bgl.Buffer creation sloooooow

Coding Python Support
1

Hi,

I have written a script that displays a bytearray (via texture) in a 3d view. So far, everything is running, however creating a bgl.Buffer from my bytearray is way too slow (80ms for 512x512).

What I do is:
buffer = bgl.Buffer(bgl.GL_BYTE, [512 * 512 * 4], data) # <-- more than 80ms here!
bgl.glTexImage2D(bgl.GL_TEXTURE_2D, 0, bgl.GL_RGBA, 512, 512, 0, bgl.GL_RGBA, bgl.GL_UNSIGNED_BYTE, buffer)

Is there a better way to do that? I could as well call an external program if that is possible and makes it faster…?

2

Don’t know if it’ll be any faster since pixel access is also slooooow but you can create a gl buffer directly from an Image class. When I was testing it after porting it over from 2.4x it didn’t seem that slow and I don’t have a VFX card anyone would consider fast.

Actually, now that I think about it, I was using bgl.Buffer to draw the webcam image in the 3dview in one of my boring ass videos so it can’t be all that slow.

----edit----

Eh, looking at the code in the video it turns out I wasn’t really using bgl.Buffer…

3

bgl buffer seems to be always slow, if you look at the code you’ll see that it’s a python object. I didn’t fully understand the code, but it seems like it’s often re-initilized and changes are made the python-way. That said, I suggest to write things in plain C if speed is important.

4

Thanks for your replies. Yes, the issue here is that bgl.Buffer is a Python object, which then allocates memory and copies data in every update view call.

In the meantime I wrote a python extension module (C module, as also suggested by CoDEmanX) and now do everything there. This is - even for small resolutions - roughly 100x faster :slight_smile:

I initially wanted to avoid a separate extension module, but this seems to be the only way to go.

5

wow sounds great! could you share (at least parts) of the code? I’m mainly interested into how to create such a python extension and i’m also interested into bgl buffer - i’d like to read the buffer of the 3d view for several frames and combine that into a new buffer and render it in the 3d view (for ghosting).

6

Posting the whole code here does not really make sense as it is already very long and specialized. A good starting point for extension modules is http://docs.python.org/2/extending/extending.html

What you need to do is

  1. Write your own extension with a method including all your OpenGL draw code
  2. compile
  3. put it in somewhere where blender/python can find it (e.g. in the folder containing your script)
  4. call it from your python script (e.g. from view_draw of a class that extends bpy.types.RenderEngine).

Hope that helped somehow.

I still have to figure out how to trigger redraw events from a script. Does anybody have a clue?

7

Thanks for that outline nb5000!
4. Does it mean the 3d view in blender will only show what is drawn by the opengl code of the extension? For my puposes, I would need the usual screen content of the POST_VIEW state as input for processing (multiple 3d view captures blended with varying transparency)

bpy lets you tag regions and areas for redraw, but it doesn’t force an immediate redraw:

e.g. bpy.context.area.tag_redraw()

Also read this:
http://www.blender.org/documentation/blender_python_api_2_68_2/info_gotcha.html#can-i-redraw-during-the-script