glCompressedTexImage2D and NULL data

OpenGL OpenGL: Advanced Coding
1

Is it allowed to pass NULL as data to {Compressed}TexImage() to “reserve space” without downloading anything ?

It seems to crash on both Nvidia and ATI

The motivation for doing this is to download a large precompressed texture in small pieces by first creating an empty texture with glCompressedTexImage2D(…, NULL) and then subloading small blocks of texture data with glCompressedTexSubImage2D(…, data)

Thanks

2

I’m not sure why this crashes, but surely you could use glTexImage2D(…, NULL) for reserving the texture (with a compressed internalformat). And then using subsequent glCompressedTexSubImage calls to upload the small pieces.

3

I experienced something similar some time ago.

Worth noting might be that the code in that thread does not crash on an Intel implementation I tested it on (though it should be noted I didn’t really check the resulting texture to see if it indeed was compressed).

4

Originally posted by golem:
Is it allowed to pass NULL as data to {Compressed}TexImage() to “reserve space” without downloading anything ?
The spec doesn’t say NULL is allowed, so that would mean it’s not. But as andras said, you can use glTexImage2D().

5

Andras,

Your suggestion is working !
Thank you very much

I suppose this is the normal way to use the API. I think this should have been explained better in the documentation. I am pretty sure i am not the only one to be confused.

6

Sorry to revive the thread, but I’m having a similar problem that when I call glTexImage2D(… NULL) for compressed textures, and then glCompressedTexSubImage2D(…, data), it always displays a white texture.

I’m running on nVidia ForceWare ver. 81.63 (Quadro FX 1400).

Does anyone else have similar problems?

7

glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);

8

Thanks, knackered. Apparently, the MIN filter does matter (is mip map the default?), but my problem was that I was calling glTexImage2D() with the wrong parameters.

I was calling glTexImage2D() with GL_COMPRESSED_RGB_S3TC_DXT1_EXT instead of GL_RGB for the “format” parameter.

Here’s my testing code for anyone that might bump into similar problems:

  glEnable(GL_TEXTURE_2D);
  glBindTexture(GL_TEXTURE_2D, texHandle);

  unsigned char red_pixels[4][4][4];
  for (int i=0; i < 4; ++i) {
    for (int j=0; j < 4; ++j) {
      red_pixels[i][j][0] = 0xff;
      red_pixels[i][j][1] = 0x00;
      red_pixels[i][j][2] = 0x00;
      red_pixels[i][j][3] = 0xff;
    }
  }

  unsigned char red_dxt1[4 * 4 / 2] =
  { 0x00, 0xf8, 0x00, 0xf8, 0x00, 0x00, 0x00, 0x00 };

  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
  glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
  static int chooser = 2;
  switch (chooser) {
    case 0:
      // Load uncompressed image in one shot.
      // Should work.
      glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, 4, 4, 0, GL_RGBA, GL_UNSIGNED_BYTE, (void *)red_pixels);
      break;
    case 1:
      // Load compressed image in one shot.
      // Should work.
      glCompressedTexImage2D(GL_TEXTURE_2D, 0, GL_COMPRESSED_RGB_S3TC_DXT1_EXT, 4, 4, 0, 8, (void *)red_dxt1);
      break;
    case 2:
      // Make space for uncompressed image, then load through subtexture region.
      // Should work.
      glTexImage2D(GL_TEXTURE_2D, 0, GL_COMPRESSED_RGB_S3TC_DXT1_EXT, 4, 4, 0, GL_RGB, GL_UNSIGNED_BYTE, NULL);
      glCompressedTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, 4, 4, GL_COMPRESSED_RGB_S3TC_DXT1_EXT, 8, (void *)red_dxt1);
      break;
    case 3:
      // Just load subtexture region.
      // Should NOT work.
      glCompressedTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, 4, 4, GL_COMPRESSED_RGB_S3TC_DXT1_EXT, 8, (void *)red_dxt1);
      break;
  }


  glBegin(GL_TRIANGLE_STRIP);
  glTexCoord2f(-1, -1);
  glVertex3f(-1, -1, 0.5);
  glTexCoord2f(1, -1);
  glVertex3f(1, -1, 0.5);
  glTexCoord2f(-1, 1);
  glVertex3f(-1, 1, 0.5);
  glTexCoord2f(1, 1);
  glVertex3f(1, 1, 0.5);
  glEnd();
9

Originally posted by canuckle:
is mip map the default?
Yes. Specifically, it’s GL_NEAREST_MIPMAP_LINEAR.