[openib-commits] r5014 - gen2/trunk/src/userspace/mpi/mvapich2-gen2
surs at openib.org
surs at openib.org
Sun Jan 15 14:52:37 PST 2006
Author: surs
Date: 2006-01-15 14:52:36 -0800 (Sun, 15 Jan 2006)
New Revision: 5014
Added:
gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES_MPICH
Removed:
gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES
Log:
-- Rename RELEASE_NOTES to RELEASE_NOTES_MPICH
Deleted: gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES
===================================================================
--- gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES 2006-01-15 22:36:10 UTC (rev 5013)
+++ gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES 2006-01-15 22:52:36 UTC (rev 5014)
@@ -1,83 +0,0 @@
--------------------------------------------------------------------------------
- Major Changes Impacting Usability
--------------------------------------------------------------------------------
-
-- A new and improved process manager that spawns processes using fork() in now
- available. This process manager, named "gforker", replaces the "forker"
- process manager found in previous releases.
-
-- (see the CHANGES file for a more extensive list of changes and new features)
-
--------------------------------------------------------------------------------
- Known Deficiencies
--------------------------------------------------------------------------------
-
-- The sock channel is the only channel that implements dynamic process support
- (i.e., MPI_COMM_SPAWN, MPI_COMM_CONNECT, MPI_COMM_ACCEPT, etc.). All other
- channels will experience failures for tests exercising dynamic process
- functionality.
-
-- Support for the "external32" data representation is incomplete. This affects
- the MPI_PACK_EXTERNAL and MPI_UNPACK_EXTERNAL routines, as well the external
- data representation capabilities of ROMIO.
-
-- The CH3 device does not presently support heterogeneous communication. That
- is to say that the processes involved in a job must use the same basic type
- sizes and format. The sizes and format are typically determined by the
- processor architecture, although it may also be influenced by compiler
- options.
-
-- The support for MPI_THREAD_SERIALIZED and MPI_THREAD_MULTIPLE is preliminary,
- and only available with the sock channel. Rudimentary testing has been
- performed; however, several problems still exist. Several of the MPI I/O
- routines hang. Some of the internal routines can cause a hang if they
- encounter an error. We plan to correct these issues in the next release.
- That said, we encourage people to experiment and provide feedback.
-
-- MPI_IRECV operations that are not explicitly completed before MPI_FINALIZE is
- called may fail to complete before MPI_FINALIZE returns, and thus never
- complete. Furthermore, any matching send operations may erroneously fail.
- By explicitly completed, we mean that the request associated with the
- operation is completed by one of the MPI_TEST or MPI_WAIT routines.
-
-- MPI_INTERCOMM_CREATE does not work correctly when the union of the local and
- remote communicators are not already part of an existing (intra or inter)
- communicator. For example, if communicator A spawns a set of processes whose
- MPI_COMM_WORLD we will refer to as communicator B. Likewise, communicator B
- spawns a set of processes whose MPI_COMM_WORLD we will refer to as
- communicator C. Despite being a legal request, MPI_INTERCOMM_CREATE will
- fail to create a new intercommunicator between A and C using a commom process
- in B.
-
-- C++ Binding:
-
- The MPI datatypes corresponding to Fortran datatypes are not available
- (e.g., no MPI::DOUBLE_PRECISION).
-
- The C++ binding does not implement a separate profiling interface,
- as allowed by the MPI-2 Standard (Section 10.1.10 Profiling).
-
- With the exception of the profiling interface, future releases of MPICH2
- will address these limitations of the C++ binding.
-
-- For passive target RMA, there is no asynchronous agent at the target
- that will cause progress to occur. Progress occurs only when the user
- calls an MPI function at the target (which could well be MPI_WIN_FREE).
-
-- --enable-sharedlibs=gcc does not work on Solaris because of difference
- between the GNU ld program and the Solaris ld program
-
-
--------------------------------------------------------------------------------
- Issues for Developers
--------------------------------------------------------------------------------
-
-- MPE2, which is distributed as part of MPICH2, requires autoconf 2.52 or
- higher. By default, MPICH2 and MPE use the autoconf programs in one's path.
- The environment variables AUTOCONF and AUTOHEADER can be set to specify
- alternative commands for MPICH2. Likewise, MPE_AUTOCONF and MPE_AUTOHEADER
- can be set to specify alternative commands for MPE2. Obviously, these
- environment variables must be set before running "maint/updatefiles" or
- "make dist". In order to handle incompatibilities between different
- minor versions of autoconf, we've selected autoconf 2.59 as the version
- required for MPE2 and MPICH2.
Added: gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES_MPICH
===================================================================
--- gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES_MPICH 2006-01-15 22:36:10 UTC (rev 5013)
+++ gen2/trunk/src/userspace/mpi/mvapich2-gen2/RELEASE_NOTES_MPICH 2006-01-15 22:52:36 UTC (rev 5014)
@@ -0,0 +1,83 @@
+-------------------------------------------------------------------------------
+ Major Changes Impacting Usability
+-------------------------------------------------------------------------------
+
+- A new and improved process manager that spawns processes using fork() in now
+ available. This process manager, named "gforker", replaces the "forker"
+ process manager found in previous releases.
+
+- (see the CHANGES file for a more extensive list of changes and new features)
+
+-------------------------------------------------------------------------------
+ Known Deficiencies
+-------------------------------------------------------------------------------
+
+- The sock channel is the only channel that implements dynamic process support
+ (i.e., MPI_COMM_SPAWN, MPI_COMM_CONNECT, MPI_COMM_ACCEPT, etc.). All other
+ channels will experience failures for tests exercising dynamic process
+ functionality.
+
+- Support for the "external32" data representation is incomplete. This affects
+ the MPI_PACK_EXTERNAL and MPI_UNPACK_EXTERNAL routines, as well the external
+ data representation capabilities of ROMIO.
+
+- The CH3 device does not presently support heterogeneous communication. That
+ is to say that the processes involved in a job must use the same basic type
+ sizes and format. The sizes and format are typically determined by the
+ processor architecture, although it may also be influenced by compiler
+ options.
+
+- The support for MPI_THREAD_SERIALIZED and MPI_THREAD_MULTIPLE is preliminary,
+ and only available with the sock channel. Rudimentary testing has been
+ performed; however, several problems still exist. Several of the MPI I/O
+ routines hang. Some of the internal routines can cause a hang if they
+ encounter an error. We plan to correct these issues in the next release.
+ That said, we encourage people to experiment and provide feedback.
+
+- MPI_IRECV operations that are not explicitly completed before MPI_FINALIZE is
+ called may fail to complete before MPI_FINALIZE returns, and thus never
+ complete. Furthermore, any matching send operations may erroneously fail.
+ By explicitly completed, we mean that the request associated with the
+ operation is completed by one of the MPI_TEST or MPI_WAIT routines.
+
+- MPI_INTERCOMM_CREATE does not work correctly when the union of the local and
+ remote communicators are not already part of an existing (intra or inter)
+ communicator. For example, if communicator A spawns a set of processes whose
+ MPI_COMM_WORLD we will refer to as communicator B. Likewise, communicator B
+ spawns a set of processes whose MPI_COMM_WORLD we will refer to as
+ communicator C. Despite being a legal request, MPI_INTERCOMM_CREATE will
+ fail to create a new intercommunicator between A and C using a commom process
+ in B.
+
+- C++ Binding:
+
+ The MPI datatypes corresponding to Fortran datatypes are not available
+ (e.g., no MPI::DOUBLE_PRECISION).
+
+ The C++ binding does not implement a separate profiling interface,
+ as allowed by the MPI-2 Standard (Section 10.1.10 Profiling).
+
+ With the exception of the profiling interface, future releases of MPICH2
+ will address these limitations of the C++ binding.
+
+- For passive target RMA, there is no asynchronous agent at the target
+ that will cause progress to occur. Progress occurs only when the user
+ calls an MPI function at the target (which could well be MPI_WIN_FREE).
+
+- --enable-sharedlibs=gcc does not work on Solaris because of difference
+ between the GNU ld program and the Solaris ld program
+
+
+-------------------------------------------------------------------------------
+ Issues for Developers
+-------------------------------------------------------------------------------
+
+- MPE2, which is distributed as part of MPICH2, requires autoconf 2.52 or
+ higher. By default, MPICH2 and MPE use the autoconf programs in one's path.
+ The environment variables AUTOCONF and AUTOHEADER can be set to specify
+ alternative commands for MPICH2. Likewise, MPE_AUTOCONF and MPE_AUTOHEADER
+ can be set to specify alternative commands for MPE2. Obviously, these
+ environment variables must be set before running "maint/updatefiles" or
+ "make dist". In order to handle incompatibilities between different
+ minor versions of autoconf, we've selected autoconf 2.59 as the version
+ required for MPE2 and MPICH2.
More information about the commits
mailing list