-
Notifications
You must be signed in to change notification settings - Fork 1
2026 01 12 webex
AI can make mistakes. Review for accuracy.
The team discussed preparations for the upcoming March forum, including a plenary reading on dynamic process sets and a proposal presentation that may attract attention. They reviewed updates to the MPI standard text and documentation, focusing on improving clarity and consistency across different MPI objects. The group also addressed the status of tools-related work and planned to revisit various topics in March, with Dominik preparing a slide set on the dictionary approach for dynamic resources. Next steps
* Sonja: Contact Dominik and Martin to discuss compatibility between spawn proposal and dynamic process sets
* Sonja: Prepare a draft proposal ready for a Sessions working group meeting in February
* Sonja: Let Howard know once draft is ready to bring up in February working group meeting
* Dominik: Check recording of previous virtual forum presentation
* Dominik: Prepare slide set for dictionary presentation to discuss in next Sessions meetings for potential March forum presentation
* Sonja: Include in dictionary presentation how it relates to existing MPI publish/lookup name functions and clarify advantages
* Howard: Remove "has no side effects" phrase from pull request text based on Sonja's feedback
* Howard: Check with Wes whether pull request should be errata or full proposal and add appropriate changelog
* Howard: Follow up on tools group discussion, potentially ping Bill Williams from Dresden about arranged meetings between tools people and Sessions group
* Someone : Read Howard's pull request addressing Sessions attributes feedback at March forum
Howard discussed the upcoming March forum and noted his absence, mentioning difficulties in arranging someone from his institution to attend. He suggested having a plenary reading on dynamic process sets, pending time availability from Dominik or Martin. Howard also mentioned a pull request he had opened to address issues from a previous attributes proposal and asked if there were other preparations needed for the March forum.
Sonja discussed her work (MPI_ISession_spawn) on a proposal and mentioned a story that might be presented at a forum, attracting unexpected attention. She expressed concerns about the complexity of the standard chapter and her need for feedback. Howard suggested considering the compatibility of Sonja's proposal with Dominik and Martin's proposal, noting that they could be treated as orthogonal if necessary.
Dominik plans to prepare a draft proposal for the working group meeting in February to discuss dynamic processes and potentially present it at the forum plenary. Dominik notes that a previous virtual forum discussion on the topic was cut short due to time constraints and lack of detailed feedback. Howard suggests presenting the proposal in full at the forum to allow for more comprehensive discussion and greater attendance. Dominik agrees to review the recording of the previous presentation to confirm details and may seek further input from Wes. Dictionary Approach for Dynamic Resources
The group discussed the status of tools-related work, noting that previous requests to associate sessions with various objects had not been addressed due to a lack of time. They agreed to revisit the topic in March, with Dominik offering to prepare a slide set on the dictionary approach for dynamic resources. Sonja suggested highlighting the advantages of the dictionary method over existing MPI functions like publish/lookup, as it may not be necessary for implementing spawning. MPI Standard Text Updates
Howard presented updates to the MPI standard text, addressing feedback from pull request #1125. He removed the adjective "simple" when describing default copy and delete attribute functions, and replaced phrases like "does nothing other than returning" with more precise technical descriptions. Howard also made these changes symmetric across different MPI objects (communicators, types, and windows) to maintain consistency in the standard. MPI Documentation Language Simplification
Howard and Sonja discussed a proposed change to MPI documentation, focusing on how to describe functions with no side effects. They agreed to remove the "has no side effects" phrase and simplify the language, with Sonja suggesting that the MPI common null copy function's description could be made clearer. Howard mentioned he would consult with Wes about whether the change should be an errata or a full proposal, as this would affect where the changelog should be placed. Dynamic Process Chapter Updates
The team discussed updates needed for the dynamic process chapter of the standard, noting inconsistencies in writing style and outdated references. Howard agreed to follow up on Sonja's suggestions, including checking for mentions of an IBM SP switch. The group reviewed the status of Dominik's slide deck focused on starting with the dictionary, which Sonja and Dominik will prepare for the next meeting. They confirmed their next meeting would be held on the 26th, with all participants agreeing to the time slot.