open source wg 2024-03-26 meeting notes

meetings/2024-01-24.rst

@@ -0,0 +1,122 @@
+1/24/2024
+=========
+
+Attendees:
+----------
+
+* Andrew Wafaa, Arm (Open Source Working Group chair)
+* Abhishek Jain, Fujitsu
+* Alex Pim, Imagination Technologies
+* Alexey Kukanov, Intel
+* Ankit Manerikar, Intel
+* Denis Samoilov, Intel
+* Igor Safanov, Intel
+* Steve Capper, Arm
+* Maria Kraynyuk, Intel
+* Masahiro Doteguchi, Fujitsu
+* Nikolay Petrov, Intel
+* Pavel Kumbrasev, Intel
+* Penporn Koanakatool, Google
+* Ragesh Hajela, Fujitsu
+* Robert Cohn, Intel
+* Sarah Knepper, Intel
+* Timmie Smith, Intel
+* Vadim Pirogov, Intel
+
+Agenda:
+-------
+
+* Present priorities for the Open Source Working Group
+* Discuss issues and areas for work
+
+Notes
+-----
+
+The first meeting of the UXL Foundation Open Source Working Group was
+led by Andrew Wafaa from Arm.
+
+The group went through a set of priorities that have been provided by
+the UXL Steering Committee.
+
+* Build Infrastructure
+  * The intention of this is to bring independent builds of the
+    projects and infrastructure in place that facilitates community
+    contributions to the projects.
+
+* Open Source Best Practice
+  * The intention of this is to ensure that the projects follow the
+    best practices of open source development so that there is a clear
+    and frictionless way for the community to contribute to the
+    projects.
+
+* Open Source Kernels
+  * The intention of this is to bring more open source code to the
+    projects where there are opportunities to add new kernel code
+    particularly to help target more vendors and architectures.
+
+* Architecture
+  * Identify opportunities to simplify projects or make it easier to
+    support more vendor and architecture targets and also to simplify
+    the integration of the projects with other frameworks and
+    software, for example TensorFlow or PyTorch.
+
+* Compatibility Testing
+  * In order to ensure that a developer using the projects has a
+    consistent and good experience, compatibility tests could be made
+    available so that they can be run on different vendors and
+    architectures.
+
+
+* Build Infrastructure Discussion
+  * Due to the history of the projects the infrastructure is mostly
+    hosted by Intel but we would like to integrate CI, build and
+    testing for different vendor and architecture targets.  Penporn
+    from Google shared information about how this is done with the
+    TensorFlow project:
+    * `Official Builds (maintained by Google)`_
+    * `Community Builds (maintained by 3rd parties)`_
+
+A similar model could be adopted for UXL Foundation projects.  Options
+for using public cloud infrastructure for the projects include:
+
+* The foundation funding infrastructure (this is unlikely due to the
+  costs involved)
+* Obtaining donations of credits from public cloud infrastructure 
+  providers (note that donations cannot be used in lieu of membership cost)
+* Organisations providing infrastructure in a similar way to how Intel
+  currently does this
+
+Some points to note from the discussions:
+
+* TensorFlow has a dedicated team looking after build and CI
+* The build machines needed is dictated by the volume of Pull Requests
+  made so can increase as contributions increase otherwise the process
+  can get backed up
+* Some organisations require security scanning when builds are done,
+  some of these would be difficult to do in public
+* A big improvement for projects would be to move infrastructure or
+  build logs in public so that the community can see this, but some
+  infrastructure can still live inside the corporate network
+* Here is a link to an example of the infrastructure TensorFlow uses
+  `with Arm`_
+* The best solution is to find maintainers for specific target devices
+  and enable them to do this through the GitHub projects
+
+Penporn passed on the details of some Intel people working on CI for
+TensorFlow who may have some advice or help.
+
+Ragesh from Fujitsu is engaging with the oneDAL team to add an Arm
+target. They have discussed modifications and changes needed to make
+the external contributions work and it will be good to share this with
+other projects.
+
+Discussions will continue on the Slack channel and at the next Working
+Group meeting.
+
+**The group agreed that the Working Group will initially focus on
+specific projects, with oneDAL and oneDNN the focus for the next
+meeting.**
+
+.. _`Official Builds (maintained by Google)`: https://github.com/tensorflow/tensorflow?tab=readme-ov-file#official-builds
+.. _`Community Builds (maintained by 3rd parties)`: https://github.com/tensorflow/build#community-supported-tensorflow-builds
+.. _ `with Arm`: https://github.com/tensorflow/tensorflow/actions/workflows/arm-ci.yml

meetings/2024-03-26.rst

@@ -0,0 +1,67 @@
+===========================
+ Open Source WG: 3/26/2024
+===========================
+
+John M. chaired the meeting. Andy could not attend because of travel.
+
+Security: Roman Z.
+==================
+
+* Roman gave an update on security
+* Group has formed and met once
+* There is a `security slack channel`_
+* Roman has created a github project: `security work package`_
+* Work is in progress
+* Deciding how to track (GitHub vs Jira)
+* Concerns about having the bandwidth to get this work done:
+
+  * Roman will make some work estimations and set deadlines
+
+.. _`security slack channel`: https://uxlfoundation.slack.com/archives/C06G978GAHG
+.. _`security work package`: https://github.com/orgs/uxlfoundation/projects/3
+
+Open Source recommendations: Rod B.
+===================================
+
+* Rod had a follow-up discussion about the findings from looking at
+  the `Red Hat open source checklist`_.
+  * `Summary spreadsheet`_
+  * `oneMKL report`_
+  * `oneDPL report`_
+  * `oneDNN report`_
+  * `oneCCL report`_
+  * `oneDAL report`_
+
+* Q: More discussion about private vs public discussion.
+  * Prioritize the kinds of information that other developers will
+    need to know. Look at oneDNN RFC's as an example.
+
+* Need to have some deadlines. Rod has created a github project: `open
+  source checklist work package`_
+
+.. _`open source checklist work package`: https://github.com/orgs/uxlfoundation/projects/5/views/1?layout=board
+.. _`Red Hat open source checklist`: https://www.redhat.com/en/resources/open-source-project-health-checklist
+.. _`Summary spreadsheet`: https://docs.google.com/spreadsheets/d/1UgOGCpskfu1LotfEh1MPwuV-GGdzB2FQQiw0iOd0Cb0
+.. _`oneMKL report`: https://docs.google.com/document/d/1F_WtqzdJuvEKgmBJR9P8nEjmalYLtxmS
+.. _`oneDPL report`: https://docs.google.com/document/d/1l_B_MM0Sk7qbIdjgBBAUgtW7VwmJOBfa
+.. _`oneDNN report`: https://docs.google.com/document/d/1e9vHtVcqWvusIs_U2mP1B8L5DtO5Qb8b
+.. _`oneCCL report`: https://docs.google.com/document/d/1NNK_BrBb0Rz_kcDDprfyuZPyktqePDtY
+.. _`oneDAL report`: https://docs.google.com/document/d/1QRJVhcUv_JKSQmqlb43NkL37n5rliKw4
+
+GitHub Org Transition: Robert C.
+================================
+
+* `GitHub Org Slides`_
+* We have to move GitHub repos to LF/UXLF ownership
+* examined:
+  1. taking oneapi-src
+  2. moving to uxlfoundation
+  3. every project has its own org: https://github.com/onednn-project/onednn
+
+* Recommend option 3
+* See the slides for details
+* There were no strong objections, but we did not have representation
+  from all the projects. Will follow-up with some slides that makes
+  the recommendation clear, and a follow-up discussion.
+
+.. _`GitHub Org Slides`: ../presentations/2024-03-26-GitHub-org.pdf

meetings/notes.rst

@@ -2,111 +2,7 @@
  Meeting Notes
 ===============
 
+`2024-03-26 <2024-03-26.rst>`__
 `2024-02-27 <2024-02-27.rst>`__
-
-1/24/2024
-=========
-
-Attendees:
-----------
-
-* Andrew Wafaa, Arm (Open Source Working Group chair)
-* Abhishek Jain, Fujitsu
-* Alex Pim, Imagination Technologies
-* Alexey Kukanov, Intel
-* Ankit Manerikar, Intel
-* Denis Samoilov, Intel
-* Igor Safanov, Intel
-* Steve Capper, Arm
-* Maria Kraynyuk, Intel
-* Masahiro Doteguchi, Fujitsu
-* Nikolay Petrov, Intel
-* Pavel Kumbrasev, Intel
-* Penporn Koanakatool, Google
-* Ragesh Hajela, Fujitsu
-* Robert Cohn, Intel
-* Sarah Knepper, Intel
-* Timmie Smith, Intel
-* Vadim Pirogov, Intel
-
-Agenda:
--------
-
-* Present priorities for the Open Source Working Group
-* Discuss issues and areas for work
-
-Notes
------
-
-The first meeting of the UXL Foundation Open Source Working Group was led by Andrew Wafaa from Arm.
-
-The group went through a set of priorities that have been provided by the UXL Steering Committee.
-
-* Build Infrastructure
-  * The intention of this is to bring independent builds of the projects and infrastructure in place that facilitates community contributions to the projects.
-
-* Open Source Best Practice
-  * The intention of this is to ensure that the projects follow the best practices of open source development so that there is a clear and frictionless way for the community to contribute to the projects.
-
-* Open Source Kernels
-  * The intention of this is to bring more open source code to the projects where there are opportunities to add new kernel code particularly to help target more vendors and architectures.
-
-* Architecture
-  * Identify opportunities to simplify projects or make it easier to support more vendor and architecture targets and also to simplify the integration of the projects with other frameworks and software, for example TensorFlow or PyTorch.
-
-* Compatibility Testing
-  * In order to ensure that a developer using the projects has a consistent and good experience, compatibility tests could be made available so that they can be run on different vendors and architectures.
-
-
-* Build Infrastructure Discussion
-Due to the history of the projects the infrastructure is mostly hosted by Intel but we would like to integrate CI, build and testing for different vendor and architecture targets.
-Penporn from Google shared information about how this is done with the TensorFlow project:
-
-  * `Official Builds (maintained by Google)`_
-  * `Community Builds (maintained by 3rd parties)`_
-
-A similar model could be adopted for UXL Foundation projects.
-Options for using public cloud infrastructure for the projects include:
-
-* The foundation funding infrastructure (this is unlikely due to the 
-  costs involved)
-* Obtaining donations of credits from public cloud infrastructure 
-  providers (note that donations cannot be used in lieu of membership cost)
-* Organisations providing infrastructure in a similar way to how Intel
-  currently does this
-
-Some points to note from the discussions:
-
-* TensorFlow has a dedicated team looking after build and CI
-* The build machines needed is dictated by the volume of Pull Requests 
-  made so can increase as contributions increase otherwise the process 
-  can get backed up
-* Some organisations require security scanning when builds are done, 
-  some of these would be difficult to do in public
-* A big improvement for projects would be to move infrastructure or 
-  build logs in public so that the community can see this, but some 
-  infrastructure can still live inside the corporate network
-* Here is a link to an example of the infrastructure TensorFlow uses 
-  `with Arm`_
-* The best solution is to find maintainers for specific target devices 
-  and enable them to do this through the GitHub projects
-
-Penporn passed on the details of some Intel people working on CI for 
-TensorFlow who may have some advice or help.
-
-Ragesh from Fujitsu is engaging with the oneDAL team to add an Arm 
-target. They have discussed modifications and changes needed to make 
-the external contributions work and it will be good to share this 
-with other projects.
-
-Discussions will continue on the Slack channel and at the next 
-Working Group meeting.
-
-**The group agreed that the Working Group will initially focus on 
-specific projects, with oneDAL and oneDNN the focus for the next 
-meeting.**
-
-.. _`Official Builds (maintained by Google)`: https://github.com/tensorflow/tensorflow?tab=readme-ov-file#official-builds
-.. _`Community Builds (maintained by 3rd parties)`: https://github.com/tensorflow/build#community-supported-tensorflow-builds
-.. _ `with Arm`: https://github.com/tensorflow/tensorflow/actions/workflows/arm-ci.yml
+`2024-01-24 <2024-01-24.rst>`__