Regarding Quad-Core support
Author: bgood
Creation Date: 7/14/2009 4:26 PM
profile picture

bgood

#1
Hello all, I searched and found discussions w.r.t. WL4 and quad-core processing, but I cannot find anything regarding WL5 and supporting multiple-core processors. The latest update with the optimizations is very cool, but WL5 appears to only utilize 1 CPU. While I am sure that there are many features you are working on adding, is there any chance that multi-core support could make the list?
profile picture

Eugene

#2
WL5 utilizes multi-core CPUs by virtue of .NET. What you're probably observing is that the CPU load isn't 100%.

As a workaround, you could launch 4 copies (not workspaces) as suggested in this FAQ and set processor affinity in Task Manager. Still it's possible to run several copies of Wealth-Lab 6 under different Windows user names.
profile picture

bgood

#3
Hi Eugene. Let me start by saying that you guys have been a great support. Next, there are many features of Wealth-Lab that I have not yet explored, and workspaces is one of them. I have been testing using the new optimize link. I have had one optimization running many simulations (30 minutes), and my quad-core only was working about 26%. It might be that be using workspaces I could utilize more of my CPU. I did try to bring up a second strategy and view it while first strategy was optimizing, and I encountered issues. Unfortunately I was not able to quantify the issue well enough to report a cogent defect.

Regarding the link above, I am confused regarding what is exactly being described. The phrase "under different Windows user names" appears that you need to log into windows as 1 user, connect Wealth-Lab and start it running. Next "Start->Switch users" leaving the first user running and repeat the above process. Do I understand correctly? If my understanding is correct, then can multiple computers be simultaneously running Wealth-Lab each streaming data, under the same users connection?
profile picture

Eugene

#4
Hi Bryce,

Although I'm not sure that it's desired and you really should be judging the application's performance by its CPU load -- considering that Wealth-Lab 5 backtests and optimizes several times faster than legacy versions did -- yes, this is a workaround that I was suggesting to utilize all of the 4 cores to max the CPU load.

Note that it's not possible to optimize on streaming data anyway, so for static data breaking a huge optimization task in 4 can be a reasonable workaround.

I remember that multi-core optimization was suggested to the software architects at Fidelity.
profile picture

bgood

#5
I just attempted to run one optimization in the first workspace and a second optimization in a second workspace. Wealth-Lab crashed!
profile picture

Eugene

#6
A crash is not good. Optimizing in 2 workspaces should work. Could you please share more details about the estimated optimization time, data loading settings and bar scale, number of runs, number of parameters etc?
profile picture

bgood

#7
Hello Eugene,

First let me start by saying that I know the example is not exactly correct, but it demonstrates the defect. First get GOOG and BIDU 1 minute bars. Next optimize for 3 months of 30 Minute candles using the Monte Carlo method for Profit with 2000 runs and 10 tests (20000 total runs). With the program below, Wealth-Lab will stop crash in seconds.

CODE:
Please log in to see this code.
profile picture

Eugene

#8
First of all, sorry for not pointing you to the workspace-based multi-core optimization right from the start; glad you found it by yourself.

Thanks for the report. Would you mind creating a support ticket and providing some more information, such as:

How many RAM do you have installed?
Operating system?
Is it x86 or x64?
What about the /3GB switch, IncreaseUserVA parameter, the 4GB patch?
profile picture

bgood

#9
I will be happy to create a support ticket. I am not a windows expert, so what is the easiest way to find out the /3GB switch, IncreaseUserVA, and 4GB patch?
profile picture

Eugene

#10
OK thanks. Don't bother about finding out: if you don't know what they're about, you're most likely not using them --
WL5 Wiki FAQ | General > Memory
This website uses cookies to improve your experience. We'll assume you're ok with that, but you can opt-out if you wish (Read more).