February 2010 - mpiwg-tools - mailman3.cels.anl.gov

Re: [Mpi3-tools] alternative signal handler text
by William Gropp 17 Feb '10

17 Feb '10

I second Dave's note. A "signal handler" is a concept that depends on the OS; Dave's text is more appropriate for an OS-agnostic standard. Dave's text is also more in the spirit of the MPI standard, which allows implementations to make performance tradeoffs while reminding implementors what the Forum feels most users will expect. Bill On Feb 16, 2010, at 5:15 PM, Dave Goodell wrote: > (comments are relative to the draft PDF retrieved 2010-02-16 @ 4:30pm > CST with md5sum "0fd54c6370574a93db13d3f5589d4b98") > > I would like to propose alternative text for signal handling in the > MPIT performance interface. The relevant text from the current draft > (page 21, lines 3-5 and page 22, lines 24-25) is reproduced below: > > -------8<------- > 3| The routines in this section to start and stop performance > variables must be safely > 4| callable from any asynchronous execution context, such as a > signal or interrupt handler, to > 5| allow their use in sampling-based tools. > ... > 24| The routines in this section to read, write, and reset > performance variables must be > 25| callable from signal handlers to allow their use in sampling- > based tools. > -------8<------- > > My proposed alternative: > -------8<------- > Advice to implementors. Although MPI places no requirements on the > interaction with external mechanisms such as signal handlers, it is > strongly recommended that the routines in this section to start and > stop performance variables should be safe to call in asynchronous > contexts. Examples of asynchronous contexts include signal handlers > and interrupt handlers. Such safety permits the development of > sampling-based tools. High quality implementations should strive to > make the results of any such interactions intuitive to users, and > attempt to document restrictions where deemed necessary. (End of > advice to implementors.) > ... > Advice to implementors. Although MPI places no requirements on the > interaction with external mechanisms such as signal handlers, it is > strongly recommended that the routines in this section to read, write, > and reset performance variables should be safe to call in asynchronous > contexts. Examples of asynchronous contexts include signal handlers > and interrupt handlers. Such safety permits the development of > sampling-based tools. High quality implementations should strive to > make the results of any such interactions intuitive to users, and > attempt to document restrictions where deemed necessary. (End of > advice to implementors.) > -------8<------- > > MPI has been careful in the past to leave signal safety and signal > handling as a quality of implementation issue (see MPI-2.2 pp22:14, > pp23:36, pp24:24, and pp384:45). I think that this was a wise > decision and that we should carry it forward into new versions of the > MPI Standard. I believe that with my proposed text implementations > will almost always support signal safety, excepting situations where > it is too difficult or impossible to do so. > > Alternatively, the current MPIT text is incompatible with MPI-2.2 > section 2.7 (and probably elsewhere). At the very least a proposal to > change that (perfectly good, IMO) text is needed. > > -Dave > William Gropp Deputy Director for Research Institute for Advanced Computing Applications and Technologies Paul and Cynthia Saylor Professor of Computer Science University of Illinois Urbana-Champaign

1 0

[Mpi3-tools] MPIT_PERFORMANCE_ALL and layering
by Dave Goodell 16 Feb '10

16 Feb '10

I think that we should reconsider including MPIT_PERFORMANCE_ALL as it currently exists in the MPIT proposal because it discourages libraries and layering among performance tools. If you have two tool libraries and one or both of them use MPIT_PERFORMANCE_ALL to start/stop their performance variables, the tools are likely to be incompatible. One of MPI's great strengths is providing support mechanisms for parallel libraries, I don't see why we should omit those mechanisms in the tools space. I understand the convenience and (potentially) performance arguments to be made for having functionality like this, but we might want to consider scoping it sort of like we do for communicators. Perhaps make it possible to group handles or open handles with an associated context of sorts to perform the group start/stop operations later. Alternatively, if there is no substantial performance argument to be made, we could just drop MPIT_PERFORMANCE_ALL altogether. A for-loop isn't particularly burdensome to tools/applications. I don't have a concrete alternative proposal here, I just wanted to bring up the issue for discussion. I unfortunately missed the last concall due to technical difficulties, so I wasn't able to bring it up there. -Dave

1 0

[Mpi3-tools] alternative signal handler text
by Dave Goodell 16 Feb '10

16 Feb '10

(comments are relative to the draft PDF retrieved 2010-02-16 @ 4:30pm CST with md5sum "0fd54c6370574a93db13d3f5589d4b98") I would like to propose alternative text for signal handling in the MPIT performance interface. The relevant text from the current draft (page 21, lines 3-5 and page 22, lines 24-25) is reproduced below: -------8<------- 3| The routines in this section to start and stop performance variables must be safely 4| callable from any asynchronous execution context, such as a signal or interrupt handler, to 5| allow their use in sampling-based tools. ... 24| The routines in this section to read, write, and reset performance variables must be 25| callable from signal handlers to allow their use in sampling- based tools. -------8<------- My proposed alternative: -------8<------- Advice to implementors. Although MPI places no requirements on the interaction with external mechanisms such as signal handlers, it is strongly recommended that the routines in this section to start and stop performance variables should be safe to call in asynchronous contexts. Examples of asynchronous contexts include signal handlers and interrupt handlers. Such safety permits the development of sampling-based tools. High quality implementations should strive to make the results of any such interactions intuitive to users, and attempt to document restrictions where deemed necessary. (End of advice to implementors.) ... Advice to implementors. Although MPI places no requirements on the interaction with external mechanisms such as signal handlers, it is strongly recommended that the routines in this section to read, write, and reset performance variables should be safe to call in asynchronous contexts. Examples of asynchronous contexts include signal handlers and interrupt handlers. Such safety permits the development of sampling-based tools. High quality implementations should strive to make the results of any such interactions intuitive to users, and attempt to document restrictions where deemed necessary. (End of advice to implementors.) -------8<------- MPI has been careful in the past to leave signal safety and signal handling as a quality of implementation issue (see MPI-2.2 pp22:14, pp23:36, pp24:24, and pp384:45). I think that this was a wise decision and that we should carry it forward into new versions of the MPI Standard. I believe that with my proposed text implementations will almost always support signal safety, excepting situations where it is too difficult or impossible to do so. Alternatively, the current MPIT text is incompatible with MPI-2.2 section 2.7 (and probably elsewhere). At the very least a proposal to change that (perfectly good, IMO) text is needed. -Dave

1 0

Re: [Mpi3-tools] New MPIT draft available / Question regarding Verbosity
by Tobias Hilbrich 11 Feb '10

11 Feb '10

Hi Martin, Hi group, its great to see the first versions of a proposal, I went over it and tried to find issues and errors. As I am unfortunately only rarely able to join the telco, I may raise issues that had already been discussed, if so feel free to ignore these comments. p7, line 3: "Fortran bindings are not required." -> One might give a rational for that, which is I guess that this targets tools that are usually not written in Fortran. p7, line 35: Uses "MPIT_Init" to refer to "MPIT_INIT" (consistency change) (also in line 41) p8, line 4-6: "Once this routine is called no MPIT routine ... may be called ...." -> I would add an except MPIT_INITIALIZED and MPIT_FINALIZED p9: general comment to MPIT_Datatype: A MPIT_DATATYPE_NULL might make the design more consistent to the regular MPI_Datatype design ... p9, line 32: "MPI_MAX_ENUMTYPE" -> "MPIT_MAX_ENUMTYPE" p11, line 18/19: IN iterator -> INOUT iterator p11 general comment to names: "MPIT_CONFIG_QUERY" and "MPI_Configquery_iterator" seem not very consistent p11 general comment on iterators: * Is there a strong need to use distinct iterator types for the config query and the performance query, couldn't we use a MPIT_Iterator type instead ? * MPIT_CONFIG_QUERY and MPIT_PERFORMANCE_QUERY might also set the iterator to MPIT_[ITERATOR|CONFIGQUERY|PERFORMANCEQUERY]_END when the call returns an MPIT_END (for consistency with common iterator concepts) p12, line 8: "MPIT_CONFIG_INFO(name, type, ...)" -> "MPIT_CONFIG_INFO(name, datatype, count, ...)" p12 comments on "MPIT_CONFIG_INFO": * Strictly seen, when the user may set desc to NULL, I would define desc as INOUT * What is the value of desclen when the user sets desc to NULL ? p13, line 33: "...the appropriate sized and ..." -> "...the appropriate size and ..." p16, line 18: "IN set string to a set" -> "string to a set" is not very descriptive, what about "string identifier of a set" or something similar ? p16, line 35: "that explains in the use ..." -> "that explains the use ..." p16, line 38: "name and description, are ..." -> "<b>name</b> and <b>desc</b>, are ..." p16, line 39: "A conforming MPI implementation may not ... " -> "A conforming MPI implementation must not ... " p16, line 41: "MPIT_PERFORMANCE_QUERY may ..." -> "MPIT_GET_SETINFO may ..." (same in line 46!) p16, line 44: "MPI_MAX_SET_NAME" -> "MPIT_MAX_SET_NAME" p16, general comment about MPIT_GET_SETINFO: * Strictly seen, when the user may set desc to NULL, I would define desc as INOUT (Same holds for further calls, I stopped listing all of them) p17, line 2: "MPI_MAX_SET_DESC" -> "MPIT_MAX_SET_DESC" p17, line 15: IN iterator iterator .... -> INOUT .... p17 comment about naming: "MPIT_PERFORMANCE_QUERY" vs "MPIT_Performancequery_iterator" p17, line 43: "The parameter name is used to return a short and unique name for the performance variable found." -> "The parameter name is used to return a short and unique name for the performance variable found, which is used as its identifier in other calls to the MPIT performance interface." p17, line 44: "A conforming MPI implementation may not..." -> "A conforming MPI implementation must not ..." p18, line 4: "MPI_MAX_PERFORMANCEVAR_NAME" -> "MPIT_MAX_PERFORMANCEVAR_NAME" p18, line 47: "MPI_MAX_PERFORMANCEVAR_DESC" -> "MPIT_MAX_PERFORMANCEVAR_DESC" p19, line 10: "MPI_MAX_PERFORMANCEVAR_SETS(???)" -> "MPIT_MAX_PERFORMANCEVAR_SETS(???)" p19, general comment to naming of MPIT Performance interface calls: * binding names like "MPIT_Performance_changed ()" seem to suggest a totally different behavior than they actually implement, in fact the MPIT_Performance is meant as a sort of namespace here why not use MPIT_PERFORMANCE_Changed to denote that, or if this is inconsistent with the rest of the standard (which apparently is the case), then why not use an extra word here, e.g. MPIT_Performance_vars_changed. p19, comment to MPIT_Performancehandle: * none of the other MPI handles use "handle" as part of their name, so what about "MPIT_Performance[_]var" ? * what about a MPIT_PERFORMANCE[HANDLE/VAR]_NULL for consistency with the other MPI handles ? * One might add that the user of the MPIT interface should free all performance handles before calling MPIT_FINALIZE * MPIT_PERFORMANCE_ALL is inconsistent with other constant handles (that use the handle name as their base name. e.g. MPI_Comm -> MPI_COMM_WORLD), what about MPIT_PERFORMANCE[HANDLE|VAR]_ALL p21, line 33: "MPIT_ERR_PERFORANCE_WRITE" -> "MPIT_PERFORMANCE_WRITE" -Tobias Martin Schulz wrote: > Hi all, > > I uploaded a new draft of the MPIT proposal to > > https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Tools/draft > > Changes since the last draft are marked with changebars and are > colored in green. > The latest draft contains changes based on comments from Kathryn > Mohror and > the MPI group at Sun/Oracle. > > There is one open question that I would like to bring up for a general > discussion: > as discussed in the WG earlier, we need some kind of verbosity > parameter that > allows the MPI implementation to classify each variable according to > its importance > and general usage scenario. Jeff and I have been going forth and back > on this > one for a while over email, but we can't come to a useful conclusion > and hence > would like wider feedback. > > In general, we both think that we want 5 classes of verbosity (odd > number larger > than 3, but not too large). > > The original proposal had the following classes > VERY_HIGH > HIGH > MEDIUM > LOW > VERY_LOW > which was rejected by the forum (I think rightfully so) since it > doesn't really convey > any useful application level information. > > The current proposal has: > USER_BASIC > USER_DETAILED > TUNER_BASIC > TUNER_DETAILED > MPI_IMPLEMENTOR > with the idea that the first two classes are intended for general MPI > users, > the third and fourth for performance experts and system administrators > wishing to tune the performance of MPI itself, while the fifth is for MPI > implementors and would include debugging information. > > Open MPI uses a similar idea of verbosity and defines the following > five classes: > INFORMATION > TRIVIAL > MINOR > MAJOR > CRITICAL > However, both Jeff and I don't like these classes, since they don't > convey > any useful information either. > > Jeff proposes something like > BASIC > ? > ADVANCED > ? > INTERNAL > but this proposal misses good names for 2 and 4. > > Any comments or suggestions (either extending one of the above or > completely > new) would be very much appreciated. > > Thanks, > > Martin > > > ________________________________________________________________________ > Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm > CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > > > _______________________________________________ > Mpi3-tools mailing list > Mpi3-tools(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools > -- Tobias Hilbrich GWT - TUD GmbH Chemnitzer Straï¿½e 48b 01187 Dresden Tel.: +49 351 463 35876 Fax: +49 351 463 37773 http://www.gwtonline.de Geschï¿½ftsfï¿½hrer: Claus-Peter Held Prokuristin: Beate-Victoria Ermisch Die GWT ist eingetragen beim Amtsgericht Dresden - HRB: 13 840

1 0

Re: [Mpi3-tools] New MPIT draft available / Question regarding Verbosity
by Martin Schulz 08 Feb '10

08 Feb '10

Hi Marc-Andre, On Feb 8, 2010, at 12:46 AM, Marc-Andre Hermanns wrote: > Martin, > > although not part of the MPIT chapter, the first page (chapter > 1.2.1) on > line 32/33 seems to be missing a verb. The sentence ends right after > the > "... as macros (See Section x)." and I would expect a phrase like "can > be intercepted". Thanks for finding this - this is part of the existing profiling interface text and needs a separate change procedure. I started collecting these and I will add this one to my list. Thanks, Martin > > Best regards, > Marc-Andre > > PS: I had to remove the color option of the changebars package > becaus of > some "option name clash" regarding "color". Without the option is > compiles fine. Strange - the annotated MPI standard already uses color. I will see if I can fix this. > > Martin Schulz wrote: >> Hi all, >> >> I uploaded a new draft of the MPIT proposal to >> >> https://*svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Tools/draft >> >> Changes since the last draft are marked with changebars and are >> colored in green. >> The latest draft contains changes based on comments from Kathryn >> Mohror and >> the MPI group at Sun/Oracle. >> >> There is one open question that I would like to bring up for a >> general >> discussion: >> as discussed in the WG earlier, we need some kind of verbosity >> parameter that >> allows the MPI implementation to classify each variable according to >> its importance >> and general usage scenario. Jeff and I have been going forth and back >> on this >> one for a while over email, but we can't come to a useful conclusion >> and hence >> would like wider feedback. >> >> In general, we both think that we want 5 classes of verbosity (odd >> number larger >> than 3, but not too large). >> >> The original proposal had the following classes >> VERY_HIGH >> HIGH >> MEDIUM >> LOW >> VERY_LOW >> which was rejected by the forum (I think rightfully so) since it >> doesn't really convey >> any useful application level information. >> >> The current proposal has: >> USER_BASIC >> USER_DETAILED >> TUNER_BASIC >> TUNER_DETAILED >> MPI_IMPLEMENTOR >> with the idea that the first two classes are intended for general MPI >> users, >> the third and fourth for performance experts and system >> administrators >> wishing to tune the performance of MPI itself, while the fifth is for >> MPI >> implementors and would include debugging information. >> >> Open MPI uses a similar idea of verbosity and defines the following >> five classes: >> INFORMATION >> TRIVIAL >> MINOR >> MAJOR >> CRITICAL >> However, both Jeff and I don't like these classes, since they don't >> convey >> any useful information either. >> >> Jeff proposes something like >> BASIC >> ? >> ADVANCED >> ? >> INTERNAL >> but this proposal misses good names for 2 and 4. >> >> Any comments or suggestions (either extending one of the above or >> completely >> new) would be very much appreciated. >> >> Thanks, >> >> Martin >> >> >> ________________________________________________________________________ >> Martin Schulz, schulzm(a)llnl.gov, http://*people.llnl.gov/schulzm >> CASC @ Lawrence Livermore National Laboratory, Livermore, USA >> >> >> >> _______________________________________________ >> Mpi3-tools mailing list >> Mpi3-tools(a)lists.mpi-forum.org >> http://*lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools > > > -- > Marc-Andre Hermanns > Juelich Supercomputing Centre > Institute for Advanced Simulation > Forschungszentrum Juelich GmbH > D-52425 Juelich > Germany > > Phone : +49-2461-61-2054 > Fax : +49-2461-61-6656 > eMail : m.a.hermanns(a)fz-juelich.de > WWW : http://*www.*fz-juelich.de/jsc/ > > JSC is the coordinator of the > John von Neumann Institute for Computing > and member of the > Gauss Centre for Supercomputing > > Sitz der Gesellschaft: Juelich > Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 > Vorsitzende des Aufsichtsrats: MinDir'in Baerbel Brumme-Bothe > Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender), > Dr. Ulrich Krafft (stellv. Vorsitzender), > Prof. Dr. Harald Bolt, > Prof. Dr. Sebastian M. Schmidt > _______________________________________________ > Mpi3-tools mailing list > Mpi3-tools(a)lists.mpi-forum.org > http://*lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools > ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA

1 0

Re: [Mpi3-tools] New MPIT draft available / Question regarding Verbosity
by Marc-Andre Hermanns 08 Feb '10

08 Feb '10

Martin, although not part of the MPIT chapter, the first page (chapter 1.2.1) on line 32/33 seems to be missing a verb. The sentence ends right after the "... as macros (See Section x)." and I would expect a phrase like "can be intercepted". Best regards, Marc-Andre PS: I had to remove the color option of the changebars package becaus of some "option name clash" regarding "color". Without the option is compiles fine. Martin Schulz wrote: > Hi all, > > I uploaded a new draft of the MPIT proposal to > > https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Tools/draft > > Changes since the last draft are marked with changebars and are > colored in green. > The latest draft contains changes based on comments from Kathryn > Mohror and > the MPI group at Sun/Oracle. > > There is one open question that I would like to bring up for a general > discussion: > as discussed in the WG earlier, we need some kind of verbosity > parameter that > allows the MPI implementation to classify each variable according to > its importance > and general usage scenario. Jeff and I have been going forth and back > on this > one for a while over email, but we can't come to a useful conclusion > and hence > would like wider feedback. > > In general, we both think that we want 5 classes of verbosity (odd > number larger > than 3, but not too large). > > The original proposal had the following classes > VERY_HIGH > HIGH > MEDIUM > LOW > VERY_LOW > which was rejected by the forum (I think rightfully so) since it > doesn't really convey > any useful application level information. > > The current proposal has: > USER_BASIC > USER_DETAILED > TUNER_BASIC > TUNER_DETAILED > MPI_IMPLEMENTOR > with the idea that the first two classes are intended for general MPI > users, > the third and fourth for performance experts and system administrators > wishing to tune the performance of MPI itself, while the fifth is for > MPI > implementors and would include debugging information. > > Open MPI uses a similar idea of verbosity and defines the following > five classes: > INFORMATION > TRIVIAL > MINOR > MAJOR > CRITICAL > However, both Jeff and I don't like these classes, since they don't > convey > any useful information either. > > Jeff proposes something like > BASIC > ? > ADVANCED > ? > INTERNAL > but this proposal misses good names for 2 and 4. > > Any comments or suggestions (either extending one of the above or > completely > new) would be very much appreciated. > > Thanks, > > Martin > > > ________________________________________________________________________ > Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm > CASC @ Lawrence Livermore National Laboratory, Livermore, USA > > > > _______________________________________________ > Mpi3-tools mailing list > Mpi3-tools(a)lists.mpi-forum.org > http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools -- Marc-Andre Hermanns Juelich Supercomputing Centre Institute for Advanced Simulation Forschungszentrum Juelich GmbH D-52425 Juelich Germany Phone : +49-2461-61-2054 Fax : +49-2461-61-6656 eMail : m.a.hermanns(a)fz-juelich.de WWW : http://www.fz-juelich.de/jsc/ JSC is the coordinator of the John von Neumann Institute for Computing and member of the Gauss Centre for Supercomputing Sitz der Gesellschaft: Juelich Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzende des Aufsichtsrats: MinDir'in Baerbel Brumme-Bothe Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender), Dr. Ulrich Krafft (stellv. Vorsitzender), Prof. Dr. Harald Bolt, Prof. Dr. Sebastian M. Schmidt

1 0

[Mpi3-tools] New MPIT draft available / Question regarding Verbosity
by Martin Schulz 08 Feb '10

08 Feb '10

Hi all, I uploaded a new draft of the MPIT proposal to https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MPI3Tools/draft Changes since the last draft are marked with changebars and are colored in green. The latest draft contains changes based on comments from Kathryn Mohror and the MPI group at Sun/Oracle. There is one open question that I would like to bring up for a general discussion: as discussed in the WG earlier, we need some kind of verbosity parameter that allows the MPI implementation to classify each variable according to its importance and general usage scenario. Jeff and I have been going forth and back on this one for a while over email, but we can't come to a useful conclusion and hence would like wider feedback. In general, we both think that we want 5 classes of verbosity (odd number larger than 3, but not too large). The original proposal had the following classes VERY_HIGH HIGH MEDIUM LOW VERY_LOW which was rejected by the forum (I think rightfully so) since it doesn't really convey any useful application level information. The current proposal has: USER_BASIC USER_DETAILED TUNER_BASIC TUNER_DETAILED MPI_IMPLEMENTOR with the idea that the first two classes are intended for general MPI users, the third and fourth for performance experts and system administrators wishing to tune the performance of MPI itself, while the fifth is for MPI implementors and would include debugging information. Open MPI uses a similar idea of verbosity and defines the following five classes: INFORMATION TRIVIAL MINOR MAJOR CRITICAL However, both Jeff and I don't like these classes, since they don't convey any useful information either. Jeff proposes something like BASIC ? ADVANCED ? INTERNAL but this proposal misses good names for 2 and 4. Any comments or suggestions (either extending one of the above or completely new) would be very much appreciated. Thanks, Martin ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA

1 0

[Mpi3-tools] TelCon Monday 8am PT
by Martin Schulz 08 Feb '10

08 Feb '10

Hi all, Just a quick reminder: we'll have the next MPI3 tools WG on Monday 2/8 (tomorrow) at the usual time 8am PT, 11am ET, 5pm MEZ. The link to the Webex meeting is on the tools Wiki. We'll go through any tool document that is ready for discussions. Further, I would like to discuss how we will present our proposals during the next MPI forum. Martin ________________________________________________________________________ Martin Schulz, schulzm(a)llnl.gov, http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA

1 0

Re: [Mpi3-tools] MPIT chapter review
by Marty Itzkowitz 07 Feb '10

07 Feb '10

On 2/6/10 10:56 PM, Martin Schulz wrote: > Hi Marty, all, > > On Feb 6, 2010, at 11:45 AM, marty.itzkowitz(a)Sun.COM wrote: > >> [Sorry it's taken me so long; as you might imagine, life at Sun, now >> a wholly owned >> subsidiary of Oracle, and soon to disappear as an independent >> company, has >> been "interesting."] > > I bet - I hope everything is still OK at the new company and that the > old projects > continue! I see, though, that you are still using "sun.com" email > addresses. Will > this stay this way? Sun is supposed to disappear as a company on Feb 15, and I (along with all the performance tools people, and the MPI people) will become a genuine Oracle employee then. I imagine they'll keep the email relay for quite a while. > >> >> I've reviewed it, and I've had two of our MPI folks, and one other >> tools person >> review it, too. The aggregated comments are below. > > Excellent - thanks for taking the time. > >> >> Marty >> >> From Marty Itzkowitz: >> >> I dislike the use of the feminine pronouns by default. Either >> rewrite so that neither >> masculine nor feminine pronouns are needed, or always use "he or she", >> "his or her." > > I personally don't have a big opinion here, but the rest of the MPI > standard is written > that way and I think we need to be consistent with it (unless we want > to suggest > global changes). In fact, in most cases the feminine versions are > straight copies > of text used in the MPI 2.2 standard. > >> >> p5, line 47. The meaning does not get more complex, the >> interpretation does. > > Agreed - however, also this is a straight copy of the existing profile > chapter and > we would need a separate ticket to change this. I'll keep this in mind > for when > we need to make other changes in this chapter to move it under the tools > chapter. > >> >> p22, line 7, reads better as "*safely *callable from *asynchronous* >> signal handlers to allow >> their use in sampling-based tools." > > Changed - I also tried to describe it more generically to avoid the > comments > we got last time, > > As a higher level question: do you see your State proposal represented > (or > implementable) with the MPIT proposal? I think it is, but would > definitely > like to address shortcomings if you don't think so. I think do. An MPI that supports it would have that performance variable, and it's readable from a profile signal handler. We could try it as soon as you have a prototype in, say OpenMPI. One more thing: We found a performance cost to running a library with the instrumentation, and set up a means to tell mpirun to use instrumented libraries, rather than the default. It might be nice to have that in the spec, too. >> From Terry Dontje: >> 1. What does MPIT_GET_SETINFO really do? I guess my problem is it >> takes a "set name" but where does one get the set name and how does >> that differ from the out parameter "name of the set". I read the >> section preceding the function description and I still just don't get >> it. > > The idea was that the name used in the set information is just some > short name > with the restricted characterset described in the set section. This > function could > then return a "full" name (which is more explicit and descriptive) > along with some > actual description text. I tried to make this a bit more clear, but we > could also drop > returning the name, if we feel the short and restricted name is > sufficient. > >> >> 2. Seems like a lot of functions for the performance accessors. So >> to do the MPI state acquisition you will need to do the following: >> >> a. Loop on MPIT_PERFORMANCE_QUERY until you find the state >> performance variable. >> b. Call MPIT_PERFORMANCE_INFO with the state performance variable >> name to get the appropriate info you need (specifically datatype/count). >> c. Call MPIT_PERFORMANCE_GETHANDLE on the state performance variable >> name. >> d. Call MPIT_PERFORMANCE_START on the state performance handle to >> activate the use of the state code. >> e. Periodically call MPIT_PERFORMANCE_READ using the state >> performance handle while the MPI code is running >> f. Once the MPI code finishes call MPIT_PERFORMANCE_FREEHANDLE >> >> I guess it isn't that bad we're just adding 4 additional calls which >> are necessary in order to have more than one performance variable. >> Just seems like a lot on paper (in the spec). > > Not sure how to change this - the first two allow the implementation to > specify what it wants to deliver, the gethandle step allows us optimize > the read access without requiring full handles for each variable, which > could be prohibitive for some implementations, and start/read is the > typical semantics as in PAPI. Note, though, that a) is optional and can > be skipped if you know the name (e.g., through a command line > parameter) beforehand. In this case, also b) could be skipped, if you > are sure you know the type and count. It's also one-time overhead at startup, but that's not so bad. > >> >> From Eugene Loh: >> >> Have the MPIT interfaces been prototyped by anyone? [Yukon Maruyama >> asked the >> same question. None of us feels confident we can understand the >> issues before we see >> a prototype.] > > The interface has changed slightly since the last forum, but there is > a dummy > implementation in the SVN for the older version of the API (an conforming > implementation that provides no variables). I also have an early > prototype on > top of MVAPICH returning both configuration and performance variables and > implementing this has indeed led to a few changes in the API. I am also > talking to the MPICH, the MVAPICH, and the Open MPI groups and all > three said that they are interested in adding a prototype. > >> >> Since the returned information is so implementation-dependent, it's >> hard for me to imagine that there could be (m)any MPIT-based tools >> that make sense for multiple MPI implementations. Meanwhile, if a >> tool makes sense for only one MPI implementation, it should be part >> of the implementation itself. In general, I'm still scratching my >> head over this MPIT stuff. > > I see this similar to the CPU counters and their use in PAPI. Each CPU > has > their own counters and they mean something different on each platform, > but they still have been very helpful in measuring and optimizing > performance. > Also, having a unified interface, like PAPI, has been extremely helpful > for tools and I expect MPI tools to use the MPIT interface in a > similar way. > > Many tools will simply measure all variables in a class and then > report them > to the user along with the description text returned by MPI. The user > will > then ultimately be responsible in interpreting the results, but such a > profiler > will run across all platforms. > > Further, several people have suggested to create a table with a set of > names that could be used across implementations. The idea would be > to describe concepts that are present in most MPIs and then > implementations > would be required to name their variable with the common name, if it > offers such a variable (but it would not be required to offer it if it > doesn't > support that type of information). Such an extension table could be in > the MPI standard or in a support layer (like PAPI) outside of MPI. > >> >> And some nits: >> page 6, line 47: >> >> All identifier covered by this interface carry >> -> >> All identifiers covered by this interface carry > > fixed. > >> >> page 7, line 1: >> >> all conventions and principle governing >> -> >> all conventions and principles governing > > fixed > >> >> page 7, line 33: >> >> must contain exactly one call to the MPIT initialization routine >> -> >> must execute exactly one call to the MPIT initialization routine > > I agree, this would sound better. I took this text from the other parts > of the MPI standard on MPI_Init, though, for consistency. > >> >> page 8, lines 17 and 26: >> >> before MPIT_INIT and after MPIT_FINALIZE >> -> >> before MPIT_INIT or after MPIT_FINALIZE > > fixed > >> >> page 12, line 42: The table of legal "scope" values seems to be >> misplaced to the top of page 13. > > I changed this to a reference to the table. > >> >> page 12, line 44: >> >> it is not a guarantee that can be changed. >> Is there a typo in that line? >> [Marty] I'm not quite sure what this line is saying. Does it mean >> that even if the >> scope says that a variable is changeable, the call to change it is >> not guaranteed to >> succeed? If that's so, what can the tool implementor rely on? > > Yes, that's what it means - the call can fail and a tool should check > for it and > must be able to react appropriately. It is expected that a failure is > rare (e.g., > because a certain resource is busy or the MPI is currently handling > something > that prevents it from executing the variable access). However, if the > scope > says that it can't be changed at all or after Init, then the tool > knows that it > is hopeless to even try. This is problematic. Sure, any call may fail, but it's not clear from the spec what the implementor is supposed to do with the failure. Retry immediately? Loop until it goes through? > > This semantics is not unusual - any access can fail any time if the MPI > library decides to return an error. Hence, perhaps this sentence is > not necessary. I'll take it out. Ok, thanks. Talk to you tomorrow.

1 0

Re: [Mpi3-tools] MPIT chapter review
by Bronis R. de Supinski 07 Feb '10

07 Feb '10

I vote for removing it. I agree with Marty that it is irritating writing -- best is to eliminate the need for the pronoun. On Sun, 7 Feb 2010, Rajeev Thakur wrote: > > > From Marty Itzkowitz: > > > > > > I dislike the use of the feminine pronouns by default. > > Either rewrite > > > so that neither > > > masculine nor feminine pronouns are needed, or always use "he or > > > she", > > > "his or her." > > > > I personally don't have a big opinion here, but the rest of > > the MPI standard is written that way and I think we need to > > be consistent with it (unless we want to suggest global > > changes). In fact, in most cases the feminine versions are > > straight copies of text used in the MPI 2.2 standard. > > Looks like it is only in the profiling interface (PMPI), and it is has > been there since MPI 1.0. > > Rajeev > > _______________________________________________ > Mpi3-tools mailing list > Mpi3-tools(a)lists.mpi-forum.org > http://*lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools > >

1 0