Parallel processing - Revision history

Francisco: /* Parallel processing via MPI */

2015-04-30T18:23:12Z

‎Parallel processing via MPI

Francisco: /* Parallel processing via MPI */

2015-04-30T18:22:39Z

‎Parallel processing via MPI

Francisco: /* Parallel processing via MPI */

2015-04-30T17:26:01Z

‎Parallel processing via MPI

Paulochambel: /* Parallel processing via MPI */

2014-02-17T13:49:05Z

‎Parallel processing via MPI

Paulochambel: /* Parallel processing via MPI */

2014-02-17T13:47:02Z

‎Parallel processing via MPI

Paulochambel: /* Parallel processing via MPI */

2014-02-17T13:45:28Z

‎Parallel processing via MPI

Paulochambel: /* Parallel processing via MPI */

2014-02-17T13:02:15Z

‎Parallel processing via MPI

Miranda ricardo: /* Parallel processing via MPI */

2014-01-22T22:41:01Z

‎Parallel processing via MPI

Miranda ricardo: /* Parallel processing via MPI */

2014-01-22T22:40:23Z

‎Parallel processing via MPI

Miranda ricardo: /* Parallel processing via MPI */

2014-01-22T22:38:50Z

‎Parallel processing via MPI

@@ Line 10: / Line 10: @@
 Parallel processing as it is presently implemented in Mohid, could not be achieved without [[object oriented programming]] philosophy, as each model is an instance of [[class Model]] and no changes, exception made to the implementation of the MPI communications calls needed to be added. Using this feature, computational speed was improved (varying from application to application), as now the whole model will take the same time as the slowest model to run plus the time to communicate with the other processes. Here, the network communication speed plays an important role, as it can become limiting. However, the amount of information passing between models, depending of course on the memory allocated for each model, has not yet proven to be big enough to make a 100 Mbps network connection time limiting.
-[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomains that communicate among them (2-way) via MPI. See the follow report [[http://www.mohid.com/PublicData/Products/Manuals/Mohid_MPI_DomainDecomposition_Doc.pdf MPI_DD_Report]] where a description of the MPI/Domain Decomposition functionality is described.
+[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomains that communicate among them (2-way) via MPI. See the follow report [[http://www.mohid.com/PublicData/Products/Manuals/Mohid_MPI_DomainDecomposition.pdf MPI_DD_Report]] where a description of the MPI/Domain Decomposition functionality is described.
 Find here information on how to [[setup a MOHID simulation using MPI]] and on [[compiling Mohid with MPI]].

@@ Line 10: / Line 10: @@
 Parallel processing as it is presently implemented in Mohid, could not be achieved without [[object oriented programming]] philosophy, as each model is an instance of [[class Model]] and no changes, exception made to the implementation of the MPI communications calls needed to be added. Using this feature, computational speed was improved (varying from application to application), as now the whole model will take the same time as the slowest model to run plus the time to communicate with the other processes. Here, the network communication speed plays an important role, as it can become limiting. However, the amount of information passing between models, depending of course on the memory allocated for each model, has not yet proven to be big enough to make a 100 Mbps network connection time limiting.
-[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomains that communicate among them (2-way) via MPI. See the follow report [https://www.dropbox.com/s/bepaeqz9453skve/Mohid_MPI_DomainDecomposition_Doc.pdf MPI_DD_Report] where a description of the MPI/Domain Decomposition functionality is described.
+[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomains that communicate among them (2-way) via MPI. See the follow report [[http://www.mohid.com/PublicData/Products/Manuals/Mohid_MPI_DomainDecomposition_Doc.pdf MPI_DD_Report]] where a description of the MPI/Domain Decomposition functionality is described.
 Find here information on how to [[setup a MOHID simulation using MPI]] and on [[compiling Mohid with MPI]].

@@ Line 10: / Line 10: @@
 Parallel processing as it is presently implemented in Mohid, could not be achieved without [[object oriented programming]] philosophy, as each model is an instance of [[class Model]] and no changes, exception made to the implementation of the MPI communications calls needed to be added. Using this feature, computational speed was improved (varying from application to application), as now the whole model will take the same time as the slowest model to run plus the time to communicate with the other processes. Here, the network communication speed plays an important role, as it can become limiting. However, the amount of information passing between models, depending of course on the memory allocated for each model, has not yet proven to be big enough to make a 100 Mbps network connection time limiting.
-[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomains that communicate among them (2-way) via MPI. See the follow report [https://www.dropbox.com/s/bepaeqz9453skve/Mohid_MPI_DomainDecomposition_Doc.pdf MPI_DD_Report] and the follow presentation [https://www.dropbox.com/s/uraxi83rmmuv50b/MOHID%20%20MPI%20%20Domain%20Decomposition_PPT.pdf MPI_DD_PPT] where a description of the MPI/Domain Decomposition functionality is described.
+[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomains that communicate among them (2-way) via MPI. See the follow report [https://www.dropbox.com/s/bepaeqz9453skve/Mohid_MPI_DomainDecomposition_Doc.pdf MPI_DD_Report] where a description of the MPI/Domain Decomposition functionality is described.
 Find here information on how to [[setup a MOHID simulation using MPI]] and on [[compiling Mohid with MPI]].

@@ Line 14: / Line 14: @@
 Find here information on how to [[setup a MOHID simulation using MPI]] and on [[compiling Mohid with MPI]].
-The problem with the current MPI implementation is it's complexity. MPI directives are sread throughout the code. This complexity is very hard to understand. The proposed solution is to consolidate MPI calls in a single point  using the [[Actor Model]] ([http://en.wikipedia.org/wiki/Actor_model]).
+The problem with the current MPI implementation is it's complexity. MPI directives are spread throughout the code. This complexity is very hard to understand from the programming point of view. The proposed solution is to consolidate MPI calls in a single point  using the [[Actor Model]] ([http://en.wikipedia.org/wiki/Actor_model]).
 == Parallel processing via OpenMP ==

@@ Line 10: / Line 10: @@
 Parallel processing as it is presently implemented in Mohid, could not be achieved without [[object oriented programming]] philosophy, as each model is an instance of [[class Model]] and no changes, exception made to the implementation of the MPI communications calls needed to be added. Using this feature, computational speed was improved (varying from application to application), as now the whole model will take the same time as the slowest model to run plus the time to communicate with the other processes. Here, the network communication speed plays an important role, as it can become limiting. However, the amount of information passing between models, depending of course on the memory allocated for each model, has not yet proven to be big enough to make a 100 Mbps network connection time limiting.
-[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomain that communicate among them (2-way) via MPI. See the follow report [https://www.dropbox.com/s/bepaeqz9453skve/Mohid_MPI_DomainDecomposition_Doc.pdf MPI_DD_Report] and the follow presentation [https://www.dropbox.com/s/uraxi83rmmuv50b/MOHID%20%20MPI%20%20Domain%20Decomposition_PPT.pdf MPI_DD_PPT] where a description of the MPI/Domain Decomposition functionality is described.
+[[Domain Decomposition]] is an ongoing project, The present version decouples a domain into several subdomains that communicate among them (2-way) via MPI. See the follow report [https://www.dropbox.com/s/bepaeqz9453skve/Mohid_MPI_DomainDecomposition_Doc.pdf MPI_DD_Report] and the follow presentation [https://www.dropbox.com/s/uraxi83rmmuv50b/MOHID%20%20MPI%20%20Domain%20Decomposition_PPT.pdf MPI_DD_PPT] where a description of the MPI/Domain Decomposition functionality is described.
 Find here information on how to [[setup a MOHID simulation using MPI]] and on [[compiling Mohid with MPI]].

@@ Line 4: / Line 4: @@
 == Parallel processing via MPI ==
-[[Mohid Water]] uses MIP to distribute the workload with 2 different strategies that can be apllyed simultaneously or in separation: [[Nested Models]] and [[Domain Decomposition]]. A tree sutructure (linked list) of Mohid Water instances is created and a map of model dependencies so that each model has a list of models that it needs to communicate with.
+[[Mohid Water]] uses [[MIP]] to distribute the workload with 2 different strategies that can be apllyed simultaneously or in separation: [[Nested Models]] and [[Domain Decomposition]]. A tree sutructure (linked list) of Mohid Water instances is created and a map of model dependencies so that each model has a list of models that it needs to communicate with.
 The [[Mohid Water]] ability to run [[Nested Models]] was accomplished by creating a linked list of all the models and by attributing to each one a father-son identification, through which the models communicate. The first stage for introducing parallel processing in Mohid was to add the possibility of launching a process by each model to run, and then, using MPICH, establish communication between models. This enables each sub-model to run in a different processor (even if the processor belongs to a different computer, as long as it is in the same network) and in parallel, instead of running all in the same processor and each model having to wait for the others to perform their calculations. Currently Nested Models have a one-way communication, border conditions being passed from father to son. This results in a assyncronous comunication where the fathe model can calculate the next time step without wayting its sons. In most applications fathers are much lighter (faster) than their sons...
@@ Line 13: / Line 13: @@
 Find here information on how to [[setup a MOHID simulation using MPI]] and on [[compiling Mohid with MPI]].
 == Parallel processing via OpenMP ==