[RFC] Enabling oneDNN Operations for x86 CPUs

[Penporn Koanantakool]

Hi all,

The RFC to enable oneDNN operations for x86 CPUs is available for comments. 

The feedback phase will be open for two weeks until Tuesday, November 9, 2021. Please leave comments directly on the pull request linked.

Summary

This RFC proposes enabling oneDNN operations in TensorFlow x86 Linux packages by default to efficiently utilize CPU hardware features.

Thank you in advance for your feedback!

Penporn

[Penporn Koanantakool]

Hi all,

We have tentatively scheduled a public design review meeting for this RFC.
Time: Thursday, November 11th, 2021 from 3-4pm PT
Meeting link: Google Meet

Meeting participants are expected to read the RFC ahead of time as the meeting will focus on the remaining issues/questions.

If there are no outstanding issues on this RFC by the time the feedback phase closes on Tuesday November 9th, 2021, we may cancel the design review meeting. (We will update here.)

Best,
Penporn

[Penporn Koanantakool]

Hi all,

The design review meeting has been canceled since there are no outstanding issues that require discussion.

Best,

Penporn

[Penporn Koanantakool]

Hi all,

A quick follow-up that oneDNN operations were enabled by default on recent x86 CPUs on Linux starting from TF 2.9. Next, we plan to also enable it by default on:

• aarch64 CPUs on Linux (starting with Neoverse V1) 
• x86 CPUs on Windows

If anyone as any concerns, please post them on the RFC's pull request by July 12, 2023.

Thank you!

Penporn