<div dir="ltr">IIRC, per-window ordering was the intent of the working group and it also seems like the only sensible reading of the chapter, given that we allow each window to provide a different value for the ordering info key.  This might just be an erratum, but it would be helpful to create a ticket for it.</div><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Aug 10, 2015 at 10:04 AM, Jeff Hammond <span dir="ltr"><<a href="mailto:jeff.science@gmail.com" target="_blank">jeff.science@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">
                
        
        
                <div title="Page 493">
                        <div>
                                <div>
                                        <p>
                
        
        
                </p><div title="Page 493">
                        <div>
                                <div>
                                        <p><span style="font-size:12.000000pt;font-family:'CMSS12'">"11.7.2 Ordering" says: </span></p>
                                </div>
                        </div>
                </div><p><span style="font-size:11.000000pt;font-family:'CMSS10'">"MPI </span><span style="font-size:11.000000pt;font-family:'CMR10'">specifies ordering between accumulate operations from one process to the same (or
overlapping) memory locations at another process on a per-datatype granularity."</span></p>
                                </div>
                        </div>
                </div><div>This seems to say that ordering is required between overlapping windows, which is a horrible semantic.  We need to add a "per-window" caveat here.</div><div><br></div><div>Later, we talk about overlapping memory regions, rather than locations, which is confusing:</div><div>
                
        
        
                <div title="Page 493">
                        <div>
                                <div>
                                        <p><span style="font-size:11.000000pt;font-family:'CMSS10'">"MPI </span><span style="font-size:11.000000pt;font-family:'CMR10'">does not provide any guarantees
for accesses or updates from different origin processes to overlapping target memory regions."</span></p>
                                </div>
                        </div>
                </div></div><div>What is a memory region anyways?  Is it a window?  Is it a set of memory locations?</div><div><br></div><div>And to be clear, I think I know what we mean in the RMA chapter, but others may not.</div><span class="HOEnZb"><font color="#888888"><div><br></div><div>Jeff</div><div><br></div>-- <br><div>Jeff Hammond<br><a href="mailto:jeff.science@gmail.com" target="_blank">jeff.science@gmail.com</a><br><a href="http://jeffhammond.github.io/" target="_blank">http://jeffhammond.github.io/</a></div>
</font></span></div>
<br>_______________________________________________<br>
mpiwg-rma mailing list<br>
<a href="mailto:mpiwg-rma@lists.mpi-forum.org">mpiwg-rma@lists.mpi-forum.org</a><br>
<a href="http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma" rel="noreferrer" target="_blank">http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma</a><br></blockquote></div><br></div>