August 2014 - mpiwg-rma - mailman3.cels.anl.gov

Re: [mpiwg-rma] Single RMA synchronization for several window handles
by Rolf Rabenseifner 11 Aug '14

11 Aug '14

Jim and all, It is not syntactic sugar. It is a latency optimizing enhancement: If you are doing neighbor communication with 1-sided communication to 6 neighbors - based on 6 windows that are all defined in MPI_COMM_WORLD, then currently you need to call 6 times the synchronization calls which implies at maximum a 6 times larger latency, e.g., 6 times MPI_Win_fence instead of one call. If you are using MPI on shared memory windows, then the same latency problem exists, but here, the trick with dynamic windows is impossible because shared memory windows must be allocated with MPI_Win_allocate_shared. Of course, it is more complicated to define MPI_Win_combine on a superset of communicators used for all the combined windows. As a first effort, it would be already helpful to define it for the same communicator. To allow enhancements in a future MPI version, I still would recommend to have the comm argument as part of the argument list. Rolf ----- Original Message ----- > From: "Jim Dinan" <james.dinan(a)gmail.com> > To: "MPI WG Remote Memory Access working group" <mpiwg-rma(a)lists.mpi-forum.org> > Sent: Sunday, August 10, 2014 5:48:02 PM > Subject: Re: [mpiwg-rma] Single RMA synchronization for several window handles > > > > Hi Rolf, > > > I had initially proposed this in the context of passive target RMA. > Active target RMA already notifies the receiver when data has > arrived, but there is no efficient way to get such a notification in > passive target RMA. I think what you are proposing here would be > syntactic sugar on top of the existing interface --- could I > implement this by fencing every window to determine that all > transfers are completed? > > > I agree with the comments regarding dynamic windows. The merged > window would contain discontiguous buffers; thus, it would lose the > ability to do offset-based addressing and would need to use absolute > (BOTTOM-based) addressing. > > > ~Jim. > > > > On Fri, Aug 8, 2014 at 10:18 AM, Rolf Rabenseifner < > rabenseifner(a)hlrs.de > wrote: > > > Jim, > > your topic "Reducing Synchronization Overhead Through Bundled > Communication" may get also help if we would be able to > combine several window handles to one superset window handle. > > If you have several windows for different buffers, but > only one synchronization pattern, e.g. MPI_Win_fince > then currently you must call MPI_Win_fence seperately > for each window handle. > > I would propose: > > MPI_Win_combine (/*IN*/ int count, > /*IN*/ MPI_Win *win, > /*IN*/ MPI_Comm comm, > /*OUT*/ MPI_Win *win_combined) > > The process group of comm must contain the process groups of all win. > The resulting window handle win_combined can be used only > in RMA synchronization calls and other helper routines, > but not for dynamic window allocation nor for any > RMA communication routine. > Collective synchronization routines must be called by all processes > of comm. > The semantics of an RMA synchronization call using win_combined > is defined as if the calls were seperately issued for > each window handle of the array win. If group handles > are part of the argument list of the synchronization call > then the appropriate subset is used for each window handle in win. > > What do you think about this idea for MPI-4.0? > > Best regards > Rolf > > ----- Original Message ----- > > From: "Jim Dinan" < james.dinan(a)gmail.com > > > To: "MPI WG Remote Memory Access working group" < > > mpiwg-rma(a)lists.mpi-forum.org > > > Sent: Thursday, August 7, 2014 4:08:32 PM > > Subject: [mpiwg-rma] RMA Notification > > > > > > > > Hi All, > > > > > > I have added a new proposal for an RMA notification extension: > > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 > > > > > > I would like to bring this forward for the RMA WG to consider as an > > MPI-4 extension. > > > > > > Cheers, > > ~Jim. > > _______________________________________________ > > mpiwg-rma mailing list > > mpiwg-rma(a)lists.mpi-forum.org > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > -- > Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de > High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 > University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 > Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner > Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma -- Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307)

1 0

Re: [mpiwg-rma] Single RMA synchronization for several window handles
by Jim Dinan 10 Aug '14

10 Aug '14

Hi Rolf, I had initially proposed this in the context of passive target RMA. Active target RMA already notifies the receiver when data has arrived, but there is no efficient way to get such a notification in passive target RMA. I think what you are proposing here would be syntactic sugar on top of the existing interface --- could I implement this by fencing every window to determine that all transfers are completed? I agree with the comments regarding dynamic windows. The merged window would contain discontiguous buffers; thus, it would lose the ability to do offset-based addressing and would need to use absolute (BOTTOM-based) addressing. ~Jim. On Fri, Aug 8, 2014 at 10:18 AM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > Jim, > > your topic "Reducing Synchronization Overhead Through Bundled > Communication" may get also help if we would be able to > combine several window handles to one superset window handle. > > If you have several windows for different buffers, but > only one synchronization pattern, e.g. MPI_Win_fince > then currently you must call MPI_Win_fence seperately > for each window handle. > > I would propose: > > MPI_Win_combine (/*IN*/ int count, > /*IN*/ MPI_Win *win, > /*IN*/ MPI_Comm comm, > /*OUT*/ MPI_Win *win_combined) > > The process group of comm must contain the process groups of all win. > The resulting window handle win_combined can be used only > in RMA synchronization calls and other helper routines, > but not for dynamic window allocation nor for any > RMA communication routine. > Collective synchronization routines must be called by all processes > of comm. > The semantics of an RMA synchronization call using win_combined > is defined as if the calls were seperately issued for > each window handle of the array win. If group handles > are part of the argument list of the synchronization call > then the appropriate subset is used for each window handle in win. > > What do you think about this idea for MPI-4.0? > > Best regards > Rolf > > ----- Original Message ----- > > From: "Jim Dinan" <james.dinan(a)gmail.com> > > To: "MPI WG Remote Memory Access working group" < > mpiwg-rma(a)lists.mpi-forum.org> > > Sent: Thursday, August 7, 2014 4:08:32 PM > > Subject: [mpiwg-rma] RMA Notification > > > > > > > > Hi All, > > > > > > I have added a new proposal for an RMA notification extension: > > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 > > > > > > I would like to bring this forward for the RMA WG to consider as an > > MPI-4 extension. > > > > > > Cheers, > > ~Jim. > > _______________________________________________ > > mpiwg-rma mailing list > > mpiwg-rma(a)lists.mpi-forum.org > > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > -- > Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de > High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 > University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 > Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner > Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma

1 0

Re: [mpiwg-rma] Single RMA synchronization for several window handles
by Jeff Hammond 08 Aug '14

08 Aug '14

On Fri, Aug 8, 2014 at 7:18 AM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > Jim, > > your topic "Reducing Synchronization Overhead Through Bundled > Communication" may get also help if we would be able to > combine several window handles to one superset window handle. > > If you have several windows for different buffers, but > only one synchronization pattern, e.g. MPI_Win_fince > then currently you must call MPI_Win_fence seperately > for each window handle. > > I would propose: > > MPI_Win_combine (/*IN*/ int count, > /*IN*/ MPI_Win *win, > /*IN*/ MPI_Comm comm, > /*OUT*/ MPI_Win *win_combined) > Let's consider whether the following makes sense: MPI_Comm_combine(/*IN*/ int count, /*IN*/ MPI_Comm comm[], /*OUT*/ MPI_Comm *comm_combined); This function would allow me to have an object to Barrier across multiple communicators at once. Make sense? I didn't think so... I would much rather think about MPI_Win_flush(_all)(_local)v that take a vector of windows. I don't see the need for MPI_Win_fencev. If the implementation can sync all the Windows on the first call, the subsequent calls should be cheap. ... MPI_Put(win1) MPI_Put(win2) MPI_Put(win3) MPI_Win_fence(win1) <<< all three MPI_Put operations are completed here MPI_Win_fence(win2) <<< essentially a no-op MPI_Win_fence(win3) <<< essentially a no-op ... Maybe I am forgetting that MPI_Win_fence requires a Barrier... Jeff > The process group of comm must contain the process groups of all win. > The resulting window handle win_combined can be used only > in RMA synchronization calls and other helper routines, > but not for dynamic window allocation nor for any > RMA communication routine. > Collective synchronization routines must be called by all processes > of comm. > The semantics of an RMA synchronization call using win_combined > is defined as if the calls were seperately issued for > each window handle of the array win. If group handles > are part of the argument list of the synchronization call > then the appropriate subset is used for each window handle in win. > > What do you think about this idea for MPI-4.0? > > Best regards > Rolf > > ----- Original Message ----- >> From: "Jim Dinan" <james.dinan(a)gmail.com> >> To: "MPI WG Remote Memory Access working group" <mpiwg-rma(a)lists.mpi-forum.org> >> Sent: Thursday, August 7, 2014 4:08:32 PM >> Subject: [mpiwg-rma] RMA Notification >> >> >> >> Hi All, >> >> >> I have added a new proposal for an RMA notification extension: >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 >> >> >> I would like to bring this forward for the RMA WG to consider as an >> MPI-4 extension. >> >> >> Cheers, >> ~Jim. >> _______________________________________________ >> mpiwg-rma mailing list >> mpiwg-rma(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > -- > Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de > High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 > University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 > Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner > Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [mpiwg-rma] Single RMA synchronization for several window handles
by Jeff Hammond 08 Aug '14

08 Aug '14

...because dynamic windows require non-scalable metadata and can inhibit performance on networks that require memory registration. Pavan can elaborate. Jeff On Fri, Aug 8, 2014 at 8:06 AM, William Gropp <wgropp(a)illinois.edu> wrote: > Why not simply use MPI_Win_create_dynamic, attach the separate memory > regions, and use that MPI_Win? Adding an option for a collective memory > allocator would make this almost equivalent, and wouldn’t add additional > overhead beyond what is already requried. > > Bill > > William Gropp > Director, Parallel Computing Institute > Thomas M. Siebel Chair in Computer Science > University of Illinois Urbana-Champaign > > > > > > On Aug 8, 2014, at 9:18 AM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > > Jim, > > your topic "Reducing Synchronization Overhead Through Bundled > Communication" may get also help if we would be able to > combine several window handles to one superset window handle. > > If you have several windows for different buffers, but > only one synchronization pattern, e.g. MPI_Win_fince > then currently you must call MPI_Win_fence seperately > for each window handle. > > I would propose: > > MPI_Win_combine (/*IN*/ int count, > /*IN*/ MPI_Win *win, > /*IN*/ MPI_Comm comm, > /*OUT*/ MPI_Win *win_combined) > > The process group of comm must contain the process groups of all win. > The resulting window handle win_combined can be used only > in RMA synchronization calls and other helper routines, > but not for dynamic window allocation nor for any > RMA communication routine. > Collective synchronization routines must be called by all processes > of comm. > The semantics of an RMA synchronization call using win_combined > is defined as if the calls were seperately issued for > each window handle of the array win. If group handles > are part of the argument list of the synchronization call > then the appropriate subset is used for each window handle in win. > > What do you think about this idea for MPI-4.0? > > Best regards > Rolf > > ----- Original Message ----- > > From: "Jim Dinan" <james.dinan(a)gmail.com> > To: "MPI WG Remote Memory Access working group" > <mpiwg-rma(a)lists.mpi-forum.org> > Sent: Thursday, August 7, 2014 4:08:32 PM > Subject: [mpiwg-rma] RMA Notification > > > > Hi All, > > > I have added a new proposal for an RMA notification extension: > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 > > > I would like to bring this forward for the RMA WG to consider as an > MPI-4 extension. > > > Cheers, > ~Jim. > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > > -- > Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de > High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 > University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 > Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner > Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > > > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

Re: [mpiwg-rma] Single RMA synchronization for several window handles
by William Gropp 08 Aug '14

08 Aug '14

If you are going to allow windows with different communicators, then you might as well use separate fence calls. Otherwise, you are introducing a brand-new kind of object - one that involves the union of non-disjoint groups of processes. I don’t think this is a good idea. Bill William Gropp Director, Parallel Computing Institute Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign On Aug 8, 2014, at 10:50 AM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > Dynamic Windows also require that all have the same communicator > (because same win handle).

1 0

Re: [mpiwg-rma] Single RMA synchronization for several window handles
by Rolf Rabenseifner 08 Aug '14

08 Aug '14

Dynamic windows do not provide the easy-to-use window base mechanism, i.e., the origin must know the absolute address on the target process. Same for target_disp_unit. Dynamic Windows also require that all have the same communicator (because same win handle). My proposal would allow full service for all RMA operations (and not only MPI_BOTTOM Quality), full service for Win_create or Win_allocate or Win_allocate_shared and together with combined synchronization. Is this worth to consider? Rolf ----- Original Message ----- > From: "William Gropp" <wgropp(a)illinois.edu> > To: "MPI WG Remote Memory Access working group" <mpiwg-rma(a)lists.mpi-forum.org> > Sent: Friday, August 8, 2014 5:06:14 PM > Subject: Re: [mpiwg-rma] Single RMA synchronization for several window handles > > Why not simply use MPI_Win_create_dynamic, attach the separate memory > regions, and use that MPI_Win? Adding an option for a collective > memory allocator would make this almost equivalent, and wouldn’t add > additional overhead beyond what is already requried. > > Bill > > William Gropp > Director, Parallel Computing Institute Thomas M. Siebel Chair in > Computer Science > University of Illinois Urbana-Champaign > > On Aug 8, 2014, at 9:18 AM, Rolf Rabenseifner < rabenseifner(a)hlrs.de > > wrote: > > > Jim, > > your topic "Reducing Synchronization Overhead Through Bundled > Communication" may get also help if we would be able to > combine several window handles to one superset window handle. > > If you have several windows for different buffers, but > only one synchronization pattern, e.g. MPI_Win_fince > then currently you must call MPI_Win_fence seperately > for each window handle. > > I would propose: > > MPI_Win_combine (/*IN*/ int count, > /*IN*/ MPI_Win *win, > /*IN*/ MPI_Comm comm, > /*OUT*/ MPI_Win *win_combined) > > The process group of comm must contain the process groups of all win. > The resulting window handle win_combined can be used only > in RMA synchronization calls and other helper routines, > but not for dynamic window allocation nor for any > RMA communication routine. > Collective synchronization routines must be called by all processes > of comm. > The semantics of an RMA synchronization call using win_combined > is defined as if the calls were seperately issued for > each window handle of the array win. If group handles > are part of the argument list of the synchronization call > then the appropriate subset is used for each window handle in win. > > What do you think about this idea for MPI-4.0? > > Best regards > Rolf > > ----- Original Message ----- > > > From: "Jim Dinan" < james.dinan(a)gmail.com > > To: "MPI WG Remote Memory Access working group" < > mpiwg-rma(a)lists.mpi-forum.org > > Sent: Thursday, August 7, 2014 4:08:32 PM > Subject: [mpiwg-rma] RMA Notification > > > > Hi All, > > > I have added a new proposal for an RMA notification extension: > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 > > > I would like to bring this forward for the RMA WG to consider as an > MPI-4 extension. > > > Cheers, > ~Jim. > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > -- > Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de > High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 > University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 > Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner > Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma -- Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307)

1 0

Re: [mpiwg-rma] Single RMA synchronization for several window handles
by William Gropp 08 Aug '14

08 Aug '14

Why not simply use MPI_Win_create_dynamic, attach the separate memory regions, and use that MPI_Win? Adding an option for a collective memory allocator would make this almost equivalent, and wouldn’t add additional overhead beyond what is already requried. Bill William Gropp Director, Parallel Computing Institute Thomas M. Siebel Chair in Computer Science University of Illinois Urbana-Champaign On Aug 8, 2014, at 9:18 AM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> wrote: > Jim, > > your topic "Reducing Synchronization Overhead Through Bundled > Communication" may get also help if we would be able to > combine several window handles to one superset window handle. > > If you have several windows for different buffers, but > only one synchronization pattern, e.g. MPI_Win_fince > then currently you must call MPI_Win_fence seperately > for each window handle. > > I would propose: > > MPI_Win_combine (/*IN*/ int count, > /*IN*/ MPI_Win *win, > /*IN*/ MPI_Comm comm, > /*OUT*/ MPI_Win *win_combined) > > The process group of comm must contain the process groups of all win. > The resulting window handle win_combined can be used only > in RMA synchronization calls and other helper routines, > but not for dynamic window allocation nor for any > RMA communication routine. > Collective synchronization routines must be called by all processes > of comm. > The semantics of an RMA synchronization call using win_combined > is defined as if the calls were seperately issued for > each window handle of the array win. If group handles > are part of the argument list of the synchronization call > then the appropriate subset is used for each window handle in win. > > What do you think about this idea for MPI-4.0? > > Best regards > Rolf > > ----- Original Message ----- >> From: "Jim Dinan" <james.dinan(a)gmail.com> >> To: "MPI WG Remote Memory Access working group" <mpiwg-rma(a)lists.mpi-forum.org> >> Sent: Thursday, August 7, 2014 4:08:32 PM >> Subject: [mpiwg-rma] RMA Notification >> >> >> >> Hi All, >> >> >> I have added a new proposal for an RMA notification extension: >> https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 >> >> >> I would like to bring this forward for the RMA WG to consider as an >> MPI-4 extension. >> >> >> Cheers, >> ~Jim. >> _______________________________________________ >> mpiwg-rma mailing list >> mpiwg-rma(a)lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma > > -- > Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de > High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 > University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 > Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner > Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307) > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma

1 0

Single RMA synchronization for several window handles
by Rolf Rabenseifner 08 Aug '14

08 Aug '14

Jim, your topic "Reducing Synchronization Overhead Through Bundled Communication" may get also help if we would be able to combine several window handles to one superset window handle. If you have several windows for different buffers, but only one synchronization pattern, e.g. MPI_Win_fince then currently you must call MPI_Win_fence seperately for each window handle. I would propose: MPI_Win_combine (/*IN*/ int count, /*IN*/ MPI_Win *win, /*IN*/ MPI_Comm comm, /*OUT*/ MPI_Win *win_combined) The process group of comm must contain the process groups of all win. The resulting window handle win_combined can be used only in RMA synchronization calls and other helper routines, but not for dynamic window allocation nor for any RMA communication routine. Collective synchronization routines must be called by all processes of comm. The semantics of an RMA synchronization call using win_combined is defined as if the calls were seperately issued for each window handle of the array win. If group handles are part of the argument list of the synchronization call then the appropriate subset is used for each window handle in win. What do you think about this idea for MPI-4.0? Best regards Rolf ----- Original Message ----- > From: "Jim Dinan" <james.dinan(a)gmail.com> > To: "MPI WG Remote Memory Access working group" <mpiwg-rma(a)lists.mpi-forum.org> > Sent: Thursday, August 7, 2014 4:08:32 PM > Subject: [mpiwg-rma] RMA Notification > > > > Hi All, > > > I have added a new proposal for an RMA notification extension: > https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 > > > I would like to bring this forward for the RMA WG to consider as an > MPI-4 extension. > > > Cheers, > ~Jim. > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma -- Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner(a)hlrs.de High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530 University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832 Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307)

1 0

RMA Notification
by Jim Dinan 07 Aug '14

07 Aug '14

Hi All, I have added a new proposal for an RMA notification extension: https://svn.mpi-forum.org/trac/mpi-forum-web/ticket/439 I would like to bring this forward for the RMA WG to consider as an MPI-4 extension. Cheers, ~Jim.

1 0

Re: [mpiwg-rma] Short question on the ccNUMA memory reality
by maik peterson 07 Aug '14

07 Aug '14

i really wonder how a team can work on rma technologies without understanding the basics of todays computer systems ? 2014-08-05 20:01 GMT+02:00 Dave Goodell (dgoodell) <dgoodell(a)cisco.com>: > On Aug 5, 2014, at 12:37 PM, Rolf Rabenseifner <rabenseifner(a)hlrs.de> > wrote: > > > Dave, > > > > thank you for this helpful answer. > > > > My question was related to > > > >> The sentence p436:45-46 > >> "The order in which data is written is not > >> specified unless further synchronization is used." > > > > Your citation helps, because it tells the usual expectation: > >> 1. Each CPU will always perceive its own memory accesses as occurring > >> in program order. > > But in the MPI Standard nothing should be expected as usual. > > Right, since the MPI Standard is at least one more level removed from the > architecture. The behavior of an MPI program is also dependent on the > behavior of the C or Fortran programming language implementation, which > isn't addressed at all by the McKenney model. > > I think Bronis made an important, valid point earlier: > > Simply put, shared memory programming without compiler > > assistance is not something ordinary programmers should do. > > > > The text p436:43-48 may be modified into > > Advice to users. > > If accesses in the RMA unified model are not synchronized (with > > locks or flushes, see Section 11.5.3), load and store operations might > observe changes > > to the memory while they are in progress. The order in which data is > written is not > > specified unless further synchronization is used. This might lead to > inconsistent views > > on memory and programs that assume that a transfer is complete by only > checking > > parts of the message are erroneous. > > NEW: The only consistent view is that each process will always > > perceive its own memory accesses as occurring in program order. > > (End of advice to users.) > > > > But one can argue that this is already expected by everyone. > > In my opinion, it is better to expect on MPI shared Memory > > windows only semantics that is written black on White > > in the MPI Standard. > > I'll defer to the more active members of the RMA working group on this > proposed modification. In general, I'm really not fond of the entire MPI > RMA shared memory feature. I think it is something that should have been > left to a library outside of MPI, and that it's impossible to clearly > specify shared memory programming behavior in the MPI Forum's vernacular. > > -Dave > > _______________________________________________ > mpiwg-rma mailing list > mpiwg-rma(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-rma >

1 0