<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class="">The issue in the past was that sufficient information to implement this was not specified in Fortran - it required special knowledge of how each Fortran compiler implemented arrays. I admit to being out-of-date on the current Fortran standard. Is this now possible within Fortran? Or would we have to create a new requirement for MPI implementations that the Fortran interface would need to know implementation details of the Fortran compiler(s) used. That might be ok, but it would be a big change and something that we’d need to be explicit about.<div class=""><br class=""></div><div class="">Bill</div><div class=""><br class=""><div class="">
<div dir="auto" style="caret-color: rgb(0, 0, 0); color: rgb(0, 0, 0); letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px; -webkit-text-stroke-width: 0px; text-decoration: none; word-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;" class=""><div>William Gropp<br class="">Director, NCSA<br class="">Thomas M. Siebel Chair in Computer Science</div><div>University of Illinois Urbana-Champaign</div><div>IEEE-CS President-Elect</div><div class=""><br class=""></div></div><br class="Apple-interchange-newline"><br class="Apple-interchange-newline">
</div>
<div><br class=""><blockquote type="cite" class=""><div class="">On Nov 29, 2021, at 4:51 AM, Jeff Hammond via mpiwg-fortran <<a href="mailto:mpiwg-fortran@lists.mpi-forum.org" class="">mpiwg-fortran@lists.mpi-forum.org</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class=""><font face="monospace" class="">Recently, I have been writing mpi4py and Fortran 2008 MPI code (<a href="https://urldefense.com/v3/__https://github.com/ParRes/Kernels/pull/592__;!!DZ3fjg!oEX4-LRl7DOwUqMofwtH6VqvIkhOE2e-H0BiyHGjui1-iLLBsZi48CHIySt_ijPKBQ$" class="">https://github.com/ParRes/Kernels/pull/592</a>), which ends up looking quite similar except for 0-1 base indexing and MPI argument deduction.</font><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">Numpy arrays behave a lot like Fortran arrays, including how they store size information in them.<br class=""></font><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">I wonder if it is reasonable to add this same argument inference to MPI Fortran. If I pass an array argument with no type or size information, it should be inferred.</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">The first inference is type. There is no reason to ask users to specify MPI_DOUBLE_PRECISION when the argument is of type double precision. Obviously, this only works for built-in types, but as that is the common case, why not do it?</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">The second inference is size. If I pass A(100) to MPI_Bcast, why do I need to say MPI_Bcast(buf=A,count=100,...)? The dope vector for A contains the 100 already.</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">The hard part here seems to be needing 15 dimensions worth of interfaces, but those are trivial to generate.</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">Are there any hard problems here that I don't realize?</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">Thanks,</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">Jeff</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><font face="monospace" class="">PS code excerpts from the link above. Named arguments would make Fortran even more similar.</font></div><div class=""><font face="monospace" class=""><br class="">for phase in range(0,np):<br class=""> recv_from = (me + phase ) % np<br class=""> send_to = (me - phase + np) % np<br class=""> lo = block_order * send_to<br class=""> hi = block_order * (send_to+1)<br class=""> comm.Sendrecv(sendbuf=A[lo:hi,:],dest=send_to,sendtag=phase,recvbuf=T,source=recv_from,recvtag=phase)<br class=""> lo = block_order * recv_from<br class=""> hi = block_order * (recv_from+1)<br class=""> B[lo:hi,:] += T.T</font><div class=""><font face="monospace" class=""><br class=""></font></div><div class=""><div style="margin: 0cm;" class=""><font face="monospace" class="">do q=0,np-1</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> recv_from =
mod( (me + q ), np)</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> send_to = mod( (me - q + np), np)</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> lo =
block_order * send_to + 1</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> hi =
block_order * (send_to+1)</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> call
MPI_Sendrecv(A(:,lo:hi), block_order*block_order, MPI_DOUBLE_PRECISION, &</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> send_to,q, &</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> T,block_order*block_order, MPI_DOUBLE_PRECISION, &</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> recv_from, q, MPI_COMM_WORLD, MPI_STATUS_IGNORE)</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> lo =
block_order * recv_from + 1</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> hi =
block_order * (recv_from+1)</font></div><div style="margin: 0cm;" class=""><font face="monospace" class=""> B(:,lo:hi) =
B(:,lo:hi) + transpose(T)</font></div><div class=""><font face="monospace" class=""><br class=""></font></div><font face="monospace" class="">-- <br class=""></font><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><font face="monospace" class="">Jeff Hammond<br class=""><a href="mailto:jeff.science@gmail.com" target="_blank" class="">jeff.science@gmail.com</a><br class=""><a href="https://urldefense.com/v3/__http://jeffhammond.github.io/__;!!DZ3fjg!oEX4-LRl7DOwUqMofwtH6VqvIkhOE2e-H0BiyHGjui1-iLLBsZi48CHIyStfTmaCwQ$" target="_blank" class="">http://jeffhammond.github.io/</a></font></div></div></div></div></div>
_______________________________________________<br class="">mpiwg-fortran mailing list<br class=""><a href="mailto:mpiwg-fortran@lists.mpi-forum.org" class="">mpiwg-fortran@lists.mpi-forum.org</a><br class="">https://urldefense.com/v3/__https://lists.mpi-forum.org/mailman/listinfo/mpiwg-fortran__;!!DZ3fjg!oEX4-LRl7DOwUqMofwtH6VqvIkhOE2e-H0BiyHGjui1-iLLBsZi48CHIySsKbricRQ$ <br class=""></div></blockquote></div><br class=""></div></body></html>