<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
<meta name="Generator" content="Microsoft Word 15 (filtered medium)">
<style><!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0cm;
font-size:11.0pt;
font-family:"Calibri",sans-serif;
mso-fareast-language:EN-US;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:#0563C1;
text-decoration:underline;}
.MsoChpDefault
{mso-style-type:export-only;
font-size:10.0pt;}
@page WordSection1
{size:612.0pt 792.0pt;
margin:70.85pt 70.85pt 70.85pt 70.85pt;}
div.WordSection1
{page:WordSection1;}
--></style><!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]-->
</head>
<body lang="FR" link="#0563C1" vlink="#954F72" style="word-wrap:break-word">
<div class="WordSection1">
<p class="MsoNormal"><span lang="EN-US">Dear MPI-forum members,<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">This comment applies to MPI-4.0 standard (<a href="https://www.mpi-forum.org/docs/mpi-4.0/mpi40-report.pdf">https://www.mpi-forum.org/docs/mpi-4.0/mpi40-report.pdf</a>)<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">Having partitioned communication looks very promising to speed up applications, but I am wondering why did you introduce “MPI_Parrived”(page 151) routine instead of making “MPI_Precv_init”(page 147) returning an array
of MPI_Request.<br>
To my understanding, one of the need for partitioned communications is runtimes that exchange set of “tasks” : computations do not immediately need all the partitions to continue. Partitioned communication enabled message aggregation but a single partition
could be sufficient to unlock computation. In this case, processes will have to search over each partition to catch a completed ones that not have been already found. This is what MPI_Waitany (page 116) does with an array of MPI_Request. It would be have been
therefore more interesting to handle explicit MPI_Request objects.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">I would appreciate that MPI_Precv_init returns an array of MPI_request in addition of the main MPI_request (p147 line 39). MPI_Parrived could be deprecated for MPI_Test on these sub-requests. In extension this would also
avoid to duplicate MPI_Testany(p117), MPI_Waitall(p118), MPI_Testall(p119), MPI_Waitsome (p 120) and MPI_Testsome(p121).<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US">In a first idea, MPI users would provide this array of the size of the number of partitions as an extra parmeter to MPI_Precv_init. If the array is valid (for example not a special value “MPI_NO_REQUESTS”) these MPI_Request
would be created by MPI_Precv_init and freed by a call of MPI_Request_free on the main request. Waiting all the sub-requests would be equivalent to wait the main request. Like requests from MPI_Irecv, it would be invalid to call MPI_Start on these sub-requests.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-US"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-US" style="mso-fareast-language:FR">Regards,<o:p></o:p></span></p>
<p class="MsoNormal"><span style="mso-fareast-language:FR">Emmanuel Brelle<o:p></o:p></span></p>
</div>
</body>
</html>