Improper placement of figures and use of colors for text

[wgropp]

Problem

There is an inline figure, which is both against the document style and is a problem waiting to happen - it depends on there being enough space on the page for the figure. This needs to use a figure, as is used everywhere else in the document.

Proposal

In the Topology chapter, the following needs to be replaced:

On a Cartesian virtual grid, the buffer usage in a given 
direction \mpicode{d} with \mpicode{dims[d]==3} and \mpicode{1},
respectively during creation of the communicator is described 
in the following figure. 

\vspace*{2mm}
\centerline{\includegraphics[width=6.0in]{figures/cart_neighbor_allgather}}
\vspace*{1mm}%

and

For a halo communication on a Cartesian grid, the buffer usage
in a given direction \mpicode{d} with  \mpicode{dims[d]==3} and  \mpicode{1}, 
respectively during creation of the communicator is described
in the following figure. 

\vspace*{2mm}
\centerline{\includegraphics[width=6.0in]{figures/cart_neighbor_alltoall}}
\vspace*{1mm}%

Instead, the graphics image must appear in a separate figure environment, with a caption and a \label, and the text in the example must use \ref to refer to the image.

Additionally, the code example includes of an inappropriate use of color. This issue is also addressed by the second commit in the pull request.

This LaTeX code fails the make checklatex test, and causes all build checks to fail.

Changes to the Text

Make the figures into figures, as is the style of the MPI standard.

Impact on Implementations

None

Impact on Users

None

References

mpi-report-issue320-PR312.pdf
See dashed arrows in Figures 8.2 and 8.3 on pages 383 + 386 (pdf 423 + 436) and modified text in the paranthesis on page 387 (pdf 427, lines 7-9)

Pull request

https://github.com/mpi-forum/mpi-standard/pull/304 and https://github.com/mpi-forum/mpi-standard/pull/312

[RolfRabenseifner]

Dear Bill,
I hope that I correctly solved both problems. I therefore updated title of both issue and PR and abstract of the issue.
I substituted the use of colors for the sendrecv calls by an additional sentence:

The first call to \mpicode{MPI_Sendrecv} implements the upper (blue) communication pattern in each diagram of Figure~\ref{fig:cart-neighbor-alltoall}, whereas the second call is for the lower (red) pattern.

I hope that now, both problems are resolved.
Best regards
Rolf

[dholmes-epcc-ed-ac-uk]

Posted without comment:
https://www.nature.com/articles/s41467-020-19160-7
https://phys.org/news/2020-10-colours-science.html

[RolfRabenseifner]

pdf of the current version:
mpi40-report_Issue320-PR304_v03.pdf
I plan to substitute the "blue" and "red" by "solid arrows" and "dashed arrows" in text and figures.

[RolfRabenseifner]

I added https://github.com/mpi-forum/mpi-standard/pull/312 to resolve the color problem by using solid and dashed arrows.
For consistency, I did it in both Figures 8.3 and 8.2 although the problem arised only in Fig. 8.3.
For simplicity, I directly decided to used dashed arrows (and not dotted). It seems to works fine.

[RolfRabenseifner]

mpi-report-issue320-PR312.pdf
See dashed arrows in Figures 8.2 and 8.3 on pages 383 + 386 (pdf 423 + 436) and modified text in the paranthesis on page 387 (pdf 427, lines 7-9)

[RolfRabenseifner]

Bill, something is wrong with the function index: The just produced (through the Azure pipline)
mpi-report-issue320-PR312.pdf
has lost most of the references in the function index.

[wgropp]

I'll have a look