Texture Range, Client storage and pixel formats

Hi, all

I've been following the OpenGL Programming Guide for Mac OS X section on Texture techniques. I've been trying to see if the Intel integrated graphics can leverage shared memory to avoid as many texture copies as possible.

I've gotten GL_APPLE_client_storage to work appropriately (shaving-off an order of magnitude of texture load time [ 20mS vs 130mS in glTexImage3D with an 128 texel cube]).

Now I'm combining that extension with glTextureRangeAPPLE() to try and reduce that further. There is no appreciable improvement. I read further and came across the "Optimal data formats and types" section of the document and hit a brick-wall. I've been trying to use the GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV pair. The performance is TERRIBLE! load times have jumped up to 151mS in glTexImage3d for the same texture size.

Are there any suggestions? Am I completely out-to-lunch about trying to avoid a memory copy? Is the optimal data formats document incorrect?

For reference, here is the code used to load the texture naively:

glBindTexture( GL_TEXTURE_3D, texID );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_NEAREST );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP );

glTexEnvf( GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE );

glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA, width, height, depth, 0, GL_RGBA, GL_FLOAT, data );

With client-storage

glBindTexture( GL_TEXTURE_3D, texID );

glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_TRUE);

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_NEAREST );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP );

glTexEnvf( GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE );

glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA, width, height, depth, 0, GL_RGBA, GL_FLOAT, data );

With client-storage and texture range and "optimal" pixel format

glBindTexture( GL_TEXTURE_3D, texID );

glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_STORAGE_HINT_APPLE, GL_STORAGE_SHARED_APPLE);

glTextureRangeAPPLE(GL_TEXTURE_3D, width * height * depth * 4 * sizeof(GL_UINT), data);

glPixelStorei(GL_UNPACK_CLIENT_STORAGE_APPLE, GL_TRUE);

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_NEAREST );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP );

glTexParameterf(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP );

glTexEnvf( GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE );

glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA, width, height, depth, 0, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, data );

Thanks in advance for any help!

William Dillon

email@hidden