~reducedmodelling/fluidity/ROM_Non-intrusive-ann : changes

~reducedmodelling/fluidity/ROM_Non-intrusive-ann » Changes to assemble/Zoltan_global_variables.F90 from revision 4037

From Revision 4037

Rev	Summary	Authors	Date
3590	Cleaning up the way we handle detectors internally and remov... Cleaning up the way we handle detectors internally and removing legacy Bisection method for Lagrangian detectors. Branch queue passes all unit, short and medium tests on buildbot. All long tests affected will be updated accordingly. Summary of changes: =================== * Removed Bisection method, making the Runge-Kutta Guided Search the default algorithm for lagrangian detector advection. * In parallel each detector now should always be held on the processor that owns the detector's element, ie. they should not reside in a halo between timesteps. * A new module detector_parallel now provides the routine exchange_detectors(), which serialises and sends detectors point-to-point during the advection routine. This module also provides distribute_detectors(), which checks that all local detectors actually reside in a local element, and sends them if necessary. * All code paths initialising detectors (single/array/checkpoint) now first check whether we would own the according element, and only then allocate the detector. * Cleaned up the interface for detector_tools module. * Added meta-information, such as the names of fields to output, I/O parameters and parameters for the RK-GS advection, to the detector_linked_list type. * Changed the way Zoltan searches for detectors to send with an element, so that we only need to traverse each local detector once and can handle multiple detector lists. * Added the analytical 2D solution for 100 lagrangian detectors to detectors_parallel_adaptive test. * Added a test to check from_file detector initialisation to binary_detectors test. * Added Forward Euler and RK4 as pre-defined options under lagrangian_timestepping to the schema. A more detailed description of the changes can be found at: https://blueprints.launchpad.net/fluidity/+spec/detector-cleanup	Michael lange	12 years ago
3584	Addition of new option for Zoltan, additional_adapt_iteratio... Addition of new option for Zoltan, additional_adapt_iterations, documented in diamond. Change of Zoltan default final_partitioner to ParMETIS. Addition of a final load balance step with no edge-weighting applied used if tightening load_imbalance_tolerance fails to prevent empty paritions being created. General update of the Zoltan option descriptions.	Paul Woodhams	12 years ago
3580	Adding an flredecomp branch of options to diamond. Options a... Adding an flredecomp branch of options to diamond. Options available are to select the partitioner used (default is ParMETIS) and to enable or disable Zoltan debugging. Also adding checks for empty partitons based on input_procs and target_procs which are passed to Zoltan when flredecomping. When flredecomping we now use Zoltan_LB_Partition as according to the Zoltan doc for Zoltan_LB_Balance, which we were previously using, NUM_GLOBAL_PARTS and NUM_LOCAL_PARTS are invalid but we need them when flredecomping.	Paul Woodhams	12 years ago
3210	Hopefully get detectors with 2D+1 parallel adaptive jobs wor... Hopefully get detectors with 2D+1 parallel adaptive jobs working. Removes the use of tmp_mesh, we now determine element owners from new_positions%mesh instead.	paulw	13 years ago
2773	Moving the last of the callback routines, zoltan_cb_unpack_f... Moving the last of the callback routines, zoltan_cb_unpack_fields, to the zoltan_detectors module. Next step is some clearning up of the used modules in Zoltan_integration.F90	paulw	13 years ago
2772	Moving zoltan_cb_pack_fields to zoltan_callbacks module. Moving zoltan_cb_pack_fields to zoltan_callbacks module.	paulw	13 years ago
2771	Moving zoltan_cb_pack_field_sizes to zoltan_callbacks module... Moving zoltan_cb_pack_field_sizes to zoltan_callbacks module.	paulw	13 years ago
2770	There was one final detector routine that needed to be moved... There was one final detector routine that needed to be moved to zoltan_detectors. Now moved and renamed populate_detector_arrays to prepare_detectors_for_packing in zoltan_detectors. Had to add some variables to zoltan_global_variables, these could be passed in to the subroutine directly but they'll be needed by the field callbacks anyway so would have to go in zoltan_global_variables eventually.	paulw	13 years ago
2765	Adding #ifdef HAVE_ZOLTAN to all the new files to hopefully ... Adding #ifdef HAVE_ZOLTAN to all the new files to hopefully fix the error pointed out in Mantis issue 905.	paulw	13 years ago
2763	Moving zoltan_cb_unpack_nodes to the zoltan_callbacks module... Moving zoltan_cb_unpack_nodes to the zoltan_callbacks module.	paulw	13 years ago
2760	Moving zoltan_cb_pack_nodes to the zoltan_callbacks module. Moving zoltan_cb_pack_nodes to the zoltan_callbacks module.	paulw	13 years ago
2759	Moving zoltan_cb_pack_node_sizes into zoltan_callbacks modul... Moving zoltan_cb_pack_node_sizes into zoltan_callbacks module.	paulw	13 years ago
2758	Moving zoltan_cb_get_edge_list into the zoltan_callback modu... Moving zoltan_cb_get_edge_list into the zoltan_callback module. Also added a missing save statement in zoltan_global_variables.	paulw	13 years ago
2753	Moving zoltan_cb_get_num_edges into the zoltan_callbacks mod... Moving zoltan_cb_get_num_edges into the zoltan_callbacks module.	paulw	13 years ago
2751	Moving a second callback routine, get_owned_nodes, into the Moving a second callback routine, get_owned_nodes, into the Zoltan_callbacks module.	paulw	13 years ago
2750	Forgot the implicit none in global variables. Also going to ... Forgot the implicit none in global variables. Also going to be a bit more careful about scoping modules to try and make it clearer where some types and functions come from.	paulw	13 years ago
2749	Global variables are a necessary evil within the Zoltan code... Global variables are a necessary evil within the Zoltan code as there is no other way for us to make certain information available to the callback functions we must provide to Zoltan. Previously these were module level variables in Zoltan_integration and this forced us to have all the callbacks in there as well. In an effort to split up the code we're going to move the callbacks into a seperate module and have any global variables needed by them in a global variables module used by both the callbacks module and Zoltan_integration. This is the first step moving just a single callback function, owned_node_count, into Zoltan_callbacks and putting the global variables it requires in Zoltan_global_variables. If this doesn't break buildbot I'll start moving the rest of the callbacks into the new module.	paulw	13 years ago