How to control number of CPU workers in JMP Pro?

[#1]

I like that JMP parallelizes computations across all of my 8 CPUs, but the use of the one last CPU is of great inconvenience, since my PC is tied up in computations and is not responsive. How can I instruct JMP to use only 7 (or total count minus 1) CPUs on my PC, preferrably in a way that is persistent across all of my projects and documents (i.e. does not need to be configured every time I open a file or launch JMP)? Many thanks in advance. 

[#2 · Accepted solution]

This is not something that can be controlled by JMP but can be controlled by your operating system

From a thread on SuperUser/StackExchange:

In Windows 10, 8 and 7

1. Go to Task Manager
2. Right-click the process for which cpu usage is to be limited. click "Go To Details"
3. Now detail tab will show up, right-click the particular process and choose "Set Affinity" and choose the cores which you allow the particular process to use.

---

Doing this for the JMP process should limit the number of cores available for JMP to use.

[#3]

You question has started a conversation between several developers at JMP. (Justin is correct that JMP currently does not have a way to limit the number of threads beyond what the internal algorithms do.)

Can you tell us if memory was also used up (the memory usage was cut off the picture), and what analysis platform you were using, how many rows in the table, etc? We think this is not usually a problem with most of the JMP analysis platforms.

Thanks!

[#4]

Hi Craige, I'm glad to hear the conversation has started. Ideally, I would look for a way to control the number of CPU workers.

For the current model, JMP uses under a 1 GB RAM (total 32 GB and 12 used by all apps+OS), so RAM is not a concern. 

The process priority can be (persistently) controled by ProcessHacker and similar tools, but there is no direct/persistent way to specify the number of CPUs to utilize. The posted work-around is great, but it seems I need to set it every time JMP starts.

RE: Analysis platform. I could not find what you mean by this, but I suppoes you meant what model I ran. So, I ran Generalized Regression+Lasso with a custom validaiton column and some Log & Logit transformated features. There are about 20 features and 10^6 observations in this model, but some categorical features have many levels, resulting in about 200 dummy variables (hence, use of LASSO). 

Please note that I am not concerned with high CPU utilization. This was by design (I increased priority to HIGH). What I could not find is a way to use only 7 out of 8 of my CPUs, so I can continue working with a laptop.

I hope it helps.

[#5]

Hi Justin. I find your answer of high quality and to the point. Thank you for clearly and succinctly answering the question!!