************** NOTE ************************************ I have tried to provide some pointers for creating the binary on most UNIX machines. I have direct access to LINUX, SUN OS, SOLARIS, and HP. I have compiled it on all of these successfully. Please read the notes below. If you are still unable to compile it, then let me know. In the past, it has also been compiled successfully on SGI, DEC Alphas, Mach Unix for Macintosh, Cygwin for Windows NT Note in the latter two, you will have to define drand48(), srand48() as this is not in their libraries Thanks, Bard ************************************************************* **************** NOTE ***************************************** If you install the new version of XPP over existing versions it is best to start from a clean set of files. Type rm *.o first. ******************************************************************* NOTE: New versions have a simplified compilation. Just untar it and type make in the main xpp directory. If errors occur, then you will still have to fix the make files. VERSION 4.3 January 2000 a. Fixed some bugs, added the ability to have up to 400 parameters b. Added a Liapunov exponent calculation c. Added the ability to save the nullclines to a file d. Added colormap features. 6 different colormaps allowed now. e. Changed the delay handle to use a cubic interpolation. f. Added a new operator del_shft to combine the shift and the delay together. g. Extended the maximum length of expressions to 1000 characters h. Changed the command line from -noicon to -iconify. The default is to not iconify the windows. I also switched the meaning of XORFIX so that the default works on LINUX and WINDOWS. Sorry if this screws everyone up i. Added ability to make portrait postscript files j. Added two new integrators - Dormand-Prince 5 and 8(3). k. Fixed a bug in the STIFF/GEAR integrators that arose when the number of equations exceeded 300 l. Rewrote "Kinescope" so that it works much faster and doesnt crash the X server VERSION 4.01 Minor fixes here and there. Added a little file lister etc so that if you just type XPP from the command line you can navigate through files before loading one up. I have (alas) eliminated the "(c)reate" option since it uses the old impenetrable syntax. If any users really miss it, I will put it back in reluctantly. VERSION 4.00 Fixed a bug in the production of ppm files to make mpeg movies. Up to 600 equations can be integrated. Initial conditions ,parameter, etc windows have scrolling so that they are a more manageable size. A file browser has been added that lets you select files easily. ***NEW SET FILES** I have drastically altered the way to save simulations so that there is much more saves and it is all commented so that a human can read it. The new SET file command is compatible with the old format so that it can read the old format set files. However, the new set files are not compatible with the old. VERSION 3.99 (updated) I have added the ability to save into a file just what is plotted on the screen for XMGR compatibility, e.g. VERSION 3.99 Added "reverse integration " command added PHI=#,THETA=# to @ commands VERSION 3.98 Oct 1998 Bug fixes here and there and added the ability to continue periodic points of maps. VERSION 3.95 Aug 1998 Some bug fixes. Added the ability to set delayed initial data in the ODE file. VERSION 3.91 June 1998 1. Added differential-algebraic capabilities to XPP and WinPP VERSION 3.90 June 3 1998 1. Minor fixes that make it more compatible with some compilers. 2. Can now integrate up to 400 ODEs (was 300 before) 3. Fixed some problems with global flags 4. Added "derived parameters" which allow you to define parameters in terms of other parameters and which are updated only when the parameters are changed. VERSION 3.85 Feb 2 1998 1. Include and update the old version of the parser since some compilers complain about the funky pointer arithmetic that I used. VERSION 3.8 Dec 23 1997 1. Fixed some stuff 2. Added AUTO stuff for the @ commands 3. Added some cool networky stuff for doing discrete convolutions and operations with sparsely connected neural networks. 4. Doubled the speed of the calculation engine by rewriting the parser. 5. Ported to Windows 95 see xpp4w95.zip VERSION 3.6 Oct 24 1997 1. Added new command line option to prevent dead iconified windows in fvwm and some other window managers. Call xppaut -noicon 2. Added interspike interval calculation to Poincare maps VERSION 3.5 Sept 1 1997 1. Frozen nullclines 2. Fixed more bugs in table stuff and array printing 3. Fixed up the shift operator to work on parameters 4. Changed data format in browser to give more decimal places VERSION 3.0 Mar 3 1997 1. Fixed more bugs 2. Added new version for initial conditions with arrays 3 !!! ANIMATION PACKAGE FOR MAKING CARTOONS WITH SIMULATIONS !!! Make this from scratch VERSION 2.71 Nov 18 1996 1. Changed flow field to integrate both forward and backward in time 2. Added m(I)ce option to initial data so that you can click on many initial conditions in a row 3. Added fancy space-time plot to go along with array capabilities 4. Fixed miscellaneous bugs :) SOON -- version with LOCBIF another bifurcation package will be available SUMMER 1997 Windows 95 version maybe VERSION 2.6 Oct 3 1996 NOTE: !! Please make this version from scratch ( rm *.o ) !! 1. Main new feature is the addition of the industrial strength stiff solver CVODE, a C language version of LSODE. This is a big piece of code and I only implement a small portion of it. It is optional for now but if you don't want it you must change the Makefile. See below. 2. Added AUTOFREEZE option. Every curve that is integrate is automatically saved. This way you dont have to freeze after every integration. Up to 26 curves can be frozen. 3. Fixed array options for markov processes 4. Fixed arrow head bug 5. Added some new options for @ directives VERSION 2.52 July 29 1996 New Features: 1. Parameter sliders -- move sliders to change parameters 2. Edit the initial conditions, parameter, BC,etc windows directly 3. Tips on all the menus 4. "Array" handling. The following is valid: x[1..10]'= -x[j] 5. Silent mode -- supports computation without X windows and produces a space delimited ascii file for output 6. Transposition of data arrays. 7. Click and move to get x-y values in any window VERSION 2.00 of XPP/XPPAUT New features: 1. Added two new integrators: i) a quality step Runge-Kutta algorithm that is good for nonstiff problems ii) An alternate stiff integrator based on a Rosenbrock method 2. AUTO interface supports discrete problems VERSION 1.80 of XPP/XPPAUT New Features: 1. -convert command line argument converts old-style to new style 2. Symbol font and Times fonts in 5 sizes. All Text is editable 3. Add arrows and pointers to plots 4. Import bifurcation diagrams into main window 5. Include information about the graphics, numerics, storage, and X fonts in the ODE files 6. Make 3D movies and movies of trajectories as a parameter varies 7. Stability for delay equations Installation of XPP and XPPAUT Create a directory and change to it. (I suggest the name xpp) FTP the program (xppaut2.60.tar.Z) to the directory in which you want to install it. I have included as part of this distribution a modified version of the f2c libraries but they do not seem to work on every machine. If you find that they do not work then you will need to ftp f2clib.tar.Z also in the same directory. I have also included sources for the CVODE library. The program is a compressed tar file and must be uncompressed with the command: uncompresss xppaut2.60.tar.Z Then you should untar the file with the command; tar xf xppaut2.60.tar A. You must create the CVODE library. (This is optional -- see below) 1) Change to the directory cvodesrc (type: cd cvodesrc) 2) Type make to create the library. If all goes well then you should have a file called libcvode.a If something goes wrong, then I guess you won't have CVODE. Oh well. 3) Type mv libcvode.a ../. 4) Type cd .. B. You must create the f2c library Your system may already have f2c libraries on it. On LINUX, the library is called libf2c.a and on my SPARC there are two of them, libI77.a and libF77.a. I prefer to use the reduced libs anyway. Case 1. Compiling the reduced version of the f2c libraries. 1) Change to the directory called libI77. 2) Type make 3) If this goes well type the following mv libf2cm.a ../. 4) Type cd .. ** DON'T DO THE NEXT STEP UNTIL YOU ARE SURE THE DISTRIBUTED VERSION DOESNT WORK ** If you find that you need the f2c libraries (although my reduced set should work) then type uncompress f2clib.tar.Z tar xf f2clib.tar ********************************************************************* Case 2. If for some reason the reduced f2c libraries did not work then do the following. Be sure you have uncompressed the f2clib.tar.Z file. 1) cd libI77 2) Type make. 3) If all goes well type mv libI77.a ../. 4) Type cd ../libF77 5) Type make. 6) If all goes well type mv libF77.a ../. 7) Type cd .. Case 3. Your system already has the libraries. Then cd to the XPP directory. Case 4. You cannot compile the f2c libraries. You should try to get the libraries for an original source -- probably `netlib.att.com:/netlib/f2c' There you will find installation notes.) C. Create the xppaut binary You should be in the main xpp directory. 1) Edit the Makefile to suit your machine (see below) NOTE: If for some reason you decided not to use the CVODE, then delete the -DCVODE_YES directive in your CFLAGS line. Also, delete any references to libcvode in the linking stage. 2) Type make There will be lots of warnings! But you can generally ignore them. If this works then you are done. You should have a binary called xppaut Once you have the binary, you should test it on a simple problem. Type ./xppaut lin.ode and it should fire up. If it does not then it is probably because the X Server is not responding. Make sure your X environment is set up correctly. Click on Window and select Zoom In. Then click the mouse inside the main window and holding down the button, drag it on the screen. You should see a rectangle that stretches as you move the mouse. If you do not see this then you should run the program with the -xorfix command line: ./xppaut -xorfix lin.ode Repeat the above and you should see the box. I generally rename xppaut to xpp and if I need the -xorfix flag, I write a small script file called xpp that has one line: xppaut -xorfix $1 $2 This means I dont have to add -xorfix every time. Once you are satisfied that it works, move it to a directory in your path. (/usr/local/bin is standard). There is a document xpp_doc.ps that gives the full documentation for the program and a WWW tutorial. Notes on the Makefile. Copy the distributed Makefile to Makefile.old. Then edit the Makefile according to the comments within it. The main reason that compilation fails is that the required X libraries cannot be found. You should thus alter the Makefile to tell the compiler where to find the libraries. Add -I<X include path> to the CFLAGS line and add -L<X library path> to the LFLAGS line. This should do the trick. Ask your sysop for the correct pathnames. Notes on the RS6000: (thanks to Artie Sherman) 1) For those without gcc, they can run the standard IBM C-compiler, xlc, with a set of definitions that fake a Sun compiler. I just added the following stanza to the file /etc/xlc.cfg: bsdcc: use = DEFLT crt = /lib/crt0.o mcrt = /lib/mcrt0.o gcrt = /lib/gcrt0.o libraries = -lbsd, -lc proflibs = -L/lib/profiled,-L/usr/lib/profiled options = -H512,-T512, -qlanglvl=extended, -qnoro,-D_BSD, -D_NON STD_TYPES, -D_NO_PROTO,-bnodelcsect, -U__STR__, -U__MATH__ (The options must be on a single, wrapped line) and did: "ln /bin/cc /bin/xlc" 2) In the makefile, I used CC=bsdcc *** RECENT SIMPLIFICATIONS FOR RS6000 USERS: *** Artie Sherman says the following seems to work! CC= cc CFLAGS= -O -DAUTO -DCVODE_YES LIBS= -lX11 -lm $(CC) -DAUTO -o xppaut $(OBJECTS) $(AUTOOBJ) $(LDFLAGS) \ libcvode. libf2cm.a $(LIBS) Notes on SUN I have been told that it may be necessaty to add the line -lsocket to the compile line to make version 3.85 on the SUN under gcc. Notes on the new version of LINUX (thanks to Bum Rak Choi) I was able to compile xppaut program on linux but I had to link libraries with -static option as follows. gcc -static -lm -lieee $(OBJ_FILES) .... When I compile xppaut with gcc -m486 option and -lm to optimize calculation, I got the floating exception error message. So, I tried to link with -lieee option. Notes on the DEC alpha What a pain in the ass this machine was. But, thanks to the DEC I am now ready for any other 64 bit machines. I compiled and ran the test program for XPPAUT on the DEC alpha using the DEC compiler cc. I have not succeeded in getting it to work under gcc. You should definitely use the included libf2cm.a for this. I tried a precompiled version of libf2c.a and it crashed. This is because f2c uses lots of "long" declarations which on the DEC are 64 bits while I treated them as 32 bits (the old standard). My version of the f2c stuff in the subdirectory libI77 works. MORE NOTES ON DEC/OSF!! The cc compiler that comes with DEC has problems with my new parser since I use some (possibly illegal) pointer arithmetic which doesn't seem to bother GCC (or WATCOM for Win95!) Thus I have included the old version of the parser in the distribution. To use this, just change a few lines in the Makefile. In the section called SOURCES change parser.c to parserslow.c and in the section called OBJECTS change parser.o to parserslow.o Then remake and hopefully it will work