ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
6 865 коммитов 3 Ветки 0 Теги
Граф коммитов

143 Коммитов

Автор SHA1 Сообщение Дата
George Bosilca
9dd522e350 Correct the MPI_COMBINER_DUP case when we generate the data description. Now, we have an 
optimized version here, where we pack only 2 integers instead of 6.

This commit was SVN r9102.
 2006-02-21 17:16:30 +00:00
George Bosilca
3ac541a722 Remove some debugging code. 
This commit was SVN r9101.
 2006-02-21 15:17:50 +00:00
George Bosilca
d4ab61aec5 Homogeneously handle the one-sided features of the datatype. 
This commit was SVN r9096.
 2006-02-20 17:22:29 +00:00
George Bosilca
79be2dbcdb Anything else than void* cannot be print with %p ... 
This commit was SVN r9094.
 2006-02-19 16:34:15 +00:00
George Bosilca
a2a08b2282 Add the missing prototype and use the function (ompi_ddt_print_args) 
in ompi_ddt_dump.

This commit was SVN r9093.
 2006-02-18 18:15:54 +00:00
George Bosilca
19fb826caf Add a new function to print how a datatype was created based on the 
MPI informations stored in args (the MPI_COMBINER stuff). This function
print only one level of information but it expand the name of the
datatype if the data is a predefined one.

This commit was SVN r9090.
 2006-02-18 04:18:39 +00:00
George Bosilca
b62b4eb5b8 icc point out a lot of useless comparaison as well as some warnings about 
inlined functions. Thie patch fix them all.

This commit was SVN r9082.
 2006-02-17 05:15:59 +00:00
George Bosilca
f4d1bea50e Allow us to add zero length datatype ... 
This commit was SVN r9079.
 2006-02-17 05:07:28 +00:00
George Bosilca
1740ab1209 I just wanted to add a flag not to replace the old ones. 
This commit was SVN r9076.
 2006-02-16 22:39:06 +00:00
George Bosilca
09aa367563 When we create a convertor if the data size is zero mark it as completed. That avoid 
problems with 0 length datatype used for pack/unpack.

This commit was SVN r9075.
 2006-02-16 22:36:42 +00:00
George Bosilca
eef3b6b2ba A lot better with a comma inside ... 
This commit was SVN r9070.
 2006-02-16 21:13:17 +00:00
George Bosilca
7d9f1992f2 Print more information about the convertor in ompi_convertor_dump. 
This commit was SVN r9069.
 2006-02-16 21:01:29 +00:00
George Bosilca
745ec1057f Remove the compilation warning. 
This commit was SVN r9047.
 2006-02-15 15:07:40 +00:00
Galen Shipman
fca3a36bc0 if data is contiguous need_buffers should return 0, not 1 
This commit was SVN r9009.
 2006-02-13 17:49:31 +00:00
George Bosilca
bde56ce528 Cleanup the convertor construct function. Do not set any useless values, instead rely on the 
calls to convertor_create and convertor_prepare to finish setting up the correct values for
the convertor.

This commit was SVN r9008.
 2006-02-13 17:03:36 +00:00
George Bosilca
ecc3e00362 Various cleanups. 
This commit was SVN r9002.
 2006-02-12 21:36:07 +00:00
George Bosilca
8b525ef21b Include the count when computing the remote size. 
This commit was SVN r9001.
 2006-02-12 21:34:55 +00:00
George Bosilca
13170eec52 Make sure all datatype files have the correct updated copyright information. 
This commit was SVN r8998.
 2006-02-12 21:03:01 +00:00
George Bosilca
25020f58c1 Correct the MPI_Indexed bug. Actually it can be triggered by any type of 
data, if the packed data is contiguous and the unpacked one is not. The
problem was that the unpack function did not update the bConverted value
to the real number of bytes extracted from the iovecs, but limit the update
to the number of bytes moved into the user buffers. So, if there were some
bytes copied in the internal pending buffer, they were lost because the next
convertor instead of using the correct bConverted (with the total number of
bytes) has only the transfered number. The result of this bug, is a shift to
left by few bytes of the values in the user buffers.
Add a protection against such kind of errors in the new_position.c.

This commit was SVN r8997.
 2006-02-12 20:19:22 +00:00
Brian Barrett
566a050c23 Next step in the project split, mainly source code re-arranging 
- move files out of toplevel include/ and etc/, moving it into the
    sub-projects
  - rather than including config headers with <project>/include, 
    have them as <project>
  - require all headers to be included with a project prefix, with
    the exception of the config headers ({opal,orte,ompi}_config.h
    mpi.h, and mpif.h)

This commit was SVN r8985.
 2006-02-12 01:33:29 +00:00
George Bosilca
3171cc1aee Remove unused variable. 
This commit was SVN r8977.
 2006-02-11 00:13:09 +00:00
George Bosilca
c9e83658dd Datatype engine went into a "slim-fast" treatement. That's only the first stage ... 
This commit was SVN r8965.
 2006-02-10 20:56:30 +00:00
George Bosilca
f96f9afc27 Copy the packed buffer when we duplicate a basic datatype. 
This commit was SVN r8931.
 2006-02-08 05:08:39 +00:00
Brian Barrett
e0a814d3a7 * adjust the buffer position on predefined datatypes as well as user 
datatypes when creating from a buffer.

This commit was SVN r8917.
 2006-02-07 12:14:41 +00:00
George Bosilca
9d969f2b53 Update the pack/unpack of data-type description (used for one sided). Correctly handle 
the predefined types, move the packed_description into the data-type structure.

This commit was SVN r8905.
 2006-02-06 19:10:19 +00:00
George Bosilca
697849c694 Add a function to check if the data-type is predefined or not (ompi_ddt_is_predefined). 
This commit was SVN r8902.
 2006-02-06 17:54:29 +00:00
George Bosilca
6a633829cb Pack/unpack datatype representaion for one-sided communications. 
This commit was SVN r8887.
 2006-02-02 19:49:29 +00:00
George Bosilca
1c667ea45a Start tracking down all useless things. Here the logic is that the MPI request already has 
increased the datatype ref count, and there is no way to have a convertor without a valid
request, so there is no need for the convertor to increase the reference count of the datatype
again. Therefore, if one want to use the datatype for something else than MPI request, one
has to manage the reference counts outside the datatype layer.

This commit was SVN r8769.
 2006-01-20 17:10:29 +00:00
George Bosilca
4ae8c291c3 Add MPI_SIGNED_CHAR. 
This commit was SVN r8618.
 2005-12-31 14:54:57 +00:00
George Bosilca
daceaf6219 The checksum code wasn't supposed to be turned on by default ... 
This commit was SVN r8510.
 2005-12-15 18:38:59 +00:00
George Bosilca
e008196f50 First implementation of the checksum in the datatype engine. I'm using adler 32 
the same algorithm as in the zlib library, as it is supposed to be the fastest
one and still give good error detection. For those who want to activate the
checksum, there is a define at the bottom of the datatype_internal.h file.

This commit was SVN r8509.
 2005-12-15 18:36:48 +00:00
George Bosilca
4af8085d3b The return is 0 or 1 so an int it's enought. Thanks Brian for catching this one. 
This commit was SVN r8504.
 2005-12-15 06:03:55 +00:00
George Bosilca
f8e875c8fb Always export this function. 
This commit was SVN r8438.
 2005-12-10 22:03:27 +00:00
Rich Graham
137703da29 add some comments. 
This commit was SVN r8376.
 2005-12-03 00:14:58 +00:00
George Bosilca
b47aa5c142 If pending_length is not zero the datatype engine believe that a partial conversion is still in the buffer attached to the convertor. We have to force it to zero for all new convertors. This bug is still half fixed until I figure out a correct way to do it when we clone a convertor ... Anyway, this future correction will not affect open mpi. 
This commit was SVN r8324.
 2005-11-30 00:16:58 +00:00
George Bosilca
cae4b603f4 First cut of the checksum capabilities for the convertor. Right now it's always set to zero, but it will change shortly. 
This commit was SVN r8305.
 2005-11-29 06:33:21 +00:00
Jeff Squyres
ebd97afdac - Make the types of the MCA param variables be "int", not "int32_t" 
- Separate out the registration of the MCA params into a standalong
  function that is invoked by ompi_mpi_init() (so that ompi_info can
  see these params)
- Rename the params to "mpi_ddt_*" instead of "datatype_*" so that
  they fit into the common naming scheme

This commit was SVN r8196.
 2005-11-18 22:51:11 +00:00
George Bosilca
9359be2190 Add the new dt_copy.c to the Makefile.am 
Remove the old function from the dt_unpack.c and activate the new one from dt_copy.c
Add a MCA param ompi_copy_debug to get messages about the local memory copies in the new function.
Slightly change the prototype of the function to keep the compilers happy on some platforms.

This commit was SVN r8142.
 2005-11-13 23:00:50 +00:00
George Bosilca
08a29da9a9 New version of the send data to self for the case where both the send and the receive data are 
the same.

This commit was SVN r8141.
 2005-11-13 22:57:28 +00:00
George Bosilca
405d9794f8 Somehow I miss to remove one of the previous definition for the unavailable data. 
This commit was SVN r8080.
 2005-11-10 02:59:20 +00:00
George Bosilca
55051b81c4 Activate the protection against unavailable datatypes. They get a flag DT_FLAG_UNAVAILABLE. We check now this flag in all the send/recv operations via the macros on mpi/c/bindings.h. 
This flag is inherited by all datatypes create with unavailable datatypes. Basically, we let the user create the wrong datatype but we dont let him using it for any pt2pt communications or any pack/unpack.

This commit was SVN r8069.
 2005-11-09 23:43:41 +00:00
George Bosilca
a6fdc2b2b4 Turn off the missing data-type message on MPI_Init. 
This commit was SVN r8056.
 2005-11-09 17:34:44 +00:00
George Bosilca
de0676a3dd Do not do any local copy into the storage if this convertor is finished. This is usefull in the Pack/Unpack case when there is more data in the packed buffer than the one we try to extract. 
This commit was SVN r8054.
 2005-11-09 07:46:12 +00:00
George Bosilca
025a8a04c5 More optimization of the data-type description are now possibles. Some corner cases are corrected. As a result we discover more accurately the contiguous part of the data memory layout. 
This commit was SVN r8051.
 2005-11-09 00:02:39 +00:00
George Bosilca
63ba3bde11 Allow the convertor to remember the last trucated unpack. If the same convertor is used for the next unpack it will put the data back correctly. However, if the BTL/PTL create a new convertor, even if it clone the last one this magic will not happens ! 
This commit was SVN r8048.
 2005-11-08 21:48:48 +00:00
George Bosilca
cdfe5e71fd By default there is no pending length on the convertor. 
This commit was SVN r8047.
 2005-11-08 21:45:45 +00:00
George Bosilca
7582ae3ef1 A simpler way to get output about the packing/unpacking. Now there are 2 MCA parameters datatype_pack_debug and datatype_unpack_debug. When they are set to 1 the ddt engine will dump a lot of messages. Dont turn them to one by default. But if you notice any problems in the ddt you can turn them to one and send me the output. 
First step toward adding memory to the convertor. It will be able to keep partial basic datatype between calls ...

This commit was SVN r8042.
 2005-11-08 17:49:51 +00:00
George Bosilca
2b9b5500b9 Change some variable's names. 
This commit was SVN r8041.
 2005-11-08 17:44:56 +00:00
George Bosilca
c63e4dcef9 When we finish one of the loops take care of the index of the begining of the loop. If it's -1 then we just complete the full datatype ... therefore we have to do something special. 
This commit was SVN r8040.
 2005-11-08 16:53:31 +00:00
George Bosilca
579398a135 Change some variable names (from pSrc to something more clear like user_memory and/or packed_buffer). 
This commit was SVN r8034.
 2005-11-08 03:12:58 +00:00