Merge branch 'master' into pr/ib-ops

content/backmatter.tex

@@ -665,6 +665,14 @@ \section{Version 1.6}
     \FUNC{shmem\_ibput}.
 \ChangelogRef{subsec:shmem_ibget, subsec:shmem_ibput}%
 %
+\item Added \FUNC{shmem\_signal\_add} and \FUNC{shmem\_signal\_set} to
+  update a remote flag without associated data transfer of a put-with-signal operation.
+\ChangelogRef{subsec:shmem_signal_add, subsec:shmem_signal_set}%
+%
+\item Clarified that \OPR{Fence} operations only guarantee ordering for
+    operations that are performed on the same context.
+\ChangelogRef{subsec:shmem_fence}%
+%
 \item Added a team-based pointer query routine:
   \FUNC{shmem\_team\_ptr}.
 \ChangelogRef{subsec:shmem_team_ptr}%
@@ -691,7 +699,6 @@ \section{Version 1.6}
     2 to enable profiling with profile library defined effects and
     additional arguments.
 \ChangelogRef{subsec:shmem_pcontrol}
->>>>>>> master
 %
 \end{itemize}
 

content/shmem_fence.tex

@@ -24,7 +24,7 @@
     \ac{PE} on the given context prior
     to the call to \FUNC{shmem\_fence} are guaranteed to be delivered before any
     subsequent operations on symmetric data
-    objects to the same \ac{PE}. \FUNC{shmem\_fence} guarantees order of delivery,
+    objects to the same \ac{PE} on the same context. \FUNC{shmem\_fence} guarantees order of delivery,
     not completion. It does not guarantee order of delivery of nonblocking
     \GET{} or values fetched by nonblocking \ac{AMO} routines.
     If \VAR{ctx} has the value \CONST{SHMEM\_CTX\_INVALID}, no operation is

content/shmem_reductions.tex

@@ -283,9 +283,11 @@ \subsubsubsection{PROD}
     provides one element for each reduction.  The results of the reductions are placed in the
     \dest{} array on all \acp{PE} participating in the reduction.
 
+    The same \source{} and \dest{} arrays must be passed by all PEs that
+    participate in the collective.
     The \source{} and \dest{} arguments must either be the same symmetric
     address, or two different symmetric addresses corresponding to buffers that
-    do not overlap in memory. That is, they must be completely overlapping or
+    do not overlap in memory. That is, they must be completely overlapping (sometimes referred to as an ``in place'' reduction) or
     completely disjoint.
 
     Team-based reduction routines operate over all \acp{PE} in the provided team argument. All

content/shmem_signal_add.tex

@@ -0,0 +1,47 @@
+\apisummary{
+  Adds to a signal value of a remote data object.
+}
+
+\begin{apidefinition}
+
+\begin{C11synopsis}
+void @\FuncDecl{shmem\_signal\_add}@(shmem_ctx_t ctx, const uint64_t *sig_addr, uint64_t signal, int pe);
+\end{C11synopsis}
+
+\begin{Csynopsis}
+void @\FuncDecl{shmem\_signal\_add}@(const uint64_t *sig_addr, uint64_t signal, int pe);
+void @\FuncDecl{shmem\_ctx\_signal\_add}@(shmem_ctx_t ctx, const uint64_t *sig_addr, uint64_t signal, int pe);
+\end{Csynopsis}
+
+\begin{apiarguments}
+  \apiargument{IN}{ctx}{
+    A context handle specifying the context on which to perform the
+    operation. When this argument is not provided, the operation is
+    performed on the default context.
+  }
+  \apiargument{OUT}{sig\_addr}{
+    Symmetric address of the signal data object to be updated on the
+    remote \ac{PE}.
+  }
+  \apiargument{IN}{signal}{
+    Unsigned 64-bit value that is used for updating the remote
+    \VAR{sig\_addr} signal data object.
+  }
+  \apiargument{IN}{pe}{
+    \ac{PE} number of the remote \ac{PE}.
+  }
+\end{apiarguments}
+
+\apidescription{
+  \FUNC{shmem\_signal\_add} adds \VAR{value} to the signal data
+  object pointed to by \VAR{sig\_addr} on \ac{PE}~\VAR{pe}.
+  The update to the \VAR{sig\_addr} signal object at the calling
+  \ac{PE} is expected to satisfy the atomicity guarantees as described
+  in Section~\ref{subsec:signal_atomicity}.
+}
+
+\apireturnvalues{
+  None.
+}
+
+\end{apidefinition}

content/shmem_signal_set.tex

@@ -0,0 +1,47 @@
+\apisummary{
+  Sets the signal value of a remote data object.
+}
+
+\begin{apidefinition}
+
+\begin{C11synopsis}
+void @\FuncDecl{shmem\_signal\_set}@(shmem_ctx_t ctx, const uint64_t *sig_addr, uint64_t signal, int pe);
+\end{C11synopsis}
+
+\begin{Csynopsis}
+void @\FuncDecl{shmem\_signal\_set}@(const uint64_t *sig_addr, uint64_t signal, int pe);
+void @\FuncDecl{shmem\_ctx\_signal\_set}@(shmem_ctx_t ctx, const uint64_t *sig_addr, uint64_t signal, int pe);
+\end{Csynopsis}
+
+\begin{apiarguments}
+  \apiargument{IN}{ctx}{
+    A context handle specifying the context on which to perform the
+    operation. When this argument is not provided, the operation is
+    performed on the default context.
+  }
+  \apiargument{OUT}{sig\_addr}{
+    Symmetric address of the signal data object to be updated on the
+    remote \ac{PE}.
+  }
+  \apiargument{IN}{signal}{
+    Unsigned 64-bit value that is used for updating the remote
+    \VAR{sig\_addr} signal data object.
+  }
+  \apiargument{IN}{pe}{
+    \ac{PE} number of the remote \ac{PE}.
+  }
+\end{apiarguments}
+
+\apidescription{
+  \FUNC{shmem\_signal\_set} writes \VAR{value} into the signal data
+  object pointed to by \VAR{sig\_addr} on \ac{PE}~\VAR{pe}.
+  The update to the \VAR{sig\_addr} signal object at the calling
+  \ac{PE} is expected to satisfy the atomicity guarantees as described
+  in Section~\ref{subsec:signal_atomicity}.
+}
+
+\apireturnvalues{
+  None.
+}
+
+\end{apidefinition}

main_spec.tex

@@ -304,13 +304,17 @@ \subsubsubsection{\textbf{SHMEM\_ATOMIC\_FETCH\_XOR\_NBI}}
 
 \subsection{Signaling Operations}\label{sec:shmem_signal}
 This section specifies the OpenSHMEM support for \OPR{put-with-signal},
-nonblocking \OPR{put-with-signal}, and \OPR{signal-fetch} routines. The
+nonblocking \OPR{put-with-signal}, and \OPR{signal-\{add, fetch, set\}} routines. The
 put-with-signal routines provide a method for copying data from a contiguous
 local data object to a data object on a specified \ac{PE} and subsequently
-updating a remote flag to signal completion. The signal-fetch routine provides
-support for fetching a signal update operation.
-
-\openshmem \OPR{put-with-signal} routines specified in this section have two
+updating a remote flag to signal completion.
+The signal-add and signal-set routines provide methods for updating
+the signal object without the associated data transfer of a
+put-with-signal operation.
+The signal-fetch routine provides support for reading a local signal value.
+
+\openshmem \OPR{put-with-signal} and \OPR{signal-\{add, set\}}
+routines specified in this section have two
 variants. In one of the variants, the context handle, \VAR{ctx}, is explicitly
 passed as an argument. In this variant, the operation is performed on the
 specified context. If the context handle \VAR{ctx} does not correspond to a
@@ -321,16 +325,20 @@ \subsection{Signaling Operations}\label{sec:shmem_signal}
 \subsubsection{Atomicity Guarantees for Signaling Operations}
 \label{subsec:signal_atomicity}
 All signaling operations put-with-signal, nonblocking put-with-signal, and
-signal-fetch are performed on a signal data object, a remotely accessible
+signal-\{add, fetch, set\} are performed on a signal data object, a remotely accessible
 symmetric object of type \VAR{uint64\_t}. A signal operator in the
-put-with-signal routine is a \openshmem library constant that determines the
+put-with-signal routine is an \openshmem library constant that determines the
 type of update to be performed as a signal on the signal data object.
 
-All signaling operations on the signal data object completes as if performed
+All signaling operations on the signal data object complete as if performed
 atomically with respect to the following:
 \begin{itemize}
     \item other blocking or nonblocking variant of the put-with-signal routine
     that updates the signal data object using the same signal update operator;
+    \item signal-add routine when the put-with-signal routine uses the
+      \LibConstRef{SHMEM\_SIGNAL\_ADD} signal operator;
+    \item signal-set routine when the put-with-signal routine uses the
+      \LibConstRef{SHMEM\_SIGNAL\_SET} signal operator;
     \item signal-fetch routine that fetches the signal data object; and
     \item any point-to-point synchronization routine that accesses the signal
     data object.
@@ -360,9 +368,15 @@ \subsubsection{\textbf{SHMEM\_PUT\_SIGNAL}}\label{subsec:shmem_put_signal}
 \subsubsection{\textbf{SHMEM\_PUT\_SIGNAL\_NBI}}\label{subsec:shmem_put_signal_nbi}
 \input{content/shmem_put_signal_nbi.tex}
 
+\subsubsection{\textbf{SHMEM\_SIGNAL\_ADD}}\label{subsec:shmem_signal_add}
+\input{content/shmem_signal_add.tex}
+
 \subsubsection{\textbf{SHMEM\_SIGNAL\_FETCH}}\label{subsec:shmem_signal_fetch}
 \input{content/shmem_signal_fetch.tex}
 
+\subsubsection{\textbf{SHMEM\_SIGNAL\_SET}}\label{subsec:shmem_signal_set}
+\input{content/shmem_signal_set.tex}
+
 
 
 \subsection{Collective Routines}\label{subsec:coll}