Dytran Release Guide

Transcription

1 Dytran 2016 Release Guide

2 Corporate Europe, Middle East, Africa MSC.Software Corporation MSC.Software GmbH 4675 MacArthur Court, Suite 900 Am Moosfeld 13 Newport Beach, CA Munich, Germany Telephone: (714) Telephone: (49) Toll Free Number: europe@mscsoftware.com americas.contact@mscsoftware.com Japan Asia-Pacific MSC.Software Japan Ltd. MSC Software (S) Pte. Ltd. Shinjuku First West 8F 100 Beach Road 23-7 Nishi Shinjuku #16-05 Shaw Tower 1-Chome, Shinjuku-Ku Singapore Tokyo , JAPAN Telephone: Telephone: (81) (3) APAC.Contact@mscsoftware.com MSCJ.Market@mscsoftware.com Worldwide Web User Documentation: Copyright 2016 MSC.Software Corporation. Printed in U.S.A. All Rights Reserved. This document, and the software described in it, are furnished under license and may be used or copied only in accordance with the terms of such license. Any reproduction or distribution of this document, in whole or in part, without the prior written authorization of MSC.Software Corporation is strictly prohibited. MSC.Software Corporation reserves the right to make changes in specifications and other information contained in this document without prior notice. The concepts, methods, and examples presented in this document are for illustrative and educational purposes only and are not intended to be exhaustive or to apply to any particular engineering problem or design. THIS DOCUMENT IS PROVIDED ON AN AS-IS BASIS AND ALL EXPRESS AND IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, ARE DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO BE LEGALLY INVALID. MSC.Software logo, MSC, MSC., MD, Adams, Dytran, Marc, Mentat, and Patran are trademarks or registered trademarks of MSC.Software Corporation or its subsidiaries in the United States and/or other countries. NASTRAN is a registered trademark of NASA. LS-DYNA is a trademark of Livermore Software Technology Corporation. All other trademarks are the property of their respective owners. Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set forth in FAR (Commercial Computer Software) and DFARS (Commercial Computer Software and Commercial Computer Software Documentation), as applicable. DT*V2016*Z*Z*Z*DC-REL.PDF

3 Contents Contents 1 Introduction 2 Enhancements Overview 7 Enhance Partitioner 8 Improve Performance for Models with Very High Number of Segments in the Coupling Surface 9 Other Enhancements and Defect Corrections 12 Dytran Explorer Redesign 13 List of Software Corrections in Dytran List of Known Software Defects in Dytran System Information Software Installation 19 Licensing 19 Release Platforms 20 Memory Requirements 21 4 Using Dytran Running Dytran on Linux 23 Running Dytran on Windows 24

4 Chapter 1: Introduction 1 Introduction

5 5 Dytran 2016 is the latest and most comprehensive version of Dytran released by MSC Software providing you with new simulation technology and improved performance. Dytran 2016 is available on Linux X8664 (Red Hat 6.3, SUSE ES 11 sp2) and Windows 64 bit (Windows 7). Please see Chapter 3: System Information of the Dytran 2016 Release Guide for more details. Dytran 2016 includes major new capabilities that are focused in the areas of High Performance Computing (HPC) for fluid-structure interaction (FSI) applications resulting in dramatic performance improvement for CPU intensive simulations. These are: Orthogonal recursive bisection (ORB): More efficient and more user-friendly Euler domain partitioner. More efficient Couple Surface processing for models with extremely large number of elements that are part of the coupling surface with a focus on higher number of cores in the DMP simulation. In addition, several critical software defects have been corrected in this release. The Dytran 2016 DMP technology is not extended to the structural solver. The DMP capability does not require any additional licensing requirements. The Dytran 2016 online documentation is available in PDF format on all platforms. The online documentation includes the Dytran Reference Manual, Dytran Theory Manual, Dytran User s Guide, Dytran Example Problem Manual,, and the Dytran Installation Instructions. Dytran uses the Macrovision FLEXlm licensing system. If you already have a Dytran 2013 license, you will not need to obtain a new authorization code to activate Dytran 2016 on your computer. However, you will need to install the latest FLEXlm 11.9 license server. MSC Licensing software is now provided on its own Product Page on the Solution Download Center (SDC). If you need assistance while installing Dytran 2016, please call the MSC Technical Support Hotline at , or your support questions to mscdytran.support@mscsoftware.com. For the latest information on supported platforms for upcoming releases of MSC products, please visit the following web site:

6 Chapter 2: New Capabilities in HPC and Fluid-structure Interaction (FSI) 2 Enhancements Overview 7 Enhance Partitioner 8 Improve Performance for Models with Very High Number of Segments in the Coupling Surface 9 Dytran Explorer Redesign 13 List of Software Corrections in Dytran

7 7 Overview Overview This version includes two different updates all related to Fluid Structure Interaction (FSI) functionality in Dytran. Improve Partitioning: Add optimal partitioner for block Euler meshes Enhance performance FSI surface computations: to get better performance for models with very large number of segments in the coupling surface

8 Chapter 2: New Capabilities in HPC and Fluid-structure Interaction (FSI) Enhance Partitioner 8 Enhance Partitioner Dytran already offers several schemes to partition the Euler mesh. The user can select a partition scheme by using PARAM, EULERPR. The existing schemes are often not easy to use or do not result in an optimal partition. A new partitioning scheme, the Orthogonal Recursive Bisection algorithm (ORB) has been added. Option ORB was added to PARAM, EULERPR and has been made the default. An example of the benefit of the ORB implementation: A model contains 1 Euler mesh with 800 by 450 by 1 elements. The SIMPLE partitioner algoithm does not find the most optimum partitioning for 4 cores. Also it would fail for 6 and 12 cores. The ORB partitioner does find the most optimal way. In table format: Number of CPUs EULERPR,ORB 2x1x1 2x2x1 3x2x1 4x2x1 6x2x1 8x2x1 8x4x1 EULERPR,SIMPLE 2x1x1 4x1x1 FATAL 4x2x1 FATAL 8x2x1 8x4x1 Explanation: On 4 CPU's, the split 2x2x1 gives better load balancing of communication costs across cores than the 4x1x1 split. For the last split, core 1 and core 4 communicate with, respectively, core 2 and core 3. This requires communicating 2*400 faces. Core 2 has to communicate with core 1 and core 3, and core 3 has to communicate with core 2 and core 4.This requires an additional communication of 4*400 faces. Therefore, the total communication for the 4x1x1 split is 2* *400 = 2400 faces. For the 2x2x1 split, all cores communicate only = 625 faces. This gives a total communication of 4*625 = 2500 faces. For the 2x2x1 splitting, the total cost is 4% higher, but it has a good load balancing of communication costs across cores. For this particular model, the wall time went from 1100 seconds to 991 seconds. The benefit for 8, 16, and 32 cores were insignificant (because the partitioning between SIMPLE and ORB was the same). Yet, a user can now very easily choose a number other than 2, 4, 8, 16, 32, etc for the cluster simulations, because the ORB partitioner allows for this.

9 9 Improve Performance for Models with Very High Number of Segments in the Coupling Surface Improve Performance for Models with Very High Number of Segments in the Coupling Surface In Dytran, the coupling surface is determined by calculating the intersection of the structural mesh with the Euler mesh. For many problems, there is a good reduction in the wall time by increasing the number of processors. This enhancement is to improve the performance for problems which were not behaving well before. This was achieved by eliminating the need to perform the intersections on geometric regions that are not relevant. In addition, the computation of intersections was made more efficient. A option was introduced to PARAM,FASTCOUP: PARAM,FASTCOUP,option1,option2,option3,option4,option5,option6,option7 Option7: Each coupling surface segment has to be intersected with the Euler elements and Euler faces. For large Euler meshes, a substantial amount of checking has to be done do find the intersecting elements and faces. By dividing the Euler mesh in search boxes, the costs of this checking is kept to a minimum. Especially for large Euler meshes combined with large coupling surface, this checking can become expensive, and then the use of search boxes can significantly reduce the costs. For Euler meshes of less than elements, the use of search boxes will have little effect. (Character; Default = ON) OFF ON Search boxes are not used. Search boxes are used. While not all models will behave the same, for a typical tire hydroplaning applications, one can observe that the there is a greater reduction of wall time, and, in fact, the performance in Dytran 2016 with 4 cores is now as good as the behavior in Dytran 2013 with 16 cores. Using the previous methods, it should be noted that this model used over 2 million Euler elements. The improvements are less dramatic for small models.

10 Chapter 2: New Capabilities in HPC and Fluid-structure Interaction (FSI) Improve Performance for Models with Very High Number of Segments in the Coupling Surface 10 The focus of the new method was to reduce the time spent in blending and in the coupling algorithms. For example a model with Euler elements and quad structural elements that are all part of the FSI coupling surface. Here are the run-times for the old and the new algorithm: Number of Cores Dytran 2016 Total run time Coupling Blending Dytran 2013 Total run time Coupling Blending The relative performance between the two versions: Number of Cores Total run time Coupling Blending

11 11 Improve Performance for Models with Very High Number of Segments in the Coupling Surface What can be observed in the above tables is that the performance of the new algorithm is higher with higher number of cores used in the simulation. However, even the performance for serial runs has been improved. But this is only true when the model has a relatively large number of elements in the coupling surface, as is in the model used in the tables above. The performance will not be as significant for models that have a relative small number of elements in the couple surface or when there is no FSI couple surface (Euler only simulations).

12 Chapter 2: New Capabilities in HPC and Fluid-structure Interaction (FSI) Other Enhancements and Defect Corrections 12 Other Enhancements and Defect Corrections New Option for ACTIVE The ACTIVE entry allows for activation or deactivation of the Euler elements for some time. But when the Euler elements are switched on again, the Eulerian masses still originate from the Euler element initialization at cycle 0. When there has been substantial movement of the coupling surface during deactivation, the Eulerian masses can be erroneously compressed. This shows up by large velocities in the Euler elements. To avoid these instabilities the Euler initialization can be postponed by using the new INITV option. PERMCPL The PERMCPL entry has been added to define a permeable area between two coupling surface. PARAM,COSUBMAXT By default, the coupling surface computation is done every cycle. Depending on the velocity of the coupling surface, it is often possible to skip these computations for several cycles. This is called subcycling. This is activated by PARAM,COSUBMAX. For a simulations in which the velocity of the structure changes, it can be useful to make the maximal mumber of subcycles time dependent. PARAM,COSUBMAXT has been added to enable this.

13 13 Dytran Explorer Redesign Dytran Explorer Redesign The Graphical User Interface (GUI) for Dytran on Windows, called Dytran Explorer, has been revamped. The user will find the familiar options with a new look and feel. The functionality of the THS viewer has been greatly improved: Dynamic zooming and scaling options Different foreground and background chart options Curve and point click Currently the new GUI does not support all the options from the old version. User subroutines. ATB input files Execution of Shared Memory Parallel process Job queueing Please see Chapter 4: Running Dytran, Running Dytran on Windows for a work-around for each of the first three unsupported features.

14 Chapter 2: New Capabilities in HPC and Fluid-structure Interaction (FSI) List of Software Corrections in Dytran List of Software Corrections in Dytran 2016 The following software defects are fixed in this release: Defect DYT-298 DYT-391 DYT-417 DYT-421 DYT-434 DYT-465 DYT-471 DYT-474 DYT-477 DYT-516 DYT-517 DYT-526 DYT-540 DYT-541 DYT-547 Description When EXCOMP is used, the user subroutine must be compiled and used by the Dytran executable. Dytran did not detect properly that the user subroutine was not supplied. An error message was added to capture a missing subroutine. Initialization of multiple Euler meshes using the graded mesh functionality can take very long for large models. This initialization has been made more efficient. Importing an Euler archive from a 1-D spherical symmetric simulation into a 2-D-axial symmetric run was not yet possible. This has been enabled. Dytran Explorer was replaced by a new interface. The old interface was using Visual Basic features which was no longer supported on Windows. The new interface can open THS files. Solved an error that caused Arc2VTK not being able to process multiple archive files with differences in the connectivity. VTU files resulting from the changed archive file did not contain results data. When the archives were processed separately, the problem did not occur. WALLS and WALLOUT were not described in the Dytran Reference Manual. A description has been added. Adaptive contact with DMP parallel is not supported. This combination has been enabled. Grades meshing can take long to initialize. This has been made more efficient. For DMP parallel simulations, the displacements of Lagrange grid points can be erroneously zero. An update to the code has been made to always correctly retrieve displacements of Lagrange grid points. Due to the system restriction of 512 files that can be opened simultaneously per process, Arc2VTK could not process THS files with results for more than about 500 grids or elements. This turned out to be a limitation in cases where time history output has to be used as input for other programs like Actran. The code was changed such that it will process THS files with larger numbers of grids or elements in multiple sweeps, so the maximum number of open files will not be exceeded anymore. Restart of simulations with HYDSTAT fail. These types of restarts have been enabled. Restart with Roe solver but without coupling surfaces can give time step too small. This has been solved. Memory error occurred for the SETC card definition when the fixed format is used. This was fixed. The free format works ok without the fix. Displacement XDIS, YDIS, ZDIS are not written to THS file. They have been added. Using MESH, ADAPT with DMP parallel requires that the number of cores is a power of 2. This restriction has been removed.

15 15 List of Software Corrections in Dytran 2016 Defect Description DYT-549 VELMAX can sometimes postpone failure of the structure. The reason is that PARAM,VELMAX also limits the angular velocity of Langrange grid points. To skip the limiting of angular velocities, PARAM,VELMAX has been added. DYT-554 The forces defined by BODYFOR are normalized. Option NORMALIZE has been added to BODYFOR to skip the normalization and keep the original force definition. DYT-555 Activation of Euler elements in combination with moving structures can cause instabilities. This has been solved by adding a new option INITV to the ACTVE entry. DYT-557 The number of Eulerian material was limited to 50. This limit has been removed. DYT-558 Using both RJREV and RJTRA could give errors that they could not be read in correctly. This has been solved. DYT-560 Restart jobs with multiple Euler domains could terminate with error. This has been solved. DYT-561 MRSUM in OUT file by default is repeated too often in DMP parallel runs. The default cycle frequency of the MRSUM was increased to the same cycle frequency as MATSUM. DYT-562 CDAMP2 can give error when there are MATRIGS. This has been solved. DYT-567 The computation of distortional energy for beam elements can in rare cases cause NaN after which the simulation terminates. This has been solved. DYT-580 The combination of MESH, ADAPT with another MESH using submesh was not supported for DMP parallel. An update to code has been made to allow for this combination. DYT-581 Clumps consisting of more than 40 elements can cause erroneous oscillatory results. An error message is given when these large clumps occur. DYT-582 Models that contain PARAM, MATRMRG1 show errors while trying to run DMP parallel. An update to the code was made to allow PARAM, MATRMRG1 with DMP. DYT-590 Models that include RCONN may crash when DMP is activated. This problem has been fixed. DYT-592 Dytran will fail to execute when the path to the working directory is more than 80 characters. The new limit is now increased to 250 characters. DYT-596 Dytran crashes for a model that has MARKER nodes that are used in a CONTACT definition which is not allowed. An error message was added to capture this combination and Dytran will terminate properly. DYT-600 PARAM,COSUBMAXT has been added to allow time dependent coupling surface subcycling. DYT-610 Moving coordinate systems gives errors when using DMP parallel. Support for this combination has been added. DYT-621 Solved crashes of Arc2Txt on Windows that occurred in 2013 releases. DYT-624 Restart simulations using MESH, ADAPT can crash. An update to the code has been made to enable these restarts.

16 Chapter 2: New Capabilities in HPC and Fluid-structure Interaction (FSI) List of Software Corrections in Dytran Defect DYT-640 DYT-673 Description Jobs running on Windows past midnight were showing negative run-times in the.out file. This has been corrected. Subsurface output often shows zero values when using Euler sybcycling (PARAM,EUSUBMAX). This has been solved.

17 17 List of Known Software Defects in Dytran 2016 List of Known Software Defects in Dytran 2016 The following are known software defects in this release: Defect DYT-294 DYT-328 DYT-352 DYT-447 DYT-459 DYT-481 DYT-518 DYT-556 DYT-584 DYT-595 DYT-606 DYT-607 Description Remove limits on number of output requests. Current limit is 100 output requests per simulation. TIC3 doesn't work correctly with shell cone in Dytran. Rotation does not work. Could not output fiber and matrix failure index successfully when selecting Classic Laminate Theory. Beam offset leads to incorrect results. Catastrophic error occurs when BJOINs are used together with contact and PARAM, CONTACT, VERSION, V4 is used. Scheduler and Queuing of Dytran Explorer does not support DMP run. Couldn't call FORTRAN complier successfully within Dytran explorer. CONM2 with RBE2 doesn't give nodal outputs. Too many Euler cubes causes crash running DMP. RBHINGE not working properly. When SET card is incorrectly used the DMP simulation does not terminate with a proper error message Limitation of number of Euler material. When number is too high, the DMP simulation does not terminate with a proper error message.

18 Chapter 3: System Information 3 System Information Software Installation 19 Licensing 19 Release Platforms 20 Memory Requirements 21

19 19 Software Installation Software Installation Dytran 2016 on Windows and Linux platforms can be downloaded from MSC.Software s Solutions Download Center: On the Windows platforms, Dytran 2016 can easily be installed as it uses the standard Windows Installation Wizard. On Linux platforms, the MSC.Software standard installation script can be used to install the software on your system. Dytran 2016 is the successor of Dytran Licensing Dytran uses the FLEXlm license manager as the licensing system for nodelock and network licensing. To run Dytran, you need an authorization code from MSC.Software Corporation. If you already have a license for MSC Dytran 2013, you will not need to obtain a new license for Dytran DMP capability is part of Dytran Standard and no additional licenses are needed to run DMP capability in Dytran However, in all cases, you do need a new installation of the license server software. Specifically, the FlexLM license server needs to be at level 10 or higher. For this purpose, an installation of FlexLM v11.9. is part of this release on all supported platforms. It is noted that Dytran 2016 is not able to check out licenses when the FlexLM server is lower than version 10. On Windows and Linux computers, Dytran requires an Ethernet card on your computer even if your computer is not connected to a network. The FLEXlm licensing mechanism uses the Ethernet card to create the unique system identification encrypted in the license information file.

20 Chapter 3: System Information Release Platforms 20 Release Platforms Dytran 2016 was built and tested on the following hardware with the listed software installed as given in Table 3-1. Table 3-1 Release Platforms Vendor OS Hardware FORTRAN Version C Version Default MPI Linux (64-bit) RHEL 6.3 SuSE 11 sp2 Intel EM64T Intel 15.0 Inter 15.0 Intel MPI 5.0 Windows (64-bit) Windows7 Intel EM64T Intel 15.0 Microsoft VS 2013 C/C++ Miscosoft HPC Pack 2012 R2

21 21 Memory Requirements Memory Requirements In general, the size of the memory required by Dytran depends on the size of the engineering problem you wish to solve. The default memory size is calculated by the Dytran Solver. It makes an estimate based on the number of elements, number of grid points, boundary conditions, output requests, and others. Typically, the memory does not have to be adjusted anymore since the Dytran 2013 release, but it may still occasionally need adjustment. On Linux platforms, the user has the command-line option (size=small/medium/large), or the user can add the MEMORY-SIZE definition in the input file. Dytran traces the usage of memory and prints a summary at the end of the output file of each analysis. The memory size listed in the summary is exact. It reflects the memory required for storing the model in core memory after one integration step. Additional memory required during the analysis is automatically allocated and de-allocated. Under certain conditions, Dytran may stop and issue a message that it cannot allocate the required memory. Since the memory allocation in Dytran is dynamic, the system may require additional memory during an analysis. If the memory is available, it will be allocated and de-allocated when it is no longer needed. When the computer runs out of memory, the Dytran analysis may stop when it needs more memory to continue. The user can solve this problem by closing applications on the computer that are no longer needed, or the user can decrease the size of the core memory that Dytran allocates for the analysis if Dytran decided to use substantially more than the analysis requires. The information on the memory size requirements of the analysis can be found in the memory summary at the end of the analysis. MSC Software Corporation recommends using Dytran on a computer that has at least 4 GB of RAM.

22 Chapter 4: Using Dytran 4 Using Dytran Running Dytran on Linux 23 Running Dytran on Windows 24 User Subroutines 24 Running Dytran Shared Memory Parallel (SMP) 26 Running Dytran with ATB input files 26

23 23 Running Dytran on Linux Running Dytran on Linux On Linux platform, you would use the command line interface like: dytran jid=xxx (to submit a regular Dytran job) dytran jid=xxx bat=no (to submit Dytran in interactive mode) dytran jid=xxx ncpus=2 (to submit Dytran using two cores for Shared Memory Parallel) dytran jid=xxx exe=my_exe.exe (to submit a Dytran job with a customized executable) To submit DMP jobs, you must specify the number of processors as well: dytran jid=xxx dmp=yes ncpus=2 (to submit Dytran using two cores Distributed Memory Parallel) dytran jid=xxx dmp=yes hlist=hostlist.txt bat=no (to submit Dytran in a cluster using a host list) Note: xxx should be replaced by the name of your input deck without the.dat extension.

24 Chapter 4: Using Dytran Running Dytran on Windows 24 Running Dytran on Windows On Windows, analysis can be submitted with Dytran.Explorer, a graphical interface to control Dytran jobs and post process the results files. Double click on the Dytran icon on your desktop after the installing Dytran 2016 to start Dytran.Explorer. Alternatively, you can use the start Menu to locate Dytran.Explorer under the Programs Folder. The Dytran.Explorer provides an on-line help system which includes online documentation. Basic post processing and animation tools are available by right-clicking on the results files displayed in Dytran Job window. To submit DMP jobs, open a DOS command window and type: {Full path to Dytran installation}\dytran\2016\dytranexe\run_dytran.bat jid={job-name} dmp=dytran nproc={number of processors} For instance: 1. Name input model is bunker.dat. 2. Dytran is installed at C:\MSC.Software. 3. Model needs to run on four CPUs in DMP mode. The correct command would be: C:\MSC.Software\Dytran\2016\dytranexe\run_dytran.bat jid=bunker dmp=dytran nproc=4 User Subroutines For user subroutines, the correct version of the Intel Fortran compiler and Microsoft Visual Studio must be installed (See Chapter 3: System Information for detailed version information). Before Version 2016, the user had to compile the individual user subroutine and create new Dytran executable. In this new version, the user will create a user dynamic library which the existing Dytran executable will try to location and access during the simulation. Also new this version is that we allow the user to include either FORTRAN subroutines or C subroutines. The template for the user subroutines are included in the Dytran 2016 installation. The FORTRAN version is: C:\MSC.Software\Dytran\2016\usr-subrtns\dytran_usersub.f The C-subroutine is: C:\MSC.Software\Dytran\2016\usr-subrtns\dytran_usersub.c The user should copy this file to the running directory, and then replace the dummy subroutine with the user-defined subroutine. Note: All unused subroutines should remain inside the template!

25 25 Running Dytran on Windows Compile the FORTRAN subroutine with: ifort.exe /nologo /c /cm /MD /4I8 /fp:precise /traceback /O2 /Qsave /Qzero /X /Og /Ot /IC:\MSC.Software\Dytran\2016\usrsubrtns dytran_usersub.f Or: Compile the C subroutine with: cl.exe /c /nologo /Og /Op /Ot /X /MD /traceback /O2 -Dwintel64 - DCALLED /IC:\MSC.Software\Dytran\2016\hinc /IC:\Program Files (x86)\microsoft Visual Studio 12.0\VC\include dytran_usersub.c Before linking, copy the following utility files from the installation to the local working directory: get_time.obj get_timestep.obj iget_nswrap.obj iget_runtype.obj iget_step.obj iput_nswrap.obj print_out.obj Now create the dynamic library with: ifort.exe /o./dytran_usersub.dll /nologo /LD get_time.obj get_timestep.obj iget_nswrap.obj iget_runtype.obj iget_step.obj iput_nswrap.obj print_out.obj dytran_usersub.obj -link /INCREMENTAL:NO /def:c:\msc.software\dytran\2016\usr-subrtns /dytran_usersub.def dytran_1.lib dytran_2.lib dytran_3.lib dytran_4.lib dytran_5.lib dytran_6.lib dytran_7.lib lapi_imp.lib sfstubs.a dmpstub.a dystubs.admpstub.a /LIBPATH:C:\MSC.Software\Dytran\2016\dytranexe In order to run the serial version of Dytran 2016, first set the following environment variables: set MP_SPIN= set DPL_NCPUS=1 set OMP_NUM_THREADS=1 set DPL_MAXCPUS=1 set DPL_FLEXLM=True Then add to the PATH the local working directory, so the executable can find the dynamic user library: set PATH={full path to local working directory};%path% Then execute Dytran with: C:\MSC.Software\Dytran\2016\dytranexe\dytran.exe jid={input file name},output={root name of all output files} In order to run DytranDMP with user subroutines, use the batch script that is available in the installation and use the new option called userdll. The value of this option is the complete path the location of the dll: C:\MSC.Software\Dytran\2016\dytranexe\run_dytran.bat jid={input file name} dmp=dytran userdll={full path to local working directory} nproc={number of processors}

26 Chapter 4: Using Dytran Running Dytran on Windows 26 Running Dytran Shared Memory Parallel (SMP) In order to run the serial version of Dytran 2016 using SMP, first set the following environment variables: set MP_SPIN= set DPL_FLEXLM=True Then add to the PATH to following directory: set PATH=C:\MSC.Software\Dytran\2016\usr-subrtns;%PATH% Then set the following variables to the desired number of SMP processes desired: set DPL_NCPUS={n} set OMP_NUM_THREADS={n} set DPL_MAXCPUS={n} Note: The value of n must be the same for all three environment variables! Then execute Dytran with: C:\MSC.Software\Dytran\2016\dytranexe\dytran.exe jid={input file name},output={root name of all output files} Running Dytran with ATB input files In order to run the serial version of Dytran 2016 with ATB, first set the following environment variables: set MP_SPIN= set DPL_FLEXLM=True set DPL_NCPUS=1 set OMP_NUM_THREADS=1 set DPL_MAXCPUS=1 Then add to the PATH to following directory: set PATH=C:\MSC.Software\Dytran\2016\usr-subrtns;%PATH% Then execute Dytran with: C:\MSC.Software\Dytran\2016\dytranexe\dytran.exe jid={input file name},output={root name of all output files},atb={atb input file}