Followers 0

# OpenGL glTexSubImage2D() performance

## 2 posts in this topic

I'm trying to implement a 2D GUI system for my OpenGL test program. I have the following definitions:

#define E_RMASK 0x00ff0000#define E_RSHIFT 16#define E_GMASK 0x0000ff00#define E_GSHIFT 8#define E_BMASK 0x000000ff#define E_BSHIFT 0#define E_AMASK 0xff000000#define E_ASHIFT 24#define E_CFMT GL_BGRAtypedef struct {	int32_t  x;	int32_t  y;	int32_t  w;	int32_t  h;} Erect_t;typedef struct {	int32_t     w;	int32_t     h;	uint8_t     bpp;	uint32_t    fmt;	void       *d;	/* private members */	bool         dirty;	Erect_t      clip;} Eimage_t;typedef struct {	Eimage_t *img;	GLuint    id;} Etexture_t;static inline uint32_t pixel_blend( uint32_t dst, uint32_t src ){	uint32_t as = ( src & E_AMASK ) >> E_ASHIFT;	uint32_t ad = ( dst & E_AMASK ) >> E_ASHIFT;	uint32_t a = min( as + ad, 0x000000ff );	if ( !as ) return dst;#define RGB_MASK ( ~E_AMASK )#define RB_MASK ( E_RMASK | E_BMASK )#define AG_MASK ( E_AMASK | E_GMASK )	uint32_t rb = ( ( ( src & RB_MASK ) * as ) + ( ( dst & RB_MASK ) * ( 0xff - as ) ) ) & AG_MASK;	uint32_t  g = ( ( ( src & E_GMASK ) * as ) + ( ( dst & E_GMASK ) * ( 0xff - as ) ) ) & E_RMASK;#undef RB_MASK#undef AG_MASK	return ( ( a << E_ASHIFT ) & E_AMASK ) | ( ( ( rb | g ) >> 8 ) & RGB_MASK );}uint32_t engineMapColor( uint8_t r, uint8_t g, uint8_t b, uint8_t a ){	return ( ( ( uint32_t )r & 0x000000ff ) << E_RSHIFT ) +	       ( ( ( uint32_t )g & 0x000000ff ) << E_GSHIFT ) +	       ( ( ( uint32_t )b & 0x000000ff ) << E_BSHIFT ) +	       ( ( ( uint32_t )a & 0x000000ff ) << E_ASHIFT );}Eimage_t *engineImageCreate( uint32_t w, uint32_t h ){	Eimage_t *ret = ( Eimage_t * )calloc( 1, sizeof( Eimage_t ) );	ret->w = w;	ret->h = h;	ret->bpp = 4;	ret->fmt = engineState()->Cfmt;	ret->d = calloc( 4, w * h );	img->clip.x = 0;	img->clip.y = 0;	img->clip.w = img->w;	img->clip.h = img->h;	return ret;}void engineImageFillRect( Eimage_t *img, const Erect_t *r, uint32_t color ){	int32_t start_x = min( ( int32_t )img->w, ( r ? max( r->x, img->clip.x ) : 0 ) );	int32_t start_y = min( ( int32_t )img->h, ( r ? max( r->y, img->clip.y ) : 0 ) );	int32_t end_x = ( r ? min( ( int32_t )img->clip.w, r->x + r->w ) : ( int32_t )img->clip.w );	int32_t end_y = ( r ? min( ( int32_t )img->clip.h, r->y + r->h ) : ( int32_t )img->clip.h );	uint32_t *data = ( uint32_t * )img->d + start_y * img->w;	for ( int32_t j = start_y; j < end_y; j++ ) {		int32_t i = end_x - start_x;		int32_t c = ( i + 7 ) / 8;		switch ( i % 8 ) {			case 0: do { { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; }			case 7: { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; }			case 6: { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; }			case 5: { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; }			case 4: { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; }			case 3: { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; }			case 2: { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; }			case 1: { data[start_x + i] = pixel_blend( data[start_x + i], color ); i -= 1; } } while ( --c > 0 );		}		data += img->w;	}	img->dirty = true;}Etexture_t *engineTextureCreate( uint32_t w, uint32_t h ){	Etexture_t *ret = ( Etexture_t * )calloc( 1, sizeof( Etexture_t ) );	ret->img = engineImageCreate( w, h  );	glGenTextures( 1, &ret->id );	glBindTexture( GL_TEXTURE_2D, ret->id  );	glTexImage2D( GL_TEXTURE_2D, 0, ret->img->bpp, ret->img->w, ret->img->h, 0, ret->img->fmt, GL_UNSIGNED_BYTE, ret->img->d );	glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR );	glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR );	return ret;}void engineTextureBind( Etexture_t *tex ){	if ( tex ) {		glBindTexture( GL_TEXTURE_2D, tex->id );		if ( tex->img->dirty ) {			glTexSubImage2D( GL_TEXTURE_2D, 0, 0, 0, tex->img->w, tex->img->h, tex->img->fmt, GL_UNSIGNED_BYTE, tex->img->d );			tex->img->dirty = false;		}	}}

Then I have (Estate_t is a structure holding various flags, the width/height of the window, etc. I'm not pasting it here since it's pretty big):
static void begin_2d_overlay( Estate_t *st ){	glDisable( GL_DEPTH_TEST );	glDisable( GL_LIGHTING );	glBlendFunc( GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA );	glEnable( GL_BLEND );	if ( st->wireframe )		glPolygonMode( GL_FRONT_AND_BACK, GL_FILL );	glMatrixMode( GL_PROJECTION );	glPushMatrix();	glLoadIdentity();	glOrtho( 0.0, st->wm_width, st->wm_height, 0.0, -1, 1 );	glMatrixMode( GL_MODELVIEW );	glPushMatrix();	glLoadIdentity();	glColor3f( 1.0, 1.0, 1.0 );}static void end_2d_overlay( Estate_t *st ){	glEnable( GL_DEPTH_TEST );	glEnable( GL_LIGHTING );	glDisable( GL_BLEND );	glMatrixMode( GL_PROJECTION );	glPopMatrix();	glMatrixMode( GL_MODELVIEW );	glPopMatrix();	if ( st->wireframe )		glPolygonMode( GL_FRONT_AND_BACK, GL_LINE );}

and my rendering function:

Etexture_t *overlay = engineTextureCreate( st->wm_width, st->wm_height );static void render_function( Estate_t *st ){	/* draw the 3D scene here */	begin_2d_overlay( st );	Erect_t r;	r.x = rand() % st->wm_width;	r.y = rand() % st->wm_height;	r.w = rand() % 300 + 1;	r.h = rand() % 300 + 1;	engineImageFillRect( overlay->img, &r, engineMapColor( rand() % 255, rand() % 255, rand() % 255, 128 ) );		overlay->img->dirty = false;  /* this shouldn't be here, see below code block for explanation */	engineTextureBind( overlay );	glBegin( GL_QUADS );		glTexCoord2f( 0.0, 1.0 ); glVertex2f( 0.0, st->wm_height );		glTexCoord2f( 1.0, 1.0 ); glVertex2f( st->wm_width, st->wm_height );		glTexCoord2f( 1.0, 0.0 ); glVertex2f( st->wm_width, 0.0 );		glTexCoord2f( 0.0, 0.0 ); glVertex2f( 0.0, 0.0 );	glEnd();	end_2d_overlay( st );	/* more stuff like swapping OpenGL buffers, etc. */}

Notice the "overlay->img->dirty = false;" line. When uncommented it causes engineTextureBind() to not call glTexSubImage2D() which in turn causes my application to jump from rendering ~360 frames per second to rendering ~3600 frames per second (yes, 10 times more). I only added that line to see if the slowdown is caused by my rectangle drawing code or something else. After multiple tries I managed to isolate glTexSubImage() as the cause of the slow-down.

My questions are:
(1) is this anywhere near the proper way to do a 2D GUI with OpenGL ?
(2) am I doing something wrong that causes glTexSubImage() to run that slow ?
(3) should I be using something else than glTexSubImage() to update the GUI ?

[Edited by - ileonte on November 21, 2010 4:29:57 AM]
0

##### Share on other sites
The difference between 360fps and 3600fps is only 2ms render time. That is not much. Don't use fps to measure performance, use seconds per frame instead.
1) No, you should do all the rendering in OpenGL. Currently you are doing a lot of computation on the CPU and then sending huge amount of pixel data to the GPU. I don't know at what resolution you are working but the situation will become worth on higher resolutions. At some point even the CPU might become the bottleneck.
2) I have not checked the code in details but using the format the GPU uses internally can improve performance. PBOs are also a good choice.
Fast Texture Transfers
3) Yes, render everything using OpenGL or use a library that does this, like Qt.
0

##### Share on other sites
The optimal glTex(Sub)Image2D parameters are:

internalformal: GL_RGBA
format: GL_BGRA
type: GL_UNSIGNED_INT_8_8_8_8_REV

Use these (and set up your source data to be happy with them) and you'll get good performance on all hardware with no stalls, no copying back to system memory to do the update, and likely no need to use a PBO (which may give you a lot of work for zero performance gain otherwise).

The reason why is that OpenGL stores your texture using 4 components irrespective of what you specify in your glTex(Sub)Image2D call - check the spec if you doubt me. Using any combination of parameters that doesn't match the real internal storage will mean that your driver will need to copy, expand, swizzle, fill in unspecified components and all the rest itself, and - in especially evil cases (Intel) - copy the texture data back to system memory.

People often choose suboptimal formats and internalformats such as GL_RGB (or GL_BGR) because they think it "saves memory" - it doesn't. OpenGL will always expand textures to 4 components, so not only does it not "save memory" but it also slows down your program.

A format of GL_BGRA (instead of GL_RGBA) can give up to a 6x increase in transfer speed on even recent NVIDIA hardware.

GL_UNSIGNED_INT_8_8_8_8_REV instead of GL_UNSIGNED_BYTE is optional but I recommend it as it can give a hint to some drivers that you're transferring 4-component texels and therefore put them into a more optimal special case path (Intel again).

http://www.opengl.org/wiki/Common_Mistakes#Paletted_textures (the last sentence)

The best code structure for efficient use of glTexSubImage2D looks like this:
Identify what needs to be updated.Update textures.Draw the scene.
The worst looks like this:
Update subrect.Draw stuff.Update another subrect.Draw stuff.Update another subrect.Draw stuff.Etc.
The latter will result in multiple pipeline stalls per frame as OpenGL will need to wait until you're finished drawing with the texture before it can update each next subrect.

Using these techniques I've had reports of an over 90x overall performance improvement from one user - not to be sniffed at.

[Edited by - mhagain on November 21, 2010 9:07:54 AM]
1

## Create an account

Register a new account

Followers 0

• ### Similar Content

• Hello, I have been working on SH Irradiance map rendering, and I have been using a GLSL pixel shader to render SH irradiance to 2D irradiance maps for my static objects. I already have it working with 9 3D textures so far for the first 9 SH functions.
In my GLSL shader, I have to send in 9 SH Coefficient 3D Texures that use RGBA8 as a pixel format. RGB being used for the coefficients for red, green, and blue, and the A for checking if the voxel is in use (for the 3D texture solidification shader to prevent bleeding).
My problem is, I want to knock this number of textures down to something like 4 or 5. Getting even lower would be a godsend. This is because I eventually plan on adding more SH Coefficient 3D Textures for other parts of the game map (such as inside rooms, as opposed to the outside), to circumvent irradiance probe bleeding between rooms separated by walls. I don't want to reach the 32 texture limit too soon. Also, I figure that it would be a LOT faster.
Is there a way I could, say, store 2 sets of SH Coefficients for 2 SH functions inside a texture with RGBA16 pixels? If so, how would I extract them from inside GLSL? Let me know if you have any suggestions ^^.
• By KarimIO
EDIT: I thought this was restricted to Attribute-Created GL contexts, but it isn't, so I rewrote the post.
Hey guys, whenever I call SwapBuffers(hDC), I get a crash, and I get a "Too many posts were made to a semaphore." from Windows as I call SwapBuffers. What could be the cause of this?
Update: No crash occurs if I don't draw, just clear and swap.
static PIXELFORMATDESCRIPTOR pfd = // pfd Tells Windows How We Want Things To Be { sizeof(PIXELFORMATDESCRIPTOR), // Size Of This Pixel Format Descriptor 1, // Version Number PFD_DRAW_TO_WINDOW | // Format Must Support Window PFD_SUPPORT_OPENGL | // Format Must Support OpenGL PFD_DOUBLEBUFFER, // Must Support Double Buffering PFD_TYPE_RGBA, // Request An RGBA Format 32, // Select Our Color Depth 0, 0, 0, 0, 0, 0, // Color Bits Ignored 0, // No Alpha Buffer 0, // Shift Bit Ignored 0, // No Accumulation Buffer 0, 0, 0, 0, // Accumulation Bits Ignored 24, // 24Bit Z-Buffer (Depth Buffer) 0, // No Stencil Buffer 0, // No Auxiliary Buffer PFD_MAIN_PLANE, // Main Drawing Layer 0, // Reserved 0, 0, 0 // Layer Masks Ignored }; if (!(hDC = GetDC(windowHandle))) return false; unsigned int PixelFormat; if (!(PixelFormat = ChoosePixelFormat(hDC, &pfd))) return false; if (!SetPixelFormat(hDC, PixelFormat, &pfd)) return false; hRC = wglCreateContext(hDC); if (!hRC) { std::cout << "wglCreateContext Failed!\n"; return false; } if (wglMakeCurrent(hDC, hRC) == NULL) { std::cout << "Make Context Current Second Failed!\n"; return false; } ... // OGL Buffer Initialization glClear(GL_DEPTH_BUFFER_BIT | GL_COLOR_BUFFER_BIT); glBindVertexArray(vao); glUseProgram(myprogram); glDrawElements(GL_TRIANGLES, indexCount, GL_UNSIGNED_SHORT, (void *)indexStart); SwapBuffers(GetDC(window_handle));
• By Tchom
Hey devs!

I've been working on a OpenGL ES 2.0 android engine and I have begun implementing some simple (point) lighting. I had something fairly simple working, so I tried to get fancy and added color-tinting light. And it works great... with only one or two lights. Any more than that, the application drops about 15 frames per light added (my ideal is at least 4 or 5). I know implementing lighting is expensive, I just didn't think it was that expensive. I'm fairly new to the world of OpenGL and GLSL, so there is a good chance I've written some crappy shader code. If anyone had any feedback or tips on how I can optimize this code, please let me know.

uniform mat4 u_MVPMatrix; uniform mat4 u_MVMatrix; attribute vec4 a_Position; attribute vec3 a_Normal; attribute vec2 a_TexCoordinate; varying vec3 v_Position; varying vec3 v_Normal; varying vec2 v_TexCoordinate; void main() { v_Position = vec3(u_MVMatrix * a_Position); v_TexCoordinate = a_TexCoordinate; v_Normal = vec3(u_MVMatrix * vec4(a_Normal, 0.0)); gl_Position = u_MVPMatrix * a_Position; } Fragment Shader
precision mediump float; uniform vec4 u_LightPos["+numLights+"]; uniform vec4 u_LightColours["+numLights+"]; uniform float u_LightPower["+numLights+"]; uniform sampler2D u_Texture; varying vec3 v_Position; varying vec3 v_Normal; varying vec2 v_TexCoordinate; void main() { gl_FragColor = (texture2D(u_Texture, v_TexCoordinate)); float diffuse = 0.0; vec4 colourSum = vec4(1.0); for (int i = 0; i < "+numLights+"; i++) { vec3 toPointLight = vec3(u_LightPos[i]); float distance = length(toPointLight - v_Position); vec3 lightVector = normalize(toPointLight - v_Position); float diffuseDiff = 0.0; // The diffuse difference contributed from current light diffuseDiff = max(dot(v_Normal, lightVector), 0.0); diffuseDiff = diffuseDiff * (1.0 / (1.0 + ((1.0-u_LightPower[i])* distance * distance))); //Determine attenuatio diffuse += diffuseDiff; gl_FragColor.rgb *= vec3(1.0) / ((vec3(1.0) + ((vec3(1.0) - vec3(u_LightColours[i]))*diffuseDiff))); //The expensive part } diffuse += 0.1; //Add ambient light gl_FragColor.rgb *= diffuse; } Am I making any rookie mistakes? Or am I just being unrealistic about what I can do? Thanks in advance
• By yahiko00
Hi,
Not sure to post at the right place, if not, please forgive me...
For a game project I am working on, I would like to implement a 2D starfield as a background.
I do not want to deal with static tiles, since I plan to slowly animate the starfield. So, I am trying to figure out how to generate a random starfield for the entire map.
I feel that using a uniform distribution for the stars will not do the trick. Instead I would like something similar to the screenshot below, taken from the game Star Wars: Empire At War (all credits to Lucasfilm, Disney, and so on...).

Is there someone who could have an idea of a distribution which could result in such a starfield?
Any insight would be appreciated

• I have just noticed that, in quake 3 and half - life, dynamic models are effected from light map. For example in dark areas, gun that player holds seems darker. How did they achieve this effect ? I can use image based lighting techniques however (Like placing an environment probe and using it for reflections and ambient lighting), this tech wasn't used in games back then, so there must be a simpler method to do this.
Here is a link that shows how modern engines does it. Indirect Lighting Cache It would be nice if you know a paper that explains this technique. Can I apply this to quake 3' s light map generator and bsp format ?

• 9
• 17
• 10
• 28
• 14