Frequently Asked Questions

How to install Version5?

Two types of installation are possible. To install both Dragon5 and Donjon5:

Without hdf5 support:
cd Version5_evxxx/Donjon/ make
With hdf5 support:
cd Version5_evxxx/Donjon/ make hdf5=1

Is Dragon Version5 an official version? What is its status relative to Dragon 3.0x?

Dragon 3.0x releases are intended to be Industrial Standard Toolset (IST) components for the Canadian nuclear industry. They are used for design and safety studies. Consequently, IST components are verified according to the procedures described in CSA-N286.7. Some modules of Dragon Version5, such as NXT:, LIB:, EVO: and SHI: are similar with those found in Dragon 3.0x and are therefore fully verified. Other modules of Dragon Version5, such as FLU: and EDI:, have been completely rewritten and are not verified according to CSA-N286.7.

However, the development of Dragon Version5 is made with modern quality assurance (QA) traceability techniques. This traceability goal is achieved through the application of versioning control (using Subversion) and issue tracking systems to the following project components:

Ansi-C and Fortran sources of the Version5 components
Non-regression tests
Software documentation

Can I use my Dragon 3.0x input files with Dragon Version5?

Yes, with one exception. Version5 has eliminated all side effects. In Dragon 3.0x, it is usual to use a side effect in module UTL: to change the root directory in a LCM object. In Version5, this operation must be accomplished with the equality (:=) module, as shown in the following example.

In Dragon 3.0x:
LINK_LIST LIBRARY TRACK GEOM ; ... UTL: LIBRARY :: STEP UP MACROLIB ; TRACK := BIVACA: GEOM LIBRARY ; UTL: LIBRARY :: STEP DOWN ;

In Dragon Version5:
LINK_LIST MACRO LIBRARY TRACK GEOM ; ... MACRO := LIBRARY :: STEP UP MACROLIB ; TRACK := BIVACA: GEOM MACRO ;

Also, a few modules, such as JPMT:, SAD:, PER: and MCU: are not available in Dragon Version5. The capabilities of modules EXCELL: and MOCC: have been integrated in ASM: and FLU:, respectively.

Can I use my Donjon 3.0x input files with Donjon Version5?

No, although Trivac-related modules have similar arguments. All modules related to L_MAP and L_DEVICE objects are different. Objects L_TABLE no longer exist in Donjon5. L_COMPO and L_MULTICOMPO objects are interpolated in place in Donjon5, without having to first copy them in a L_TABLE object. The programmation of Donjon5 is greatly simplified and cleaned.

Is Dragon Version5 works on Windows?

Yes, in two possible ways:

A native build is possible, provided that Microsoft Visual Studio and Intel Visual Fortran are available. A script named instver5.bat is available to create executables (.exe) from a MS-DOS command window, but is not actively maintained. Follow the following steps:
- Click Start, point to All Programs, Intel(R) Software Development Tools, Intel(R) Visual Fortran Compiler Professional and finally Fortran Build Environment for applications running on IA-32. The Visual Studio Command Prompt automatically sets the correct paths of both C and Fortran compilers, together with their associated libraries.
- In the command window, type
  cd %homepath%\Version5\
  .\script\instver5
A "Unix-type" build is possible, provided that Windows Subsystem for Linux (WSL) is properly installed on your PC. WSL is a native implementation of a Unix shell. After installation of WSL, you obtain a native Unix terminal window where all Version5 components can be executed.

Is Dragon Version5 works on Mac OS?

Yes. Many components of Version5 are developed on a Mac. However, you must install Xcode and the gfortran compiler before installing Version5.

A tutorial is available to explain the installation procedure on a Mac.

Why using Fortran 2003 in Version5 distribution?

We are using Fortran 2003 for performance and serviceability reasons. The style of Fortran programming we use is nevertheless modern as we don't use obsolete constructs such as COMMON blocks for static storage needs. All LCM object addresses are stored in TYPE(C_PTR) variables and memory allocation is performed with ALLOCATE and DEALLOCATE statements.

Do you support the Flang compiler?

The Flang compiler in LLVM is the Fortran compiler of the future. It is supported on Mac OS starting with version 5.0.11. You first install the Flang compiler using Homebrew:

brew install flang

Next, include the following lines in the .profile or .zprofile script:

# Support for flang-new compiler export LIBRARY_PATH=/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/lib export LIBRARY_PATH="$LIBRARY_PATH:/Library/Developer/CommandLineTools/usr/lib/clang/16/lib/darwin" export LLVMTOOLS=/opt/homebrew/Cellar/flang/19.1.6/lib # directory with libfortran.a

The code is compiled using:

make llvm=1

and executed using

make tests llvm=1

What is the status of Version5 distribution?

Version5 is a 64-bit clean distribution of the reactor physics codes developed at Polytechnique Montréal. This distribution is intended to users of recent computer systems experiencing obsolescence problems with previous distributions. Release 5.0.1 is iso-functional with Version4 distribution at release 4.1.0. The Ganlib5 kernel is mostly programmed in ANSI-C language and the computational modules are programmed in Fortran 2003. Consequently, the g77 compiler can no longer be used. The Version5 distribution is based on the architecture presented in report IGE-332 (PDF). Using Version5 distribution, we successfully performed the following exercices:

Construct a complete build using only 64-bit libraries (i.e., without using -m32 flag). This characteristics is important as some Linux distributions are installed without 32-bit C and Fortran libraries.
Use MPI standard libraries
Make Salome couplings using YACS
Use the optimiser (the -O flag) with recent gfortran compilers
Use hdf5 files
Integrate Version5 components into a Python3 program (using PyGan).

Can I use Version5 components for commercial applications? What are the terms of the license?

The Version5 components are released under the GNU Lesser General Public License (LGPL). This means that you can integrate Version5 components into your industrial framework and use it for commercial applications. The only condition is to inform and make available to us any modification or improvement you make in an existing Version5 component.

Enhancements to Version5 can be written as procedure files in CLE-2000 language or as modifications to the Fortran source. CLE-2000 procedures or new Fortran modules written by you are not covered by the LGPL. We encourage you to release your Fortran source enhancements under the LGPL for inclusion in future versions of Version5.

Why not giving full remote access to the Subversion repository?

All components of the Version5 repository are now open source. We consider giving public access to the repository when Version 5.1 will be released.

Which cross-section libraries can I use with Dragon Version5?

You can use existing libraries in the following formats: MATXS, Apolib-1, Apolib-2, Apolib-3, WIMS-D4, WIMS-AECL and Draglib. Most available libraries in these formats are proprietary, but open source libraries exist in WIMS-D4 (see WLUP site) and Draglib formats.

I try to save a file and to recover it at the end of the DRAGON or DONJON run. After the run, it is not there, but the output indicates a successful run. Some code reading (and debugging) shows that the file is correctly closed and kept after the module call, but is deleted afterwards in the CLE-2000 procedure in "kdrcln".

Any file that is candidate to be kept after the DRAGON or DONJON run must be defined with the FILE attribute. For example, a postscript file to be created in the PSP: module must be declared as

SEQ_ASCII Fig_reg :: FILE 'Fig_reg.ps' ;

Moreover, if the main CLE-2000 dataset file is named data.x2m, a data.save script file must be defined to recover the Fig_reg.ps file. An example of such a data.save script file follows:

#!/bin/sh # if [ $# = 0 ] then echo "usage: data.save directory" 1>&2 exit 1 fi echo access data.save MACH=`uname -s` Sysx="`echo $MACH | cut -b -6`" if [ $Sysx = "CYGWIN" ]; then MACH=`uname -o` elif [ $Sysx = "AIX" ]; then MACH=`uname -s` else MACH=`uname -sm | sed 's/[ ]/_/'` fi ls -l mv *.ps $1/"$MACH" echo "data.save completed"

This script file must be set as executable using a Bourne directive:

chmod 755 data.save

Is it possible to detect a memory leak in DRAGON or in DONJON?

The Valgrind utility can be used together on systems that support it. To enable this verification, all components of Version5 distribution must be compiled with the -debug option. Moreover, the main CLE-2000 dataset file must be executed by a command similar to

valgrind --leak-check=yes ./rdragon data.x2m

Your program will run much slower (eg. 20 to 30 times) than normal, and use a lot more memory. Memcheck will issue messages about memory errors and leaks that it detects.

The power is the product of FLUX-INTG and H-FACTOR records. Why the H-FACTOR record is missing from the Macrolib produced by the EDI: module?

You have to set the DEPL line option in LIB: module.

In the .result file, there are two fluxes that I get (region integrated flux and averaged regional flux). Normally I would expect the flux to be in E14 level. The values that I get are in like 2.14E-1. Is the flux normalized there?

The flux output by EDI: are not normalized. The normalization factor is computed by module EVO:, but is not applied. Only the fluxes produced by modules COMPO:, SAP: or CPO:are normalized.

Cross section information is missing in the .result file. As the cross-section files in DRAGON are in BINARY format, is there any software to read those binary files? Also, is it possible to print the cross-sections using Dragon scripts as well?

All LCM objects produced by DRAGON5 (including the DRAGLIB) can be converted (serialized) into ASCII format using the := operator. Moreover, their internal organization is fully described in IGE351 report.

Here is an example of a case with data exportation. I also made an export of the multicompo containing all the merged/condensed data produced by EDI:. This is the standard way of using DRAGON5. Any Dragon LCM object can be exported that way.

Don't forget to declare chmod 755 the .save and .access scripts.

`casl_pin.access`

`casl_pin.x2m`

INTEGER istep := 1 ;
REAL evobeg evoend ;
REAL step2 step3 step4 step5 := 1.0 27.1739 67.9348 135.8696 ;

* structure and Modules

LINKED_LIST
  LIBRARY LIBRARY2 MOSTELAS MOSTELA TRACKS TRACK SYS
  FLUX BURNUP EDITION DATABASE ISOT ;

MODULE
  GEO: SYBILT: LIB: USS: ASM: FLU: EVO: EDI: COMPO: DELETE: UTL: END: ;
  
SEQ_ASCII _DATABASE :: FILE '_DATABASE.txt' ;

PROCEDURE assertS ;
LIBRARY := LIB: ::
  NMIX 3 PT CTRA NONE
  DEPL LIB: DRAGON FIL: DLIB_7R1_361
  MIXS LIB: DRAGON FIL: DLIB_7R1_361
  MIX 1 600.0
  U235 = 'U235'   1.66078E-4 1
  U238 = 'U238'   2.28994E-2 1
  U236 = 'U236'   0.0        1
  Pu239 = 'Pu239' 0.0        1
  O16  = 'O16'   4.61309E-2
  MIX 2 600.0
  Zr91 = 'Zr91'   3.83243E-2
  MIX 3 600.0
  O16    = 'O16' 2.33753E-02
  H1     = 'H1' 4.67505E-02
  B10    = 'B10' 1.00874E-05
  B11    = 'B11' 4.06030E-05
;

*----------------
*  Geometry MOSTELAS : 3 regions annular cell for self-shielding
*           MOSTELA  : 4 regions annular cell for transport
*---------------
MOSTELAS := GEO: :: TUBE 3
  R+ REFL RADIUS 0.0 0.39306 0.45802 0.71206
  MIX 1  2  3 ;
MOSTELA  := GEO:  MOSTELAS ::
  SPLITR 2  1  1 ;

*------------
*  Create the reactor database
*-----------
DATABASE := COMPO: ::
  EDIT 5
  COMM  'Multi-parameter reactor database' ENDC
  PARA  'BURN' IRRA
  PARA  'FLUB' FLUB
  INIT
  ;

*-------------
*  Self-Shielding calculation SYBIL
*  Transport calculation      SYBIL
*  Flux calculation for keff with imposed buckling
*  using B1 homogeneous leakage model
*------------
TRACKS := SYBILT: MOSTELAS ::
  EDIT 1 MAXR 3 ;
LIBRARY2 := USS: LIBRARY TRACKS :: EDIT 0 PASS 2 ;
TRACK := SYBILT: MOSTELA ::
  EDIT 1 MAXR 4 ;
SYS := ASM: LIBRARY2 TRACK ;
FLUX := FLU: SYS LIBRARY2 TRACK ::
  TYPE K B1 PNL BUCK 0.2948E-2 ;
EDITION := EDI: LIBRARY2 TRACK FLUX ::
  EDIT 3 MICR RES MERG COMP COND 0.625 SAVE  ;

*----------------
*  Burnup loop: for first step BURNUP is created
*  while for other steps it is modified
*  two burnup per step:
*  1) get a first approximation of final composition followed
*     by a transport calculation
*  2) use approximation for final flux distribution to get a
*     better approximation for final composition
*---------------

EVALUATE evoend := 0.0 ;
WHILE evoend step2 < DO
  EVALUATE evobeg := evoend ;
  EVALUATE evoend := step2 ;
  IF istep 1 = THEN
    BURNUP LIBRARY2 := EVO: LIBRARY2 FLUX TRACK ::
     SAVE <<evobeg>> DAY POWR 36.8
     DEPL <<evobeg>> <<evoend>> DAY POWR 36.8
     SET <<evoend>> DAY ;
  ELSE
    BURNUP LIBRARY2 := EVO: BURNUP LIBRARY2 FLUX TRACK ::
     SAVE <<evobeg>> DAY POWR 36.8
     NOEX DEPL <<evobeg>> <<evoend>> DAY POWR 36.8
     SET <<evoend>> DAY ;
  ENDIF ;

LIBRARY2 := USS: LIBRARY LIBRARY2 TRACKS :: EDIT 0 PASS 2 ;
  SYS := DELETE: SYS ;
  SYS := ASM: LIBRARY2 TRACK ;
  FLUX := FLU: FLUX SYS LIBRARY2 TRACK ::
    TYPE K B1 PNL BUCK 0.2948E-2 ;

BURNUP LIBRARY2 := EVO: BURNUP LIBRARY2 FLUX TRACK ::
   SAVE <<evoend>> DAY POWR 36.8
   DEPL <<evobeg>> <<evoend>> DAY POWR 36.8
   SET <<evoend>> DAY ;
  LIBRARY2 := USS: LIBRARY LIBRARY2 TRACKS :: EDIT 0 PASS 2 ;
  SYS := DELETE: SYS ;
  SYS := ASM: LIBRARY2 TRACK ;
  FLUX := FLU: FLUX SYS LIBRARY2 TRACK :: TYPE K B1 PNL BUCK 0.2948E-2 ;

EDITION := EDI: EDITION LIBRARY2 TRACK FLUX ::
    EDIT 4 MICR ALL SAVE ;
  BURNUP LIBRARY2 := EVO: BURNUP LIBRARY2 FLUX TRACK ::
   SAVE <<evoend>> DAY POWR 36.8 ;

DATABASE := COMPO: DATABASE EDITION BURNUP ::
   EDIT 3
   SET <<evoend>> DAY ;
  EVALUATE step2 step3 step4 step5 :=
   step3 step4 step5 step2 ;
  EVALUATE istep := istep 1 + ;

ENDWHILE ;

ISOT := DATABASE :: STEP UP default
 STEP UP MIXTURES STEP AT 1
 STEP UP CALCULATIONS STEP AT 2
 STEP UP ISOTOPESLIST STEP AT 1 ;

UTL: ISOT :: DUMP ;

_DATABASE := DATABASE ;

END: ;
QUIT "LIST" .

`casl_pin.save`

What is the status of HDF5 support in Version5?

HDF5 support is an optional capability of the Version5 platform in version 5.0.8 or newer. A HDF5 file is a hierarchical structure made of groups (sub-structures) and datasets (arrays). HDF5 files are popular in competing platforms such as ODYSSEE and APOLLO3. They provide a more accepted alternative to XSM files, already used in Version5.

To activate HDF5 capability in Version5, you need:

Install HDF5 API on your OS and modify your .profile accordingly. If you don't activate HDF5 file support, you don't need to make these modifications. For example, on the Ubuntu OS for x86_64, include the following lines:
# Support for HDF5 export HDF5_INC="/usr/include/hdf5/serial/" # HDF5 include directory export HDF5_API="/usr/lib/x86_64-linux-gnu/hdf5/serial" # HDF5 C API export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$HDF5_API"
On the Ubuntu OS for aarm, include the following lines:
# Support for HDF5 export HDF5_INC="/usr/include/hdf5/serial/" # HDF5 include directory export HDF5_API="/usr/lib/aarch64-linux-gnu/hdf5/serial" # HDF5 C API export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$HDF5_API"
Install Version5 codes using the "make hdf5=1" command.

Why the absorption cross section occurs negative is some fast energy groups?

A negative absorption is a neutronic absorption (ABS) defined as Eq. (3.7) at Sect. 3.2.4 of report IGE335. This value may be negative due to (n,xn) reactions. Another definition is the depletion absorption defined as NTOT0 - (NELAS + NINEL). This value is always positive. The definition of absorption is always ambiguous, in any code. We strongly recommend using ENDF-type reactions such as NTOT0 (total), NG (radiative capture), NELAS (elastic scattering), NINEL (inelastic scattering), etc.

How to install the Python3 API (PyGan) for Version5?

Install the Version5 distribution as before. Next, install the three Python3 extension modules lcm, lifo and cle2000 using PyGan makefiles. Two types of installation are available:

Install the extension modules in the Version5_evxxx/PyGan/lib/ directory:
cd Version5_evxxx/PyGan/ make hdf5=1
Install the extension modules in the /site-packages/ directory of your OS, together with other extension modules such as numpy or matplotlib. This capability is only possible on recent distributions.
cd Version5_evxxx/PyGan/ make hdf5=1 pip=1