maxNumCompThreads() is limited to 1 on a multi-CPU Linux system

Issue:

>> maxNumCompThreads()

ans =

1

Steps taken:

1) Fresh network install of R2024b and R2025b in default folder using sudo

2) All available toolboxes (including Parallel Computing Toolbox) were installed.

3) Other non-MATLAB software using multithreading runs fine and can utilize all available CPUs

Question

Has anybody experienced similar issue before please? The lscpu shows "Core(s) per socket: 1" with 24 sockets available. "cat /proc/cpuinfo | grep "physical id" outputs 24 lines all pointing to physical ID: 0. Could it be because of the one-core-per-socket split?

Importantly, other multiprocessing tools, for instance python-based torch implementations, or tools like MRtrix, all recognize 24 cores without any problems.

I will appreciate any help on this.

Many thanks in advance!

Supporting information

System specifications

% Operating system:        Debian GNU/Linux 12 (bookworm) x86_64
% Kernel:                  6.1.0-41-amd64

CPU specifications (given by lscpu):

% Architecture:            x86_64
% CPU op-mode(s):          32-bit, 64-bit
% Address sizes:           46 bits physical, 48 bits virtual
% Byte Order:              Little Endian
% CPU(s):                  24
% On-line CPU(s) list:     0-23
% Vendor ID:               GenuineIntel
% Model name:              Intel(R) Core(TM) Ultra 9 285
% CPU family:              6
% Model:                   198
% Thread(s) per core:      1
% Core(s) per socket:      1
% Socket(s):               24
% Stepping:                2
% CPU(s) scaling MHz:      19%
% CPU max MHz:             7000.0000
% CPU min MHz:             800.0000
% BogoMIPS:                4992.00

"cat /proc/cpuinfo | grep "physical id"

physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0
physical id	: 0

[Walter Roberson]

maxNumCompThreads() would be 1 if the Parallel Computing Toolbox is not installed or is not licensed.

[Walter Roberson]

% Kernel: 6.1.0-41-amd64

That suggests you are using an AMD system.

% Model name: Intel(R) Core(TM) Ultra 9 285

That suggests that you are using a hardcore Intel system.

Perhaps you need an Intel-based Linux kernel?

[Daniel]

1) Parallel Computing Toolbox is licensed and installed. It is a university-wide license for the whole MATLAB suite, so all toolboxes are licensed. The toolbox was installed as a part of the installation.

ver

2) I think amd64 is just a x64 architecture, which is also what Intel uses, it is not a brand-specific solution. If this would be the issue, I believe other multithreading software would fail as well.

[Walter Roberson]

Hmmm....

There is the possibility that somehow the process pool settings are messed up.

https://www.mathworks.com/help/parallel-computing/parallel-preferences.html

[Daniel]

Thank you Walter for your help, your input guided me into the right direction and looking deeper into the local parpool cluster.

Answer 1

[Daniel]

I think I solved it. The solution involved:

1) adjust the number of workers within the local cluster

c = parcluster('local');
c.NumWorkers = 24;

2) save the adjusted cluster profile

saveProfile(c)

3) stop parpool and restart

delete(gcp('nocreate')); 
parpool

The NumWorkers within the parcluster('local') was set to 1 by default. Same for the number of threads. With the procedure above, I managed to spawn the desired amount of parallel workers.

*EDIT: removing an unnecessary step from the solution.

Answer 2

[Justin Hontz]

It seems that this issue is caused by a bug in the Linux kernel for this particular CPU model (Intel(R) Core(TM) Ultra 9 285K). See here:

https://github.com/util-linux/util-linux/issues/3902

Upgrading to a newer Linux kernel version (e.g. 6.12.73+deb12-amd64) appears to fix the issue.