- •Contents
- •List of Figures
- •List of Tables
- •Acknowledgments
- •Introduction to MPI
- •Overview and Goals
- •Background of MPI-1.0
- •Background of MPI-1.1, MPI-1.2, and MPI-2.0
- •Background of MPI-1.3 and MPI-2.1
- •Background of MPI-2.2
- •Who Should Use This Standard?
- •What Platforms Are Targets For Implementation?
- •What Is Included In The Standard?
- •What Is Not Included In The Standard?
- •Organization of this Document
- •MPI Terms and Conventions
- •Document Notation
- •Naming Conventions
- •Semantic Terms
- •Data Types
- •Opaque Objects
- •Array Arguments
- •State
- •Named Constants
- •Choice
- •Addresses
- •Language Binding
- •Deprecated Names and Functions
- •Fortran Binding Issues
- •C Binding Issues
- •C++ Binding Issues
- •Functions and Macros
- •Processes
- •Error Handling
- •Implementation Issues
- •Independence of Basic Runtime Routines
- •Interaction with Signals
- •Examples
- •Point-to-Point Communication
- •Introduction
- •Blocking Send and Receive Operations
- •Blocking Send
- •Message Data
- •Message Envelope
- •Blocking Receive
- •Return Status
- •Passing MPI_STATUS_IGNORE for Status
- •Data Type Matching and Data Conversion
- •Type Matching Rules
- •Type MPI_CHARACTER
- •Data Conversion
- •Communication Modes
- •Semantics of Point-to-Point Communication
- •Buffer Allocation and Usage
- •Nonblocking Communication
- •Communication Request Objects
- •Communication Initiation
- •Communication Completion
- •Semantics of Nonblocking Communications
- •Multiple Completions
- •Non-destructive Test of status
- •Probe and Cancel
- •Persistent Communication Requests
- •Send-Receive
- •Null Processes
- •Datatypes
- •Derived Datatypes
- •Type Constructors with Explicit Addresses
- •Datatype Constructors
- •Subarray Datatype Constructor
- •Distributed Array Datatype Constructor
- •Address and Size Functions
- •Lower-Bound and Upper-Bound Markers
- •Extent and Bounds of Datatypes
- •True Extent of Datatypes
- •Commit and Free
- •Duplicating a Datatype
- •Use of General Datatypes in Communication
- •Correct Use of Addresses
- •Decoding a Datatype
- •Examples
- •Pack and Unpack
- •Canonical MPI_PACK and MPI_UNPACK
- •Collective Communication
- •Introduction and Overview
- •Communicator Argument
- •Applying Collective Operations to Intercommunicators
- •Barrier Synchronization
- •Broadcast
- •Example using MPI_BCAST
- •Gather
- •Examples using MPI_GATHER, MPI_GATHERV
- •Scatter
- •Examples using MPI_SCATTER, MPI_SCATTERV
- •Example using MPI_ALLGATHER
- •All-to-All Scatter/Gather
- •Global Reduction Operations
- •Reduce
- •Signed Characters and Reductions
- •MINLOC and MAXLOC
- •All-Reduce
- •Process-local reduction
- •Reduce-Scatter
- •MPI_REDUCE_SCATTER_BLOCK
- •MPI_REDUCE_SCATTER
- •Scan
- •Inclusive Scan
- •Exclusive Scan
- •Example using MPI_SCAN
- •Correctness
- •Introduction
- •Features Needed to Support Libraries
- •MPI's Support for Libraries
- •Basic Concepts
- •Groups
- •Contexts
- •Intra-Communicators
- •Group Management
- •Group Accessors
- •Group Constructors
- •Group Destructors
- •Communicator Management
- •Communicator Accessors
- •Communicator Constructors
- •Communicator Destructors
- •Motivating Examples
- •Current Practice #1
- •Current Practice #2
- •(Approximate) Current Practice #3
- •Example #4
- •Library Example #1
- •Library Example #2
- •Inter-Communication
- •Inter-communicator Accessors
- •Inter-communicator Operations
- •Inter-Communication Examples
- •Caching
- •Functionality
- •Communicators
- •Windows
- •Datatypes
- •Error Class for Invalid Keyval
- •Attributes Example
- •Naming Objects
- •Formalizing the Loosely Synchronous Model
- •Basic Statements
- •Models of Execution
- •Static communicator allocation
- •Dynamic communicator allocation
- •The General case
- •Process Topologies
- •Introduction
- •Virtual Topologies
- •Embedding in MPI
- •Overview of the Functions
- •Topology Constructors
- •Cartesian Constructor
- •Cartesian Convenience Function: MPI_DIMS_CREATE
- •General (Graph) Constructor
- •Distributed (Graph) Constructor
- •Topology Inquiry Functions
- •Cartesian Shift Coordinates
- •Partitioning of Cartesian structures
- •Low-Level Topology Functions
- •An Application Example
- •MPI Environmental Management
- •Implementation Information
- •Version Inquiries
- •Environmental Inquiries
- •Tag Values
- •Host Rank
- •IO Rank
- •Clock Synchronization
- •Memory Allocation
- •Error Handling
- •Error Handlers for Communicators
- •Error Handlers for Windows
- •Error Handlers for Files
- •Freeing Errorhandlers and Retrieving Error Strings
- •Error Codes and Classes
- •Error Classes, Error Codes, and Error Handlers
- •Timers and Synchronization
- •Startup
- •Allowing User Functions at Process Termination
- •Determining Whether MPI Has Finished
- •Portable MPI Process Startup
- •The Info Object
- •Process Creation and Management
- •Introduction
- •The Dynamic Process Model
- •Starting Processes
- •The Runtime Environment
- •Process Manager Interface
- •Processes in MPI
- •Starting Processes and Establishing Communication
- •Reserved Keys
- •Spawn Example
- •Manager-worker Example, Using MPI_COMM_SPAWN.
- •Establishing Communication
- •Names, Addresses, Ports, and All That
- •Server Routines
- •Client Routines
- •Name Publishing
- •Reserved Key Values
- •Client/Server Examples
- •Ocean/Atmosphere - Relies on Name Publishing
- •Simple Client-Server Example.
- •Other Functionality
- •Universe Size
- •Singleton MPI_INIT
- •MPI_APPNUM
- •Releasing Connections
- •Another Way to Establish MPI Communication
- •One-Sided Communications
- •Introduction
- •Initialization
- •Window Creation
- •Window Attributes
- •Communication Calls
- •Examples
- •Accumulate Functions
- •Synchronization Calls
- •Fence
- •General Active Target Synchronization
- •Lock
- •Assertions
- •Examples
- •Error Handling
- •Error Handlers
- •Error Classes
- •Semantics and Correctness
- •Atomicity
- •Progress
- •Registers and Compiler Optimizations
- •External Interfaces
- •Introduction
- •Generalized Requests
- •Examples
- •Associating Information with Status
- •MPI and Threads
- •General
- •Initialization
- •Introduction
- •File Manipulation
- •Opening a File
- •Closing a File
- •Deleting a File
- •Resizing a File
- •Preallocating Space for a File
- •Querying the Size of a File
- •Querying File Parameters
- •File Info
- •Reserved File Hints
- •File Views
- •Data Access
- •Data Access Routines
- •Positioning
- •Synchronism
- •Coordination
- •Data Access Conventions
- •Data Access with Individual File Pointers
- •Data Access with Shared File Pointers
- •Noncollective Operations
- •Collective Operations
- •Seek
- •Split Collective Data Access Routines
- •File Interoperability
- •Datatypes for File Interoperability
- •Extent Callback
- •Datarep Conversion Functions
- •Matching Data Representations
- •Consistency and Semantics
- •File Consistency
- •Random Access vs. Sequential Files
- •Progress
- •Collective File Operations
- •Type Matching
- •Logical vs. Physical File Layout
- •File Size
- •Examples
- •Asynchronous I/O
- •I/O Error Handling
- •I/O Error Classes
- •Examples
- •Subarray Filetype Constructor
- •Requirements
- •Discussion
- •Logic of the Design
- •Examples
- •MPI Library Implementation
- •Systems with Weak Symbols
- •Systems Without Weak Symbols
- •Complications
- •Multiple Counting
- •Linker Oddities
- •Multiple Levels of Interception
- •Deprecated Functions
- •Deprecated since MPI-2.0
- •Deprecated since MPI-2.2
- •Language Bindings
- •Overview
- •Design
- •C++ Classes for MPI
- •Class Member Functions for MPI
- •Semantics
- •C++ Datatypes
- •Communicators
- •Exceptions
- •Mixed-Language Operability
- •Problems With Fortran Bindings for MPI
- •Problems Due to Strong Typing
- •Problems Due to Data Copying and Sequence Association
- •Special Constants
- •Fortran 90 Derived Types
- •A Problem with Register Optimization
- •Basic Fortran Support
- •Extended Fortran Support
- •The mpi Module
- •No Type Mismatch Problems for Subroutines with Choice Arguments
- •Additional Support for Fortran Numeric Intrinsic Types
- •Language Interoperability
- •Introduction
- •Assumptions
- •Initialization
- •Transfer of Handles
- •Status
- •MPI Opaque Objects
- •Datatypes
- •Callback Functions
- •Error Handlers
- •Reduce Operations
- •Addresses
- •Attributes
- •Extra State
- •Constants
- •Interlanguage Communication
- •Language Bindings Summary
- •Groups, Contexts, Communicators, and Caching Fortran Bindings
- •External Interfaces C++ Bindings
- •Change-Log
- •Bibliography
- •Examples Index
- •MPI Declarations Index
- •MPI Function Index
410 |
CHAPTER 13. I/O |
1MPI_FILE_IREAD_AT is a nonblocking version of the MPI_FILE_READ_AT interface.
2
3
4
5
6
7
8
9
10
11
12
13
MPI_FILE_IWRITE_AT(fh, o set, buf, count, datatype, request)
INOUT |
fh |
le handle (handle) |
IN |
o set |
le o set (integer) |
IN |
buf |
initial address of bu er (choice) |
IN |
count |
number of elements in bu er (integer) |
IN |
datatype |
datatype of each bu er element (handle) |
OUT |
request |
request object (handle) |
14 |
int MPI_File_iwrite_at(MPI_File fh, MPI_Offset offset, void *buf, |
|
|
||
15 |
int count, MPI_Datatype datatype, MPI_Request *request) |
|
|
||
16 |
MPI_FILE_IWRITE_AT(FH, OFFSET, BUF, COUNT, DATATYPE, REQUEST, IERROR) |
|
17 |
||
<type> BUF(*) |
||
18 |
||
INTEGER FH, COUNT, DATATYPE, REQUEST, IERROR |
||
19 |
||
INTEGER(KIND=MPI_OFFSET_KIND) OFFSET |
||
20 |
||
|
||
21 |
fMPI::Request MPI::File::Iwrite_at(MPI::Offset offset, const void* buf, |
|
|
||
22 |
int count, const MPI::Datatype& datatype) (binding deprecated, see |
|
|
||
23 |
Section 15.2) g |
|
24 |
||
MPI_FILE_IWRITE_AT is a nonblocking version of the MPI_FILE_WRITE_AT interface. |
||
25 |
||
|
||
26 |
|
2713.4.3 Data Access with Individual File Pointers
28MPI maintains one individual le pointer per process per le handle. The current value
29of this pointer implicitly speci es the o set in the data access routines described in this
30section. These routines only use and update the individual le pointers maintained by MPI.
31The shared le pointer is not used nor updated.
32The individual le pointer routines have the same semantics as the data access with
33explicit o set routines described in Section 13.4.2, page 407, with the following modi cation:
34
35the o set is de ned to be the current value of the MPI-maintained individual le
36pointer.
37
After an individual le pointer operation is initiated, the individual le pointer is updated
38
to point to the next etype after the last one that will be accessed. The le pointer is updated
39
relative to the current view of the le.
If MPI_MODE_SEQUENTIAL mode was speci ed when the le was opened, it is erroneous to call the routines in this section, with the exception of MPI_FILE_GET_BYTE_OFFSET.
13.4. DATA ACCESS |
411 |
MPI_FILE_READ(fh, buf, count, datatype, status)
INOUT |
fh |
le handle (handle) |
OUT |
buf |
initial address of bu er (choice) |
IN |
count |
number of elements in bu er (integer) |
IN |
datatype |
datatype of each bu er element (handle) |
OUT |
status |
status object (Status) |
int MPI_File_read(MPI_File fh, void *buf, int count, MPI_Datatype datatype, MPI_Status *status)
MPI_FILE_READ(FH, BUF, COUNT, DATATYPE, STATUS, IERROR) <type> BUF(*)
INTEGER FH, COUNT, DATATYPE, STATUS(MPI_STATUS_SIZE), IERROR
fvoid MPI::File::Read(void* buf, int count, const MPI::Datatype& datatype, MPI::Status& status) (binding deprecated, see Section 15.2) g
fvoid MPI::File::Read(void* buf, int count, const MPI::Datatype& datatype)
(binding deprecated, see Section 15.2) g
MPI_FILE_READ reads a le using the individual le pointer.
Example 13.2 The following Fortran code fragment is an example of reading a le until the end of le is reached:
!Read a preexisting input file until all data has been read.
!Call routine "process_input" if all requested data is read.
!The Fortran 90 "exit" statement exits the loop.
integer |
bufsize, numread, totprocessed, status(MPI_STATUS_SIZE) |
parameter |
(bufsize=100) |
real |
localbuffer(bufsize) |
call MPI_FILE_OPEN( MPI_COMM_WORLD, 'myoldfile', & MPI_MODE_RDONLY, MPI_INFO_NULL, myfh, ierr )
call MPI_FILE_SET_VIEW( myfh, 0, MPI_REAL, MPI_REAL, 'native', & MPI_INFO_NULL, ierr )
totprocessed = 0 do
call MPI_FILE_READ( myfh, localbuffer, bufsize, MPI_REAL, & status, ierr )
call MPI_GET_COUNT( status, MPI_REAL, numread, ierr ) call process_input( localbuffer, numread ) totprocessed = totprocessed + numread
if ( numread < bufsize ) exit
enddo
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
write(6,1001) numread, bufsize, totprocessed |
48 |
1
2
3
4
5
6
7
412 |
CHAPTER 13. I/O |
1001 format( "No more data: read", I3, "and expected", I3, & "Processed total of", I6, "before terminating job." )
call MPI_FILE_CLOSE( myfh, ierr )
8
9
10
11
12
13
14
15
16
MPI_FILE_READ_ALL(fh, buf, count, datatype, status)
INOUT |
fh |
le handle (handle) |
OUT |
buf |
initial address of bu er (choice) |
IN |
count |
number of elements in bu er (integer) |
IN |
datatype |
datatype of each bu er element (handle) |
OUT |
status |
status object (Status) |
int MPI_File_read_all(MPI_File fh, void *buf, int count,
17
MPI_Datatype datatype, MPI_Status *status)
18
19MPI_FILE_READ_ALL(FH, BUF, COUNT, DATATYPE, STATUS, IERROR)
20<type> BUF(*)
21INTEGER FH, COUNT, DATATYPE, STATUS(MPI_STATUS_SIZE), IERROR
22 |
fvoid MPI::File::Read_all(void* buf, int count, |
|
23 |
||
24 |
const MPI::Datatype& datatype, MPI::Status& status) (binding |
|
deprecated, see Section 15.2) g |
||
25 |
||
26 |
fvoid MPI::File::Read_all(void* buf, int count, |
|
27 |
||
|
const MPI::Datatype& datatype) (binding deprecated, see |
|
28 |
Section 15.2) g |
|
29 |
30
31
32
33
MPI_FILE_READ_ALL is a collective version of the blocking MPI_FILE_READ interface.
MPI_FILE_WRITE(fh, buf, count, datatype, status)
34 |
INOUT |
fh |
le handle (handle) |
|
35 |
IN |
buf |
initial address of bu er (choice) |
|
|
|
|
||
36 |
IN |
count |
number of elements in bu er (integer) |
|
37 |
||||
|
|
|
||
38 |
IN |
datatype |
datatype of each bu er element (handle) |
|
39 |
OUT |
status |
status object (Status) |
|
|
|
|
40
41
int MPI_File_write(MPI_File fh, void *buf, int count,
42
MPI_Datatype datatype, MPI_Status *status)
43
44MPI_FILE_WRITE(FH, BUF, COUNT, DATATYPE, STATUS, IERROR)
45<type> BUF(*)
46INTEGER FH, COUNT, DATATYPE, STATUS(MPI_STATUS_SIZE), IERROR
47
48
13.4. DATA ACCESS |
413 |
fvoid MPI::File::Write(const void* buf, int count,
const MPI::Datatype& datatype, MPI::Status& status) (binding deprecated, see Section 15.2) g
fvoid MPI::File::Write(const void* buf, int count,
const MPI::Datatype& datatype) (binding deprecated, see Section 15.2) g
MPI_FILE_WRITE writes a le using the individual le pointer.
MPI_FILE_WRITE_ALL(fh, buf, count, datatype, status)
INOUT |
fh |
le handle (handle) |
IN |
buf |
initial address of bu er (choice) |
IN |
count |
number of elements in bu er (integer) |
IN |
datatype |
datatype of each bu er element (handle) |
OUT |
status |
status object (Status) |
int MPI_File_write_all(MPI_File fh, void *buf, int count, MPI_Datatype datatype, MPI_Status *status)
MPI_FILE_WRITE_ALL(FH, BUF, COUNT, DATATYPE, STATUS, IERROR) <type> BUF(*)
INTEGER FH, COUNT, DATATYPE, STATUS(MPI_STATUS_SIZE), IERROR
fvoid MPI::File::Write_all(const void* buf, int count,
const MPI::Datatype& datatype, MPI::Status& status) (binding deprecated, see Section 15.2) g
fvoid MPI::File::Write_all(const void* buf, int count,
const MPI::Datatype& datatype) (binding deprecated, see Section 15.2) g
MPI_FILE_WRITE_ALL is a collective version of the blocking MPI_FILE_WRITE inter-
face.
MPI_FILE_IREAD(fh, buf, count, datatype, request)
INOUT |
fh |
le handle (handle) |
OUT |
buf |
initial address of bu er (choice) |
IN |
count |
number of elements in bu er (integer) |
IN |
datatype |
datatype of each bu er element (handle) |
OUT |
request |
request object (handle) |
int MPI_File_iread(MPI_File fh, void *buf, int count, MPI_Datatype datatype, MPI_Request *request)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
MPI_FILE_IREAD(FH, BUF, COUNT, DATATYPE, REQUEST, IERROR) |
48 |
1
2
3
4
5
6
414 CHAPTER 13. I/O
<type> BUF(*)
INTEGER FH, COUNT, DATATYPE, REQUEST, IERROR
fMPI::Request MPI::File::Iread(void* buf, int count,
const MPI::Datatype& datatype) (binding deprecated, see Section 15.2) g
7MPI_FILE_IREAD is a nonblocking version of the MPI_FILE_READ interface.
8
9Example 13.3 The following Fortran code fragment illustrates le pointer update seman-
10 tics:
11
12! Read the first twenty real words in a file into two local
13! buffers. Note that when the first MPI_FILE_IREAD returns,
14! the file pointer has been updated to point to the
15! eleventh real word in the file.
16
17
integer |
bufsize, req1, req2 |
18integer, dimension(MPI_STATUS_SIZE) :: status1, status2
19parameter (bufsize=10)
20 |
real |
buf1(bufsize), buf2(bufsize) |
21 |
|
|
22
23
24
25
26
27
28
29
30
call MPI_FILE_OPEN( MPI_COMM_WORLD, 'myoldfile', & MPI_MODE_RDONLY, MPI_INFO_NULL, myfh, ierr )
call MPI_FILE_SET_VIEW( myfh, 0, MPI_REAL, MPI_REAL, 'native', & MPI_INFO_NULL, ierr )
call MPI_FILE_IREAD( myfh, buf1, bufsize, MPI_REAL, & req1, ierr )
call MPI_FILE_IREAD( myfh, buf2, bufsize, MPI_REAL, & req2, ierr )
31
32
33
call MPI_WAIT( req1, status1, ierr ) call MPI_WAIT( req2, status2, ierr )
34
35
36
37
38
39
40
41
42
43
44
45
46
call MPI_FILE_CLOSE( myfh, ierr )
MPI_FILE_IWRITE(fh, buf, count, datatype, request)
INOUT |
fh |
le handle (handle) |
IN |
buf |
initial address of bu er (choice) |
IN |
count |
number of elements in bu er (integer) |
IN |
datatype |
datatype of each bu er element (handle) |
OUT |
request |
request object (handle) |
int MPI_File_iwrite(MPI_File fh, void *buf, int count,
47
MPI_Datatype datatype, MPI_Request *request)
48
13.4. DATA ACCESS |
415 |
MPI_FILE_IWRITE(FH, BUF, COUNT, DATATYPE, REQUEST, IERROR) <type> BUF(*)
INTEGER FH, COUNT, DATATYPE, REQUEST, IERROR
fMPI::Request MPI::File::Iwrite(const void* buf, int count, const MPI::Datatype& datatype) (binding deprecated, see Section 15.2) g
MPI_FILE_IWRITE is a nonblocking version of the MPI_FILE_WRITE interface.
MPI_FILE_SEEK(fh, o set, whence)
INOUT |
fh |
le handle (handle) |
IN |
o set |
le o set (integer) |
IN |
whence |
update mode (state) |
int MPI_File_seek(MPI_File fh, MPI_Offset offset, int whence)
MPI_FILE_SEEK(FH, OFFSET, WHENCE, IERROR)
INTEGER FH, WHENCE, IERROR
INTEGER(KIND=MPI_OFFSET_KIND) OFFSET
fvoid MPI::File::Seek(MPI::Offset offset, int whence) (binding deprecated, see Section 15.2) g
MPI_FILE_SEEK updates the individual le pointer according to whence, which has the following possible values:
MPI_SEEK_SET: the pointer is set to o set
MPI_SEEK_CUR: the pointer is set to the current pointer position plus o set
MPI_SEEK_END: the pointer is set to the end of le plus o set
The o set can be negative, which allows seeking backwards. It is erroneous to seek to a negative position in the view.
MPI_FILE_GET_POSITION(fh, o set)
IN |
fh |
le handle (handle) |
OUT |
o set |
o set of individual pointer (integer) |
int MPI_File_get_position(MPI_File fh, MPI_Offset *offset)
MPI_FILE_GET_POSITION(FH, OFFSET, IERROR)
INTEGER FH, IERROR
INTEGER(KIND=MPI_OFFSET_KIND) OFFSET
fMPI::Offset MPI::File::Get_position() const (binding deprecated, see Section 15.2) g
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48