Question

I’m using the following code to draw characters on screen (each UTF8 character is a texture):

int row = 0;

glTexCoordPointer(2, GL_FLOAT, 0, texCoords);
glVertexPointer(2, GL_FLOAT, 0, vertices);

for (StdStr* line in _lines) {
    const char* str = [line cString];

    for (int i = 0; i < strlen(str); i++) {
        ((XX::OGL::GLESContext*)context)->viewport(C_WIDTH*i,
                                                   C_HEIGHT*row,
                                                   C_WIDTH,
                                                   C_HEIGHT);

        glColor4f(1.0, 1.0, 1.0, 1.0);

        glBindTexture(GL_TEXTURE_2D, _textures[0] + *(str + i));
        glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);  
    }

    row++;
}

When there are a lot of characters, the code takes longer to run. In this case, almost 99% of the time is spent in the glDrawArrays routine. Is it possible to minimise the amount of calls to glDrawArrays? The OpenGL ES version is 1.1.

Answer 1

Actually I think that you should try to limit the amount of calls to viewport, glBindTexture and glDrawArrays.

Technically, you should pack all your characters in a single texture, so that you can bind it once.

Then, you could compute the vertices and texcoords in a loop like you do actually, but doing the viewport maths yourself, and accumulating results in a CPU array. Once your array constituted, you should submit a draw call once, providing this array.

You can probably find inpiration here:

http://www.angelcode.com/products/bmfont/

http://sourceforge.net/projects/oglbmfont/