
ORD(image_scale_hint_get);
ORD(image_stride_get);
ORD(image_map_draw);
ORD(image_map_surface_new);
ORD(image_map_surface_free);
ORD(image_map_clean);
ORD(image_content_hint_set);
ORD(image_content_hint_get);
ORD(image_cache_flush);
ORD(image_cache_set);
ORD(image_cache_get);
ORD(gl_surface_create);
ORD(gl_surface_destroy);
ORD(gl_context_create);
ORD(gl_context_destroy);
ORD(gl_make_current);
From: "Sung W. Park" <sungwoo@gmail.com> Subject: Re: [E-devel] [E-Devel][Review][Patch] Evas GL Extensions + a bug fix Here's an initial attempt at the GL extensions issue for Evas GL. I have been in discussion with a few EFL developers regarding how we should provide extensions. Essentially, there are two ways to go about doing this. 1. provide evas_gl_proc_address_get() function as it is done in other glue layers 2. provide all the extension functions in the EVAS_GL_API struct. #1 approach is how it's done in other glue layers and the driver implementor can provide new extensions easily. It is however pretty annoying to get the function prototypes right and use the function pointers and etc. #2 approach provides all the extension functions in the struct so it's definitely easier to use. Adding new extensions can be a pain as people may have to wait for new version releases. For now, we thought it was OK to just throw them in the struct as in #2 and try it out. So, I've implemented this approach. As for the extensions, I've basically included all the extensions in gl2ext.h as EvasGL currently provides all the GLES 2.0 functions. In order to display the right information, I had to override glGetString() with GL_EXTENSIONS as parameter to properly display the supported extensions. Also, I've added a few EGL extensions that have been modified for EvasGL use. For example, eglCreateImage/eglDestroyImage has been defined as folllows. EvasGLImage (*evasglCreateImage) (int target, void* buffer, int* attrib_list); void (*evasglDestroyImage) (EvasGLImage image); const char *evas_gl_string_query() function was added to return a string of supported EvasGL extensions. So essentially, a user can search this string to see if the desired extension is supported. if it is, he can use the functions. He can always check if the function pointers are NULL as well. Take a look at the pach and let me know what you think. ______________ While I was adding the extension code, I've added a few fixes/ changes to the EvasGL. 1. glDeletBuffers bug - When I wad destroying evasgl context, I was deleting the context FBO with glDeleteBuffers instead of glDeleteFramebuffers. This code in effect was deleting BOs in other contexts and we had some funky behaviors as a result. The bug has been fixed. 2. make_current - I've made some changes to the make current code and also added a resource context to the engine data. the resource context is used for creating surface texture/ fbos when surface/ context is created. Before, i was using evas' context but thought it'd be a good idea to use a separate context. SVN revision: 64139
2011-10-18 01:13:23 -07:00
ORD(gl_string_query);
ORD(gl_proc_address_get);
ORD(gl_native_surface_get);
ORD(gl_api_get);
ORD(gl_img_obj_set);
ORD(image_load_error_get);
/* now advertise out own api */
ORD(image_animated_get);
ORD(image_animated_frame_count_get);
ORD(image_animated_loop_type_get);
ORD(image_animated_loop_count_get);
ORD(image_animated_frame_duration_get);
ORD(image_animated_frame_set);
ORD(image_max_size_get);
/* now advertise out own api */
em->functions = (void *)(&func);
return 1;
}
static void
module_close(Evas_Module *em EINA_UNUSED)
{
eina_log_domain_unregister(_evas_engine_GL_X11_log_dom);
evas_gl_common_module_close();
}
static Evas_Module_Api evas_modapi =
{
EVAS_MODULE_API_VERSION,
"gl_x11",
"none",
{
module_open,
module_close
}
};
EVAS_MODULE_DEFINE(EVAS_MODULE_TYPE_ENGINE, engine, gl_x11);
#ifndef EVAS_STATIC_BUILD_GL_XLIB
EVAS_EINA_MODULE_DEFINE(engine, gl_x11);
#endif
/* vim:set ts=8 sw=3 sts=3 expandtab cino=>5n-2f0^-2{2(0W1st0 :*/