How to get client script to run before rendering on Network Render

Question

Using Blender's Network Render addon, I am trying to render on a client's GPU using this script:

import bpy

FlagCUDA = False

#Select CUDA
try:
    bpy.context.user_preferences.system.compute_device_type = 'CUDA'
    FlagCUDA = True
except:
    print("No CUDA-Device")
    FlagCUDA = False

if FlagCUDA:    
    try:
        bpy.context.user_preferences.system.compute_device = 'CUDA_MULTI_2'
    except:
        bpy.context.user_preferences.system.compute_device = 'CUDA_0'
#Check which GPUs are there
print("Rendering on: " + bpy.context.user_preferences.system.compute_device)

However, when I try to render it, the script doesn't seem to be run until after the slave has already rendered the file! For example, here's the slave's console output:

Created directory: /tmp/slave_1942577ae3d0b404da672d202b2a9a58
Created directory: /tmp/slave_1942577ae3d0b404da672d202b2a9a58/job_7
Fullpath /home/tylere/untitled.blend
File: untitled.blend and 0 other files
frame 1
Color management: using fallback mode for management
Read new prefs: /home/tylere/.config/blender/2.69/config/userpref.blend
read blend: /home/tylere/untitled.blend
Fra:1 Mem:7.10M (9.89M, Peak 16.99M) | Mem:0.00M, Peak:0.00M | Scene, RenderLayer | Synchronizing object | Cube
...
RenderLayer | Updating Images
Fra:1 Mem:7.11M (9.89M, Peak 17.00M) | Mem:0.00M, Peak:0.00M | Scene, RenderLayer | Updating Camera
Fra:1 Mem:7.11M (9.89M, Peak 17.00M) | Mem:0.00M, Peak:0.00M | Scene, RenderLayer | Updating Objects
Fra:1 Mem:7.11M (9.89M, Peak 17.00M) | Mem:0.00M, Peak:0.00M | Scene, RenderLayer | Updating Objects | Copying Transformations to device
Fra:1 Mem:7.11M (9.89M, Peak 17.00M) | Mem:0.00M, Peak:0.00M | Scene, 
...
RenderLayer | Path Tracing Tile 134/135
Fra:1 Mem:7.11M (9.89M, Peak 17.07M) | Mem:0.51M, Peak:0.59M | Scene, RenderLayer | Path Tracing Tile 135/135
Fra:1 Mem:7.11M (9.89M, Peak 17.07M) | Mem:0.02M, Peak:0.59M | Scene, RenderLayer | Finished
Fra:1 Mem:7.08M (9.89M, Peak 17.07M) Sce: Scene Ve:0 Fa:0 La:0
Saved: /tmp/slave_1942577ae3d0b404da672d202b2a9a58/job_7/000001.exr Time: 00:01.44 (Saving: 00:00.09)

No CUDA-Device
Rendering on: CPU

Blender quit
status 0

(The client I am currently testing scripts on has no GPU, hence the "No CUDA-Device")

How do I get the script to run BEFORE the rendering occurs on the slave?

Answer 1

I expect it is the way the addon runs. The python script is in control as it reads the file and starts rendering, only after it has finished does blender sit idle and go back to running registered text blocks from the open file step.

First I would suggest you setup the slaves to always start with gpu rendering enabled.

Failing that, you could adjust the netrender addon on each slave. You will find in netrender/operator.py that rendering starts either on line 80 or 97 you could specifically run your script before the render step or add your gpu enable code to the addon.

text = bpy.data.texts['myStartscript.py']
if text is not None:
    exec(compile(text.as_string(), 'errors.log', 'exec'))
 bpy.ops.render.render('INVOKE_AREA', animation=True)