Received: from cap.totalviewtech.com (192.168.68.251) by mymail.roguewave.com (10.22.0.14) with Microsoft SMTP Server (TLS) id 8.2.254.0; Mon, 19 Jul 2010 09:00:19 -0600 From: John DelSignore To: "mpi3-tools@lists.mpi-forum.org" Date: Mon, 19 Jul 2010 09:00:18 -0600 Subject: Re: [Mpi3-tools] Reminder: next tools WG Meeting Monday 7/19 (tomorrow) - topic: MPIR Thread-Topic: [Mpi3-tools] Reminder: next tools WG Meeting Monday 7/19 (tomorrow) - topic: MPIR Thread-Index: AcsnUxsy0psP+d5xTOiDWgxvpu4ZEQ== Message-ID: <4C446882.5020703@roguewave.com> References: In-Reply-To: Accept-Language: en-US Content-Language: en-US X-MS-Exchange-Organization-AuthAs: Internal X-MS-Exchange-Organization-AuthMechanism: 0a X-MS-Exchange-Organization-AuthSource: Fountain.Blue.Roguewave.Com X-MS-Has-Attach: X-MS-TNEF-Correlator: user-agent: Thunderbird 2.0.0.18 (X11/20081113) Content-Type: text/plain; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Hi Marc-Andr=E9, My initial replies in-line below... Marc-Andr=E9 Hermanns wrote: > Hi all, >=20 > as I am quite inexperienced with the implementation and use of debugging = interfaces I have been more silent about the MPIR discussion, yet, I wanted= to raise some questions. I apologize when the answers might seem obvious. >=20 > =3D=3D=3D=3D=3D=3D=3D >=20 > Page 7/8, Par. 1: > =85after it has set MPIR_debug_state to MPIR_DEBUG_SPAWNED and =85 >=20 > Should the variable have an initial state? MPIR_NULL? I think that it doesn't matter. The only time that the tool should look at = the variable is when the starter process calls MPIR_Breakpoint(). > =3D=3D=3D=3D=3D=3D=3D >=20 > Page 8, Par. 3: > =85, and avoid defining MPIR_debug_gate in the MPI processes. >=20 > What does avoid mean? Is it still ok to do so? If it is ok, why should it= be avoided? When it is not ok, why not prohibit it? We plan to discuss this in more detail during the call today. Dong Ahn and = I have discussed this issue in email near the end of June. > =3D=3D=3D=3D=3D=3D=3D >=20 > Page 11, Par. 3 + 6: > If MPIR_Breakpoint exists the tool should consider this a starter process= =85 > if MPIR_i_am_starter ist not defined =85. >=20 > why are there two variables/symbols that seem to do the same thing? Because MPI process 0 may also be a starter process, as was done in MPICH 1= . > =3D=3D=3D=3D=3D=3D=3D >=20 > Page 18, Par. 10: > =85is a symbol of any kind, preferably an int, =85. >=20 > If it can be any kind, why prefer an int? Why not require it to be an int= ? Because its type doesn't matter, only its definition. > =3D=3D=3D=3D=3D=3D=3D >=20 > In general: > - Some symbols are evaluated for existence or non-existence, yet others h= ave a state (1 or 0). Why? Existence or non-existence is a static property, whereas state (1 or 0) is = a dynamic property. Cheers, John D. > Best regards, > Marc-Andre >=20 > On 18.07.2010, at 23:27, Martin Schulz wrote: >=20 >> Just as a reminder: the next MPI tools WG meeting will be tomorrow morni= ng at 8am PDT, >> 11am EDT, 5pm MSZ. The main item on the agenda is the discussion on the = MPIR >> document that John has put together. >> >> The link to the Webex session is on the tools wiki and below: >> https://cisco.webex.com/ciscosales/j.php?ED=3D144317932&UID=3D0&PW=3DNYj= k1YmJhZGEz&RT=3DMiMxMQ%3D%3D >> Password: tools >> >> The latest version of the MPIR document is attached. >> >> Martin >> >> >> ________________________________________________________________________ >> Martin Schulz, schulzm@llnl.gov, http://people.llnl.gov/schulzm >> CASC @ Lawrence Livermore National Laboratory, Livermore, USA >> >> >> >> _______________________________________________ >> Mpi3-tools mailing list >> Mpi3-tools@lists.mpi-forum.org >> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools >=20 > -- > Marc-Andre Hermanns > German Research School for > Simulation Sciences GmbH > c/o Laboratory for Parallel Programming > 52056 Aachen | Germany >=20 > Tel +49 241 80 99753 > Fax +49 241 80 6 99753 > Web www.grs-sim.de >=20 > Members: Forschungszentrum J=FClich GmbH | RWTH Aachen University > Registered in the commercial register of the local court of > D=FCren (Amtsgericht D=FCren) under registration number HRB 5268 > Registered office: J=FClich > Executive board: Prof. Dr. Heiner M=FCller-Krumbhaar | Prof. Marek Behr, = Ph.D. >=20 >=20