<html dir="ltr">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<style id="owaParaStyle" type="text/css">P {margin-top:0;margin-bottom:0;}</style>
</head>
<body ocsi="0" fpstyle="1">
<div style="direction: ltr;font-family: Tahoma;color: #000000;font-size: 10pt;">Hi,<br>
<br>
it's quite unclear what Page 410, Lines 17-19<br>
<br>
A consistent view can be created in the uni?fied<br>
memory model (see Section 11.4) by utilizing the window synchronization functions (see<br>
Section 11.5)<br>
<br>
really means. Section 11.5 doesn't mention any (load/store) access to shared memory.<br>
Thus, must <br>
<br>
(*) RMA communication calls and RMA operations<br>
be interpreted as RMA communication calls (MPI_GET, MPI_PUT, ...) and <br>
ANY load/store access to shared window<br>
(*) put call as put call and any store to shared memory<br>
(*) get call as get call and any load from shared memory<br>
(*) accumulate call as accumulate call and any load or store access to shared window ?<br>
<br>
Example: Assertion MPI_MODE_NOPRECEDE<br>
<br>
Does <br>
<br>
the fence does not complete any sequence of locally issued RMA calls<br>
<br>
mean for windows created by MPI_Win_Allocate_shared ()<br>
<br>
the fence does not complete any sequence of locally issued RMA calls or<br>
any load/store access to the window memory ?<br>
<br>
It's not clear to me. I will be probably not clear for the standard MPI user.<br>
RMA operations are defined only MPI functions for window objects<br>
(as far as I can see).<br>
But possibly I'm totally wrong and the synchronization functions synchronize<br>
only the RMA communication calls (MPI_GET, MPI_PUT, ...).<br>
<br>
Hubert<br>
<br>
</div>
</body>
</html>