I am new to WL pro and based on my search on the forum, it seems WL is currently single threaded ...and I assume Neuro Lab is also single threaded? In my experience with TensorFlow for regression type problems, neural networks get significant speed up with multi core support...is there any plan to make Neuro Lab multi core?
Size:
Color:
Sorry, there isn't any plan to make Neuro-Lab multi-core.
Size:
Color:
On a related note, the Monte Carlo Lab addin once had multi-core support which later got withdrawn due to a thread-safety issue. Despite a solid speed up, building MC equity curve could be erratic. Since NL would have to rely on similar multi-threading principles, it could be affected as well.
Size:
Color:
QUOTE:
... got withdrawn due to a thread-safety issue. ... building Monte Carlo Lab equity curve could be erratic.
So you're saying there are .NET datatypes WL is using "internally" that aren't thread safe.
I know some WL indicators are written to be multi-threaded, and they function fine. But they are only dependent on a DataSeries datatype and its base classes, which are much narrower scoped than what an optimizer (e.g. Monte Carlo Lab) or a performance visualizer would see.
Size:
Color:
QUOTE:
So you're saying there are .NET datatypes WL is using "internally" that aren't thread safe.
Let's put it this way: there's a WL internal class MCL was using that isn't thread safe.
Size:
Color: