I don't see any error here. Please write down the error you get on the console. Also what is your OS, and if any extra step is required to reproduce the error (e.g., open the Performance panel? click settings?)

I am so sorry about that, here is complete steps to reproduce the error:

1. Switch to Cycles render
2. In Cycles settings change Cumpute Device from None to CUDA
3. In render settings change Device from CPU to GPU Compute
4. Activate the Auto Tile Size add-on
5. Look into console

I guess if you'll use OpenCL instead of CUDA it would result in the same error, but I cannot check this.

I can confirm the issue with dd92123 on win 7

line 149, in engine_is_gpu
    return engine == 'CYCLES' and device == 'GPU' and userpref.system.compute_device_type != 'NONE'
AttributeError: 'UserPreferencesSystem' object has no attribute 'compute_device_type'

The same problem was reported on https://developer.blender.org/D2338#54191

The userpref.system.compute_device_type check needs to be changed.

@Dalai Felinto (dfelinto) I hope you don't mind that I reopened and updated the task?
Also I can take the assignment if you guys have your hands full.

I have some code that solves the current bug, however the function get_threads will fail too

gpu_device_str = userpref.system.compute_device
if 'MULTI' in gpu_device_str:
    threads = int(gpu_device_str.split('_')[-1])
else:
    threads = 1

The number of threads is calculated by splitting the string from the end.
I'm not sure where to look in this case, so any pointers are welcomed.

would propose the following fixes:

method engine_is_gpu in line 149:

def engine_is_gpu(engine, device, userpref):
	if hasattr(userpref.system, 'compute_device_type'):
		_devType = userpref.system.compute_device_type
	else:
		_devType = userpref.addons.get('cycles').preferences.compute_device_type
	return engine == 'CYCLES' and device == 'GPU' and  _devType != 'NONE'

method get_threads in line 206:

def get_threads(context, device):
	render = context.scene.render
	engine = render.engine
	userpref = context.user_preferences

	if engine_is_gpu(engine, device, userpref):
		if hasattr(userpref.system, 'compute_device_type'):
			gpu_device_str = userpref.system.compute_device
			if 'MULTI' in gpu_device_str:
				threads = int(gpu_device_str.split('_')[-1])
			else:
				threads = 1
		else:
			threads = len([x for x in userpref.addons.get('cycles').preferences.devices if x.use])
	else:
		threads = render.threads

	return threads

However, only tested on a multi-CUDA-GPU rig, don't have OpenCL devices at hand here