From Fedora Project Wiki
(some page cleanup and clarification, still needs "ReadyForWrangler")
(adding tracker bug)
 
(7 intermediate revisions by 3 users not shown)
Line 15: Line 15:
<!-- Include you email address that you can be reached should people want to contact you about helping with your change, status is requested, or technical issues need to be resolved. If the change proposal is owned by a SIG, please also add a primary contact person. -->
<!-- Include you email address that you can be reached should people want to contact you about helping with your change, status is requested, or technical issues need to be resolved. If the change proposal is owned by a SIG, please also add a primary contact person. -->
* Email: trix@redhat.com  
* Email: trix@redhat.com  
* Name: [[User:mystro256| Jeremy Newton]]
* Email: Jeremy.Newton@amd.com
<!--- UNCOMMENT only for Changes with assigned Shepherd (by FESCo)
<!--- UNCOMMENT only for Changes with assigned Shepherd (by FESCo)
* FESCo shepherd: [[User:FASAccountName| Shehperd name]] <email address>
* FESCo shepherd: [[User:FASAccountName| Shehperd name]] <email address>
-->
-->
== Current status ==
== Current status ==
[[Category:ChangePageIncomplete]]
[[Category:ChangeAcceptedF41]]
<!-- When your change proposal page is completed and ready for review and announcement -->
<!-- When your change proposal page is completed and ready for review and announcement -->
<!-- remove Category:ChangePageIncomplete and change it to Category:ChangeReadyForWrangler -->
<!-- remove Category:ChangePageIncomplete and change it to Category:ChangeReadyForWrangler -->
Line 37: Line 40:
ON_QA -> change is fully code complete
ON_QA -> change is fully code complete
-->
-->
* [<will be assigned by the Wrangler> Discussion thread]
* [https://lists.fedoraproject.org/archives/list/devel-announce@lists.fedoraproject.org/thread/VFPMF56ZGF5DNG47H75XKIFGC2MJTINQ/ Announced]
* FESCo issue: <will be assigned by the Wrangler>
* [https://discussion.fedoraproject.org/t/f41-change-proposal-rocm-6-2-self-contained/126071 Discussion thread]
* Tracker bug: <will be assigned by the Wrangler>
* FESCo issue: [https://pagure.io/fesco/issue/3256 #3256]
* Tracker bug: [https://bugzilla.redhat.com/show_bug.cgi?id=2305771 #2305771]
* Release notes tracker: <will be assigned by the Wrangler>
* Release notes tracker: <will be assigned by the Wrangler>


Line 45: Line 49:
<!-- Expand on the summary, if appropriate.  A couple sentences suffices to explain the goal, but the more details you can provide the better. -->
<!-- Expand on the summary, if appropriate.  A couple sentences suffices to explain the goal, but the more details you can provide the better. -->
ROCm 6.2 is the latest iteration of AMD's compute libraries that work with the linux kernel to allow users to run compute workloads on their GPU's.  As many GPU's as possible are enabled so open acceleration is available and easy for to as wide an audience as possible.
ROCm 6.2 is the latest iteration of AMD's compute libraries that work with the linux kernel to allow users to run compute workloads on their GPU's.  As many GPU's as possible are enabled so open acceleration is available and easy for to as wide an audience as possible.
ROCm 6.2 has not been released yet, but upstream has confirmed it will happen long before Fedora 41 code freeze.


== Feedback ==
== Feedback ==
Line 51: Line 57:
== Benefit to Fedora ==
== Benefit to Fedora ==
By providing system level acceleration, the challenges for developers to sync out of tree kernel drivers with complicated software stacks is removed.  Developers can focus on solving their problems, not the problems of the getting hardware acceleration working.
By providing system level acceleration, the challenges for developers to sync out of tree kernel drivers with complicated software stacks is removed.  Developers can focus on solving their problems, not the problems of the getting hardware acceleration working.
Moving to ROCm 6.2 means newer features and bug fixes. It also means moving to LLVM 18, which includes bugfixes, new hardware support, and reduces package maintenance of using an older LLVM package.
<!-- What is the benefit to the distribution?  Will the software we generate be improved? How will the process of creating Fedora releases be improved?
<!-- What is the benefit to the distribution?  Will the software we generate be improved? How will the process of creating Fedora releases be improved?
    
    
Line 83: Line 91:


<!-- What work do the feature owners have to accomplish to complete the feature in time for release?  Is it a large change affecting many parts of the distribution or is it a very isolated change? What are those changes?-->
<!-- What work do the feature owners have to accomplish to complete the feature in time for release?  Is it a large change affecting many parts of the distribution or is it a very isolated change? What are those changes?-->
* Other developers: <!-- REQUIRED FOR SYSTEM WIDE CHANGES --> Fedora tools has to release compat packages for llvm18 because ROCm 6.2 is not compatible with llvm19 which is slated for inclusion with F41 (this is already planned and on their schedule). The ROCm upstreams need to finalize the 6.2 release which is expected prior to F41's release
* Other developers: <!-- REQUIRED FOR SYSTEM WIDE CHANGES --> Fedora tools has to release compat packages for llvm18 because ROCm 6.2 is not compatible with llvm19 which is slated for inclusion with F41 (this is already planned and on their schedule). The ROCm upstreams need to finalize the 6.2 release which is expected prior to F41's release
<!-- What work do other developers have to accomplish to complete the feature in time for release?  Is it a large change affecting many parts of the distribution or is it a very isolated change? What are those changes?-->
<!-- What work do other developers have to accomplish to complete the feature in time for release?  Is it a large change affecting many parts of the distribution or is it a very isolated change? What are those changes?-->

Latest revision as of 14:04, 19 August 2024

ROCm 6.2

This is a proposed Change for Fedora Linux.
This document represents a proposed Change. As part of the Changes process, proposals are publicly announced in order to receive community feedback. This proposal will only be implemented if approved by the Fedora Engineering Steering Committee.

Summary

The latest ROCm release in sync with AMD's latest upstream release. Builds on the math libraries released in F40 to also include packages to run AI workloads. Has been integrated with Fedora's PyTorch.

Owner

Current status

Detailed Description

ROCm 6.2 is the latest iteration of AMD's compute libraries that work with the linux kernel to allow users to run compute workloads on their GPU's. As many GPU's as possible are enabled so open acceleration is available and easy for to as wide an audience as possible.

ROCm 6.2 has not been released yet, but upstream has confirmed it will happen long before Fedora 41 code freeze.

Feedback

Benefit to Fedora

By providing system level acceleration, the challenges for developers to sync out of tree kernel drivers with complicated software stacks is removed. Developers can focus on solving their problems, not the problems of the getting hardware acceleration working.

Moving to ROCm 6.2 means newer features and bug fixes. It also means moving to LLVM 18, which includes bugfixes, new hardware support, and reduces package maintenance of using an older LLVM package.

Scope

  • Proposal owners: ROCm Packagers Sig has to update the current set from 6.1.2 to 6.2.0
  • Other developers: Fedora tools has to release compat packages for llvm18 because ROCm 6.2 is not compatible with llvm19 which is slated for inclusion with F41 (this is already planned and on their schedule). The ROCm upstreams need to finalize the 6.2 release which is expected prior to F41's release
  • Release engineering: N/A (not needed for this Change)
  • Policies and guidelines: N/A (not needed for this Change)
  • Trademark approval: N/A (not needed for this Change)
  • Alignment with the Fedora Strategy: "Innovation & Leadership in Technology: We integrate programming language stack ecosystems" by including the latest release of the ROCm ecosystem, Fedora is better enabling heterogeneous computation and supporting the latest hardware for ROCm.

Upgrade/compatibility impact

There should be no backwards incompatible changes for this upgrade, it is a minor release.

Early Testing (Optional)

Do you require 'QA Blueprint' support? N

How To Test

AMD GPU hardware is needed. Most of the ROCm packages have option test subpackages which need to be built locally for a single specific AMD shader family (e.g gfx1100). These will be built and run.


User Experience

Dependencies

Contingency Plan

The current ROCm set 6.1.2 has all of the current packages expected in 6.2 llvm17 will be maintained by the ROCm packagers sig (Tom Rix)

  • Contingency mechanism: (What to do? Who will do it?) N/A (not a System Wide Change)
  • Contingency deadline: N/A (not a System Wide Change)
  • Blocks release? N/A (not a System Wide Change), Yes/No


Documentation

N/A (not a System Wide Change)

Release Notes