# OpenGL Still slow with display lists

This topic is 4464 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

## Recommended Posts

Hi, the other week i startet a thread about my slow program. It seemed that the problem was that i drew in immidate mode but i discovered its not so. I got the exact same FPS in both immidiate mode and with display lists. When i disabled the textures the FPS only dropped a few values. I just dont know whats going on. Im only drawing 12*8 quads. Main.cpp:
#include "fhInclude.h"

int main(int argc, char *argv[])
{
SDL_Init( SDL_INIT_VIDEO );

SDL_GL_SetAttribute( SDL_GL_RED_SIZE, 8 );
SDL_GL_SetAttribute( SDL_GL_GREEN_SIZE, 8 );
SDL_GL_SetAttribute( SDL_GL_BLUE_SIZE, 8 );

SDL_GL_SetAttribute( SDL_GL_DEPTH_SIZE, 8 );
SDL_GL_SetAttribute( SDL_GL_DOUBLEBUFFER, 1 );

SDL_Surface *screen;
screen = SDL_SetVideoMode( 800, 600, 24, SDL_OPENGL | SDL_SWSURFACE );

glViewport( 0, 0, 800, 600 );

glMatrixMode( GL_PROJECTION );
glOrtho( 0.0f, 800.0f, 600.0f, 0.0f, 0.0f, -2.0f );

glClearColor( 0.0f, 0.0f, 0.0f, 0.0f );
glClearDepth( 2.0f );
glEnable( GL_DEPTH_TEST );
glDepthFunc( GL_LEQUAL );

glClear( GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT );

glMatrixMode( GL_MODELVIEW );
glTranslatef( 200.0f, 300.0f, 1.0f );

//variables for handling FPS (frames per second)
bool done = false;
SDL_Event evt;
int FPS     = 0;
int pastFPS = 0;
int past    = 0;
int currentTime = 0;

//variabels for key-states
int KeyLeft = 0; //0 means key is up, 1 means its down
int KeyRight = 0;
int KeyUp = 0;
int KeyDown = 0;

int SecKeyLeft = 0; //0 means key is up, 1 means its down
int SecKeyRight = 0;
int SecKeyUp = 0;
int SecKeyDown = 0;

int mode = 4;

fhTexture test;

fhSlice testSlice;
testSlice.txtr = &test;
testSlice.slice = new fhCoord;
testSlice.slice->x = 0;
testSlice.slice->y = 0;
testSlice.slice->x2 = 800;
testSlice.slice->y2 = 600;
testSlice.amount = 1;

testSlice.Compile();

int x = 0;
int y = 0;

int w = 800;
int h = 600;

int tx = 0;
int ty = 0;

while( 1 )
{
//Handling FPS counter
if( SDL_PollEvent(&evt) )
{
if (evt.type == SDL_QUIT)
done = true;
}

currentTime = SDL_GetTicks();
past = SDL_GetTicks();
FPS++;

//Checking if one second has passed and printing FPS to top of window
if ( currentTime - pastFPS >= 2000 )
{
static char buffer[20] = {0};
sprintf( buffer, "%d FPS", FPS/2 );
SDL_WM_SetCaption( buffer,0 );
FPS = 0;
pastFPS = currentTime;
}

glClear(GL_COLOR_BUFFER_BIT);

if (mode == 0)
{
//Checking key-states and changing the projection position
if (KeyLeft) x--;
if (KeyRight) x++;
if (KeyUp) y--;
if (KeyDown) y++;
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
}
else if (mode == 1)
{
if (KeyLeft) w--;
if (KeyRight) w++;
if (KeyUp) h--;
if (KeyDown) h++;
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
}
else if (mode == 2)
{
if (KeyLeft) tx--;
if (KeyRight) tx++;
if (KeyUp) ty--;
if (KeyDown) ty++;
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
}
else if (mode == 3)
{
if (KeyLeft) tx--, w--;
if (KeyRight) tx++, w++;
if (KeyUp) ty--, h--;
if (KeyDown) ty++, h++;
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
test.Draw( x, y, tx, ty, w, h );
}
else if (mode == 4)
{
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);
glCallList(*testSlice.data);

}

SDL_GL_SwapBuffers();

// Poll for events, and handle the ones we care about.
SDL_Event event;
while (SDL_PollEvent(&event))
{
switch (event.type)
{
case SDL_KEYDOWN:
switch (event.key.keysym.sym)
{
case SDLK_LEFT:
KeyLeft = 1;
break;
case SDLK_RIGHT:
KeyRight = 1;
break;
case SDLK_UP:
KeyUp = 1;
break;
case SDLK_DOWN:
KeyDown = 1;
break;
}
break;

case SDL_KEYUP:
switch (event.key.keysym.sym)
{
case SDLK_ESCAPE:
// If escape is pressed, return (and thus, quit)
test.Free();
return 0;
case SDLK_LEFT:
KeyLeft = 0;
break;
case SDLK_RIGHT:
KeyRight = 0;
break;
case SDLK_UP:
KeyUp = 0;
break;
case SDLK_DOWN:
KeyDown = 0;
break;
case SDLK_SPACE:
mode++;
if (mode == 5) mode = 0;
break;
}
break;

case SDL_QUIT:
return(0);
}
}
}

SDL_Quit();
return 0;
}


As i said last time, the code isnt in its best shape as i have been trying to change a lot of stuff...

##### Share on other sites
I think your using sdl/opengl software mode.

Try doing:

SDL_SetVideoMode( 800, 600, 24, SDL_OPENGL | SDL_HWSURFACE );

##### Share on other sites
Nope, that didnt help at all :(

##### Share on other sites
Display lists are barely any faster than immediate mode. You'll see a much better performance difference if you switch to Vertex Arrays or Vertex Buffer Objects.

##### Share on other sites
I don't believe the above is true. Most literature says Display Lists are very fast, the downside is the memory overhead for storing all the calls.

##### Share on other sites
Quote:
 Original post by BoderI don't believe the above is true. Most literature says Display Lists are very fast, the downside is the memory overhead for storing all the calls.

I've seen quite a bit of old literature suggesting display lists are fast, but if you're willing to really dig deep down into the documentation on the nvidia developer page, you'll see that for about 2-3 years, they've been trying to nudge people away from using display lists for anything more than speeding up state changes. Here's just one example:

slide 13

I don't exactly know why display lists aren't as fast as vertex buffer objects, but if you benchmark it, I'm sure you'll see a significant difference. In my own experiences, I've found the benefit of display lists to be barely noticeable while the benefit of vertex buffers is usually an order of magnitude.

##### Share on other sites
Yep, it's unfortunate that the documentation at opengl.org is version 1.1 and NeHe also shows display lists first.

Thanks for the reference cwhite.

##### Share on other sites
Actually, for purely static geometry, display lists can still beat vertex buffer objects by a few points. You have to make sure the display lists contain ONLY geometry, though, and compile them with GL_COMPILE -- if you compile with GL_COMPILE_AND_EXECUTE, it won't be fast.

I e, something like:
  set material state (textures, etc)  if don't have display list    GL_COMPILE    glBegin    glVertex ...    glEnd    capture display list  endif  draw display list

##### Share on other sites
The real problem is that OpenGL doesn't require hardware to be certified to quite nearly the same extent as DirectX. There are lots of things marked as "implementation specific" or "implementation dependent." For these things, the official OpenGL party line might be that two things should perform similarly, but Nvidia and ATI might have found that setting up their drivers or their hardware a certain way makes common API usages fast and uncommon API usages slow. For deep understanding of performance issues, you really need to sort through documents from the hardware vendors.

And to respond to hplus' insight, all of my previous benchmarking was done on code where the display lists encapsulated both state changes and geometry, so I don't have any intuitive sense for the performance of display lists when they are purely geometric. I would, though, wonder about how a display list could take advantage of the vertex cache the same way a well-optimized mesh with an index buffer can.

1. 1
2. 2
3. 3
Rutin
24
4. 4
JoeJ
18
5. 5

• 14
• 23
• 11
• 11
• 9
• ### Forum Statistics

• Total Topics
631766
• Total Posts
3002228
×