doc/mansrc: Added semantic descriptions for functions

[KLeoUIC]

Pull Request Description

Currently documentation is very minimal or boilerplate and aimed at developers and those familiar with MPI. It leaves a gap for those learning MPI through MPICH and users looking to orient themselves in the large standard. As MPIX functions are not in the standard yet users have no easy way of looking up their usage beyond the one sentence description manpages have currently. The generated descriptions provide a way of getting more information for users who do not wish to scour the codebase or search for papers where the function was proposed.

Pull request provides semantic descriptions based off of the latest MPI Standard and source code.
Directions and script for making new descriptions in Box.

Author Checklist

• Provide Description
  Particularly focus on why, not what. Reference background, issues, test failures, xfail entries, etc.
• Commits Follow Good Practice
  Commits are self-contained and do not do two things at once.
  Commit message is of the form: module: short description
  Commit message explains what's in the commit.
• Passes All Tests
  Whitespace checker. Warnings test. Additional tests via comments.
• Contribution Agreement
  For non-Argonne authors, check contribution agreement.
  If necessary, request an explicit comment from your companies PR approval manager.

[hzhou]

test:mpich/doc

[hzhou]

test:mpich/doc

[sonjahapp]

I reviewed text for MPI Session and spawn procedures. Text looks good to me except for the single comment I made on MPI_Comm_spawn.

[sonjahapp]

Suggested change

**MPI_Win_create** is a collective operation over the group of `comm` that creates a window object for remote memory access operations. Each process specifies a region of existing memory starting at `base` of `size` bytes that is exposed for RMA accesses by other processes in the group. The call returns an opaque window handle that represents the group of processes and the attributes of each window. A process may expose no memory by specifying `size` = 0. The `disp_unit` argument facilitates address arithmetic in RMA operations by scaling target displacement arguments. The `info` argument provides optimization hints including `no_locks`, `accumulate_ordering`, `accumulate_ops`, `same_size`, and `same_disp_unit`. Concurrent communications to distinct overlapping windows may lead to undefined results. Implementations may make the provided memory available for load/store accesses

This function is collective over `comm` and may not return until MPI_INIT has been called in the children. MPI_INIT in the children may not return until all parents have called MPI_Comm_spawn, forming a collective operation over the union of parent and child processes. The spawned children have their own `MPI_COMM_WORLD`, which is separate from that of the parents. The returned inter-communicator contains parent processes in the local group and child processes in the remote group. If MPI is unable to spawn `maxprocs` processes, it raises an error of class `MPI_ERR_SPAWN`. An implementation may allow the `info` argument with the `soft` key to spawn a smaller number of processes instead of raising an error; the number of spawned processes is given by the size of the remote group of `intercomm`. Processes spawned at the same time using MPI_COMM_SPAWN_MULTIPLE are placed in the same `MPI_COMM_WORLD`. The constant `MPI_ERRCODES_IGNORE` may be passed in `array_of_errcodes` if error codes are not needed.

I find the last sentence here confusing and suggest to remove it.

[sonjahapp]

Not sure what gitlab did here with my suggestion 🙈 For clarity: I mean to suggest to remove the last sentence of the text for MPI_Comm_spawn about performance.