Add a chapter about scikit-learn-intelex by Vika-F · Pull Request #40 · intel/optimization-zone

Vika-F · 2026-06-30T11:48:08Z

Add information about sklearnex configurations that can provide better performance.

CC @napetrov @david-cortes-intel @avolkov-intel @Alexandr-Solovev

david-cortes-intel

Some comments:

Could make it clear that OMP settings related to affinity will not have an effect.
Could provide commands to do all of this programmatically instead of manually listing the core IDs after looking at output.
This advise is not really specific to sklearnex. Maybe could be put as an additional section and then linked from all the relevant articles, like NumPy and scikit-learn.

david-cortes-intel · 2026-06-30T12:14:48Z

+
+To set EPB to Performance mode:
+```
+sudo cpupower set -b 0


It doesn't explain the longevity of the effect from this command. Same for the other CPU settings.

@Vika-F Bumping up comment.

Vika-F · 2026-06-30T15:03:02Z

Could make it clear that OMP settings related to affinity will not have an effect.

Ok, will do.

Could provide commands to do all of this programmatically instead of manually listing the core IDs after looking at output.

You mean, python commands doing the same things like affinity modification?

This advise is not really specific to sklearnex. Maybe could be put as an additional section and then linked from all the relevant articles, like NumPy and scikit-learn.

Yes, I thought about it as well, but was not sure. I will restructure the pages accordingly.

david-cortes-intel · 2026-06-30T15:07:28Z

You mean, python commands doing the same things like affinity modification?

I guess they'd have to be shell commands, otherwise they wouldn't have an effect on an already-launched python process.

Vika-F · 2026-07-01T08:16:33Z

I guess they'd have to be shell commands, otherwise they wouldn't have an effect on an already-launched python process.

I understand the idea, but it will be a multi-line script for sure. Not a single sed .... Will it be Ok? I mean, I do not see similar code in the other chapters.

david-cortes-intel · 2026-07-01T08:20:46Z

I guess they'd have to be shell commands, otherwise they wouldn't have an effect on an already-launched python process.

I understand the idea, but it will be a multi-line script for sure. Not a single sed .... Will it be Ok? I mean, I do not see similar code in the other chapters.

@Vika-F isn't there any built-in numactl command do something like "run only physical cores"?

Vika-F · 2026-07-01T11:36:14Z

@Vika-F isn't there any built-in numactl command do something like "run only physical cores"?

No, there is no such option.

Vika-F · 2026-07-01T11:38:51Z

@david-cortes-intel An example of a script that runs a process only on physical nodes:
run_no_ht.sh

For me it looks too heavy for the page like this. What do you think?

david-cortes-intel · 2026-07-01T11:42:49Z

@david-cortes-intel An example of a script that runs a process only on physical nodes: run_no_ht.sh

For me it looks too heavy for the page like this. What do you think?

@Vika-F How about something in a pipeline (perhaps with awk?) getting the result from lscpu with arguments, filtering, and selecting a column?

david-cortes-intel · 2026-07-02T13:40:15Z

+
+```bash
+export MIN_MHZ=3000
+cpus=$(lscpu -e=cpu,core,maxmhz | tail -n +2 | awk -v min="$MIN_MHZ" '$3 >= min && !seen[$2]++ {printf sep $1; sep=","}')


I think this MIN_MHZ could also be obtained with awk from lscpu -e=cpu,core,maxmhz.

The problem is that there are various SKUs available, i.e. with P-cores only; with P-cores and E-cores; with E-cores and LP E-cores, etc.

Without MIN_MHZ the solution can exclude not LPE, but another type of cores with the lowest frequency.
That's why I decided just to provide the MHz threshold.

david-cortes-intel · 2026-07-03T09:48:38Z

@@ -0,0 +1,133 @@
+This chapter contains information about the practices that lead to better performance of scikit-learn-intelex on Intel CPUs.


Needs title and table of contents.

Copilot

Pull request overview

Adds new documentation aimed at improving scikit-learn-intelex workload performance on Intel CPUs by describing recommended system/CPU configuration and process pinning practices, and links the new chapter from the repo’s main README.

Changes:

Add a new scikit-learn-intelex chapter covering hyper-threading and LPE-core affinity recommendations (Windows/Linux).
Add a shared “common” chapter with OS-level recommendations (EPB/EPP and CPU frequency scaling).
Link the new scikit-learn-intelex chapter from the top-level README.

Reviewed changes

Copilot reviewed 3 out of 9 changed files in this pull request and generated 6 comments.

File	Description
`software/scikit-learn-intelex/README.md`	New chapter with HT/LPE detection and affinity guidance for Windows and Linux.
`software/common/README.md`	New common system tuning guidance (EPB/EPP and CPU frequency scaling).
`README.md`	Adds the scikit-learn-intelex chapter to the main table of contents.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+
+### On Windows
+
+Hyper-threading can be deteched by running **Task Manager**. Then navigate to **Performance** > **CPU** tab.


+
+<img src="images/cpu-affinity-ht.png" alt="drawing" style="width:600px;"/>
+
+Which is equivalent to `2BFF` in hexadecimal format. Run following command to disable HT on Windows:


+The recommended affinity mask that disables both hyper-threading and LPE cores would be `2BFC`:
+


+From the output we can see that 4 logical processors (0, 1, 2, 3) are running on two physical cores (0, 1).
+To run the process on physical cores only, use one of the following commands:
+
+```
+numactl -C 0,2,4-13 python <workload.py>
+```
+or
+```
+taskset -c 0,2,4-13 python <workload.py>
+```


+Run the following command to disable HT and LPE cores on Linux:
+
+```
+numactl -C 0,2,4-11 python <workload.py>
+```


+## Energy Performance Bias (EPB)
+
+Energy Performance Bias (EPB) is an Intel Xeon hardware setting that controls the trade-off between power consumption and processing performance. For the best performance, it is recommended to set it to `0` (Performance mode).
+
+### On Windows
+
+Run the following command in `cmd`:
+
+```
+powercfg -setacvalueindex scheme_current sub_processor PERFEPP 0
+```
+
+[More info about `powercfg`](https://learn.microsoft.com/en-us/windows-hardware/customize/power-settings/options-for-perf-state-engine-perfenergypreference).
+
+### On Linux
+
+To check the current value of EPB, run:
+```
+sudo cpupower info
+```
+
+To set EPB to Performance mode:
+```
+sudo cpupower set -b 0
+```


adgubrud · 2026-07-10T20:02:29Z

    - [Redis](software/similarity-search/redis/README.md)
  - [Spark](software/spark/README.md)
  - [scikit-learn](software/scikit-learn/README.md)
+  - [scikit-learn-intelex](software/scikit-learn-intelex/README.md)


I propose that this PR is changed to live completely in the Software->Common directory

adgubrud · 2026-07-10T20:04:26Z

@@ -0,0 +1,140 @@
+# Introduction


The contents for the scikit-learn-intelex README don't seem specific to scikit-learn-intelex. I recommend merging this README with the one you created in the "common" directory.

Vika-F added 4 commits June 26, 2026 04:40

Initial commit

22e3171

Update table of context

3e1fa76

Update

6cc553b

Update

71be167

david-cortes-intel reviewed Jun 30, 2026

View reviewed changes

Vika-F added 3 commits July 1, 2026 05:12

Add common recommendations about system configuration

e9d7534

Add the commands to switch off the HT programmatically in bash

62a688e

Add command sequence for LPE cores

24ef0dc

david-cortes-intel reviewed Jul 2, 2026

View reviewed changes

david-cortes-intel reviewed Jul 3, 2026

View reviewed changes

Add table of contents

2e3b89b

david-cortes-intel approved these changes Jul 3, 2026

View reviewed changes

adgubrud requested review from adgubrud and Copilot July 10, 2026 19:58

Copilot started reviewing on behalf of adgubrud July 10, 2026 19:58 View session

Copilot AI reviewed Jul 10, 2026

View reviewed changes

adgubrud requested changes Jul 10, 2026

View reviewed changes

		@@ -0,0 +1,133 @@
		This chapter contains information about the practices that lead to better performance of scikit-learn-intelex on Intel CPUs.


		### On Windows

		Hyper-threading can be deteched by running Task Manager. Then navigate to Performance > CPU tab.


		<img src="images/cpu-affinity-ht.png" alt="drawing" style="width:600px;"/>

		Which is equivalent to `2BFF` in hexadecimal format. Run following command to disable HT on Windows:

		The recommended affinity mask that disables both hyper-threading and LPE cores would be `2BFC`:

Uh oh!

Conversation

Vika-F commented Jun 30, 2026

Uh oh!

david-cortes-intel left a comment

Choose a reason for hiding this comment

Uh oh!

david-cortes-intel Jun 30, 2026

Choose a reason for hiding this comment

Uh oh!

david-cortes-intel Jul 3, 2026

Choose a reason for hiding this comment

Uh oh!

Vika-F commented Jun 30, 2026

Uh oh!

david-cortes-intel commented Jun 30, 2026

Uh oh!

Vika-F commented Jul 1, 2026

Uh oh!

david-cortes-intel commented Jul 1, 2026

Uh oh!

Vika-F commented Jul 1, 2026

Uh oh!

Vika-F commented Jul 1, 2026

Uh oh!

david-cortes-intel commented Jul 1, 2026

Uh oh!

david-cortes-intel Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

Vika-F Jul 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

david-cortes-intel Jul 3, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

adgubrud Jul 10, 2026

Choose a reason for hiding this comment

Uh oh!

adgubrud Jul 10, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Vika-F Jul 3, 2026 •

edited

Loading