-
Notifications
You must be signed in to change notification settings - Fork 27
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
added non-blocking root communicator #1478
base: develop
Are you sure you want to change the base?
Changes from 19 commits
282ae1c
43b02da
7969aa8
14f802e
c04745a
8275b6d
0c0da25
1c120d3
c74d1f4
53c0f4e
173bb41
d946bf9
3a6c50a
d3bf633
ffb6d91
3a3a52c
5f416fd
ae83780
fe58609
a6b8c63
92fd4c8
97579e5
a0f3548
a3863ea
4a62e2b
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -30,6 +30,18 @@ namespace lumberjack | |
*/ | ||
const char* mpiBlockingReceiveMessages(MPI_Comm comm); | ||
|
||
/*! | ||
***************************************************************************** | ||
* \brief Receives any Message sent to this rank, if there are any messages | ||
* that are sent. Returns null if no messages are sent. | ||
* | ||
* \param [in] comm The MPI Communicator. | ||
* \param [in] tag The MPI tag to use for communication. When set to zero, | ||
* MPI communication uses default LJ_Tag. | ||
***************************************************************************** | ||
*/ | ||
const char* mpiNonBlockingReceiveMessages(MPI_Comm comm); | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Should we update the docs for this function (and There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. @gberg617 We should update docs as @kennyweiss suggested. There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Added a sentence about this to the function documentation, |
||
|
||
/*! | ||
***************************************************************************** | ||
* \brief Sends all Message sent to the given rank. | ||
|
@@ -40,6 +52,8 @@ const char* mpiBlockingReceiveMessages(MPI_Comm comm); | |
* \param [in] destinationRank Where the Message classes is being sent. | ||
* \param [in,out] packedMessagesToBeSent All of the Message classes to be sent | ||
* packed together. | ||
* \param [in] tag The MPI tag to use for communication. When set to zero, | ||
* MPI communication uses the default Lumberjack Tag. | ||
***************************************************************************** | ||
*/ | ||
void mpiNonBlockingSendMessages(MPI_Comm comm, | ||
|
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,96 @@ | ||
// Copyright (c) 2017-2024, Lawrence Livermore National Security, LLC and | ||
// other Axom Project Developers. See the top-level LICENSE file for details. | ||
// | ||
// SPDX-License-Identifier: (BSD-3-Clause) | ||
|
||
/*! | ||
****************************************************************************** | ||
* | ||
* \file NonCollectiveRootCommunicator.cpp | ||
* | ||
* \brief Implementation of the NonCollectiveRootCommunicator class. | ||
* | ||
****************************************************************************** | ||
*/ | ||
|
||
#include <limits> | ||
|
||
#include "axom/lumberjack/NonCollectiveRootCommunicator.hpp" | ||
#include "axom/lumberjack/MPIUtility.hpp" | ||
|
||
namespace axom | ||
{ | ||
namespace lumberjack | ||
{ | ||
void NonCollectiveRootCommunicator::initialize(MPI_Comm comm, int ranksLimit) | ||
{ | ||
MPI_Comm_dup(comm, &m_mpiComm); | ||
MPI_Comm_rank(m_mpiComm, &m_mpiCommRank); | ||
MPI_Comm_size(m_mpiComm, &m_mpiCommSize); | ||
m_ranksLimit = ranksLimit; | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Presumably, this needs to be a positive number greater than or equal to 1. Should we check it w/ a There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I put in a std::cerr to avoid including anything from SLIC in Lumberjack. |
||
} | ||
|
||
void NonCollectiveRootCommunicator::finalize() { MPI_Comm_free(&m_mpiComm); } | ||
|
||
int NonCollectiveRootCommunicator::rank() { return m_mpiCommRank; } | ||
|
||
void NonCollectiveRootCommunicator::ranksLimit(int value) | ||
{ | ||
m_ranksLimit = value; | ||
} | ||
|
||
int NonCollectiveRootCommunicator::ranksLimit() { return m_ranksLimit; } | ||
|
||
int NonCollectiveRootCommunicator::numPushesToFlush() { return 1; } | ||
|
||
void NonCollectiveRootCommunicator::push( | ||
const char* packedMessagesToBeSent, | ||
std::vector<const char*>& receivedPackedMessages) | ||
{ | ||
if(m_mpiCommRank == 0) | ||
{ | ||
const char* currPackedMessages = nullptr; | ||
bool receive_messages = true; | ||
while(receive_messages) | ||
{ | ||
currPackedMessages = mpiNonBlockingReceiveMessages(m_mpiComm); | ||
|
||
if(isPackedMessagesEmpty(currPackedMessages)) | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. It looks odd to me that There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I believe |
||
{ | ||
if(currPackedMessages == nullptr) | ||
{ | ||
receive_messages = false; | ||
} | ||
else | ||
{ | ||
delete[] currPackedMessages; | ||
} | ||
} | ||
else | ||
{ | ||
receivedPackedMessages.push_back(currPackedMessages); | ||
} | ||
|
||
currPackedMessages = nullptr; | ||
} | ||
} | ||
else | ||
{ | ||
if(isPackedMessagesEmpty(packedMessagesToBeSent) == false) | ||
{ | ||
mpiNonBlockingSendMessages(m_mpiComm, 0, packedMessagesToBeSent); | ||
} | ||
} | ||
} | ||
|
||
bool NonCollectiveRootCommunicator::isOutputNode() | ||
{ | ||
if(m_mpiCommRank == 0) | ||
{ | ||
return true; | ||
} | ||
return false; | ||
} | ||
|
||
} // end namespace lumberjack | ||
} // end namespace axom |
Original file line number | Diff line number | Diff line change | ||||||
---|---|---|---|---|---|---|---|---|
@@ -0,0 +1,135 @@ | ||||||||
// Copyright (c) 2017-2024, Lawrence Livermore National Security, LLC and | ||||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. After merging in There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Fixed. |
||||||||
// other Axom Project Developers. See the top-level LICENSE file for details. | ||||||||
// | ||||||||
// SPDX-License-Identifier: (BSD-3-Clause) | ||||||||
|
||||||||
/*! | ||||||||
******************************************************************************* | ||||||||
* \file NonCollectiveRootCommunicator.hpp | ||||||||
* | ||||||||
* \brief This file contains the class definition of the | ||||||||
* NonCollectiveRootCommunicator. | ||||||||
******************************************************************************* | ||||||||
*/ | ||||||||
|
||||||||
#ifndef NONCOLLECTIVEROOTCOMMUNICATOR_HPP | ||||||||
#define NONCOLLECTIVEROOTCOMMUNICATOR_HPP | ||||||||
|
||||||||
#include "axom/lumberjack/Lumberjack.hpp" | ||||||||
#include "axom/lumberjack/Communicator.hpp" | ||||||||
|
||||||||
namespace axom | ||||||||
{ | ||||||||
namespace lumberjack | ||||||||
{ | ||||||||
/*! | ||||||||
******************************************************************************* | ||||||||
* \class NonCollectiveRootCommunicator | ||||||||
* | ||||||||
* \brief Based off of RootCommunicator. This communicator pushes | ||||||||
messages from any rank to root non-collectively, if any messages are sent. | ||||||||
******************************************************************************* | ||||||||
*/ | ||||||||
class NonCollectiveRootCommunicator : public axom::lumberjack::Communicator | ||||||||
{ | ||||||||
public: | ||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief Called to initialize the Communicator. | ||||||||
* | ||||||||
* This performs any setup work the Communicator needs before doing any work. | ||||||||
* It is required that this is called before using the Communicator. | ||||||||
* | ||||||||
* \param [in] comm The MPI Communicator | ||||||||
* \param [in] ranksLimit Limit on how many ranks are individually tracked per | ||||||||
* Message. | ||||||||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Suggestion:
Suggested change
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. Fixed. |
||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
void initialize(MPI_Comm comm, int ranksLimit); | ||||||||
|
||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief Called to finalize the Communicator. | ||||||||
* | ||||||||
* This performs any cleanup work the Communicator needs to do before going | ||||||||
* away.It is required that this is the last function called by the | ||||||||
* Communicator. | ||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
void finalize(); | ||||||||
|
||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief Returns the MPI rank of this node | ||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
int rank(); | ||||||||
|
||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief Sets the rank limit. | ||||||||
* | ||||||||
* This is the limit on how many ranks generated a given message are | ||||||||
* individually tracked per Message. After the limit has been reached, only | ||||||||
* the Message::rankCount is incremented. | ||||||||
* | ||||||||
* \param [in] value Limits how many ranks are tracked per Message. | ||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
void ranksLimit(int value); | ||||||||
|
||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief Returns the rank limit. | ||||||||
* | ||||||||
* This is the limit on how many ranks generated a given message are | ||||||||
* individually tracked per Message. After the limit has been reached, only | ||||||||
* the Message::rankCount is incremented. | ||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
int ranksLimit(); | ||||||||
|
||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief Function used by the Lumberjack class to indicate how many | ||||||||
* individual pushes fully flush all currently held Message classes to the | ||||||||
* root node. The Communicator class's tree structure dictates this. | ||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
int numPushesToFlush(); | ||||||||
|
||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief This pushes all messages to the root node. | ||||||||
* | ||||||||
* All messages are pushed to the root node. This is the same as | ||||||||
* RootCommunicator::pushMessagesFully for this Communicator. | ||||||||
* | ||||||||
* \param [in] packedMessagesToBeSent All of this rank's Message classes | ||||||||
* packed into a single buffer. | ||||||||
* \param [in,out] receivedPackedMessages Received packed message buffers from | ||||||||
* this nodes children. | ||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
void push(const char* packedMessagesToBeSent, | ||||||||
std::vector<const char*>& receivedPackedMessages); | ||||||||
|
||||||||
/*! | ||||||||
***************************************************************************** | ||||||||
* \brief Function used by the Lumberjack to indicate whether this node should | ||||||||
* be outputting messages. Only the root node outputs messages. | ||||||||
***************************************************************************** | ||||||||
*/ | ||||||||
bool isOutputNode(); | ||||||||
|
||||||||
private: | ||||||||
MPI_Comm m_mpiComm; | ||||||||
int m_mpiCommRank; | ||||||||
int m_mpiCommSize; | ||||||||
int m_ranksLimit; | ||||||||
}; | ||||||||
|
||||||||
} // end namespace lumberjack | ||||||||
} // end namespace axom | ||||||||
|
||||||||
#endif |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
When we return from a non-blocking probe, there could be sent messages that haven't arrived yet. I suggest replacing "are sent" with "have arrived." It's pedantic but can avoid confusion when the unexpected happens.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Fixed.