nrhelp.txt

Numerical Recipes Software 提供的算法子程序集

  > NRCOPY C E xsvdcmp            [unpack the ANSI C example file]
  > NRCOPY C R svdcmp pythag      [unpack the necessary Recipes]
  > NRCOPY C O matrx3             [unpack the data file]
  > BCC -DANSI -K -mh xsvdcmp.c svdcmp.c pythag.c nrutil.obj
                                  [invoke the compiler]
  > xsvdcmp                       [run the demonstration]

  The Example programs are all self-prompting.  Just follow their
  instructions.  Press F7 for an explanation of the compiler
  switches used in the above examples, and for hints on simplifying the
  command lines.

                          Ordering Information 

     Numerical Recipes books, and diskettes for IBM/PC-compatible and
     Apple Macintosh computers, are published exclusively by Cambridge
     University Press.  Their address and telephone order number (North
     American branch) is:

         Cambridge University Press, Order Department
         110 Midland Avenue
         Port Chester, NY 10573

         telephone: toll-free 800-872-7423 (in NY: 800-227-0247)
         from abroad: 914-937-9600

     Outside of North America, orders may be placed with:

         Cambridge University Press, Customer Services Department
         Edinburgh Building
         Shaftesbury Road
         Cambridge, CB2 2RU
         U.K.                                         more...

         fax: (U.K. country code+) 223 315052
         telex: 851 817256 CUPCAM G
         
     The following books are available (prices subject to change):

     Numerical Recipes in C, 2nd Edition (hardcover) $49.95
     Numerical Recipes C Example Book, 2nd Ed. (paper) $29.95
     
     Numerical Recipes in FORTRAN, 2nd Edition (hardcover) $49.95
     Numerical Recipes FORTRAN Example Book, 2nd Ed. (paper) $29.95
     
     These First Edition books remain available:

     Numerical Recipes in Pascal, Revised Edition (hardcover) $49.95
     Numerical Recipes Example Book (Pascal), Revised Ed. (paper) $29.95

     Numerical Recipes Routines and Examples in BASIC (paper) $32.50

     Diskettes are available in C, FORTRAN, Pascal, and BASIC (Microsoft
     QuickBASIC), for IBM/PC-compatibles and Apple Macintosh.  Contact
     Cambridge University Press for pricing information.  A TrueBasic
     diskette is available from TrueBasic Inc., 12 Commerce Ave., West
     Lebanon, NH 03784.


                              Compiler Hints 

       C compilers for IBM/PC-compatible machines are quite varied.
  Since these machines have segmented, and limited, addressable memory,
  different compilers employ widely differing techniques for accommodating
  large programs and/or large data spaces.  The most common sources
  of difficulty in using the Numerical Recipes programs have to do with
  stack or segment sizes, and with choice of memory models.

       The most commonly used C compilers on MS-DOS machines are
  Microsoft C/C++, Borland C++ (which includes C) and Borland Turbo C.
  All of the Numerical Recipes routines have been validated under these
  compilers.  While most routines compile correctly with the default
  compiler options, these defaults are not always ideal for numerical or
  scientific work.  Here are some compiler options that you may wish
  to consider using (singly or more than one):

  Microsoft C:
     /AH         (use huge memory model)
     /F 8000     (set stack size to 32 Kbytes)
     /FPi87      (produce fast code requiring a math coprocessor)
     /J          (change char default to unsigned)
     /Od         (turn off all optimization; sometimes "cures" rare bugs)
     /Zd         (include source code line numbers for debugging)

     You can cause frequently used options to be included automatically
     by using the CL environment variable.  For example, the DOS command
       > SET CL=/AH /F 8000 /J
     will cause the named options to be included by default.  (You can put
     the command in your AUTOEXEC.BAT if you want to use it all the time.)

  Borland or Turbo C:
     -mh         (use huge memory model)
     -f87        (produce fast code requiring a math coprocessor)
     -K          (change char default to unsigned)
     -N          (check for stack overflow; useful when debugging)
     -O-         (turn off all optimization; sometimes "cures" rare bugs)

     You can cause frequently used options to be included automatically
     by placing them, one per line, in a configuration file in your
     working directory.  The default configuration file is named TURBOC.CFG
     To cause an alternative configuration file, e.g., "NRBCC.CFG", to
     be loaded, add "+NRBCC.CFG" to the BCC command line.  Be sure
     that any configuration file includes the location of the necessary
     include and library directories, e.g., contains lines like
        -IC:\BORLANDC\INCLUDE
        -LC:\BORLANDC\LIB

  Please consult your compiler documentation for additional, or more recent,
  information.

                    Troubleshooting and Common Questions 

 Q. I have tried it several times, but the INSTALL program just does not
    work on my machine.
 A. Write immediately to Numerical Recipes Software, P.O. Box 243,
    Cambridge, MA 02238 (USA) or fax to 617-863-1739.  Include your
    telephone or fax number and tell us when we can reach you.  We really
    want to know about this!

 Q. INSTALL seems to work, but I get a message about "can't find a valid
    license key and/or licensee name".
 A. Be sure that the directory that contains the archive files NR*.AR
    is in your DOS path.  Check that the file NR20KEY.AR is there.
    Examine this file with any text editor.  It should be one line long
    and consist of exactly 3 printable characters followed immediately
    by the licensee name.  If this is not the case, re-run the INSTALL
    program and be sure to enter a valid licensee name.  If it IS the
    case, then it is possible that your NR20KEY.AR file has been
    improperly altered.  Please write to us.

 Q. Using NRCOPY, I get a message about "can't find archive file in path".
 A. Either you did not run the INSTALL program, or else you have forgotten
    to modify your AUTOEXEC.BAT file to add the subdirectory with the
    Numerical Recipes archive files to your DOS path.

 Q. Using a Borland compiler, I keep getting the message "Floating Point
    Formats Not Linked".
 A. Believe it or not, Borland considers this a feature, not a bug!  Their
    Technical Support document no. 645 explains that you can force the
    formats to be linked by compiling the following function along with
    your program.  You do not need to call it, just include its filename on
    your compile command line:
           void LinkFloat (void)
           {
             float a=0, *b=&a;
             a=*b;
           }
    We have included this function in the OTHER archive as a file named
    BCCBUG.C.  You can unpack it with NRCOPY.

 Q. I have a nonstandard computer that is unable to display the output
    of the NRHELP command.  What do I do?
 A. A text version of NRHELP, NRHELP.TXT, is included on the diskette,
    though it is not automatically copied to your hard disk by INSTALL. 

 Q. There are some garbage characters in the copyright notice that is
    appended to unpacked files.
 A. Those are not garbage characters.  They contain information about
    the program version and your license.  Please include them if you need
    to write to us about a routine.

 Q. I am uncommonly vigilant against computer viruses, and I refuse to run
    your INSTALL or NCOPY programs on my machine.
 A. You will need to obtain the C source code for Haruhiko Okumura's
    public domain "AR" archiver from a bulletin board.  Build its
    executable on your machine.  The Numerical Recipes archive files
    are in the standard AR format, and can be dearchived with that
    program.  (The AR format is different from other, similarly named,
    formats such as ARC or ARJ.  Incidentally, Haruhiko Okumura is the
    Japanese translator for the Numerical Recipes books published in Japan.)

 Q. What happened to the old Numerical Recipes Example Diskettes?
 A. Press Home and see the topic "How to use the Example programs."

 Q. I have both the FORTRAN and C diskettes installed, but NRHELP seems
    to relate to only one of them.
 A. One diskette's NRHELP.COM has overwritten the other.  Copy them from
    the original installation diskettes to a directory in your path, but
    give them different names, e.g., NRHELPC.COM and NRHELPF.COM.  Then,
    you can use the commands NRHELPC and NRHELPF to get language-specific
    help.

 Q. What happened to Numerical Recipes in Pascal? BASIC?
 A. The Second Edition of Numerical Recipes is available in C and FORTRAN.
    Due to lack of demand, we have not prepared a Second Edition version
    of the Pascal version.  However, the First Edition of Numerical Recipes
    in Pascal continues to be available.  Ditto, BASIC.

 Q. How do I contact Numerical Recipes Software?
 A. Press F2 and then F9 for our address, fax number,
    and email address.
              Numerical Recipes in the REC archive (all .C) 
              (main book section numbers are in parentheses)  

ADDINT  interpolate and add, used by {MGLIN} (19.6)
AIRY    Airy functions (6.7)
AMEBSA  simulated annealing in continuous spaces (10.9)
AMOEBA  minimize in N-dimensions by downhill simplex method (10.4)
AMOTRY  evaluate a trial point, used by {AMOEBA} (10.4)
AMOTSA  evaluate a trial point, used by {AMEBSA} (10.9)
ANNEAL  traveling salesman problem by simulated annealing (10.9)
ANORM2  utility used by {MGFAS} (19.6)
ARCMAK  construct an arithmetic code (20.5)
ARCODE  encode or decode a character using arithmetic coding (20.5)
ARCSUM  add integer to byte string, used by {ARCODE} (20.5)
ASOLVE  used by {LINBCG} for preconditioner (2.7)
ATIMES  used by {LINBCG} for sparse multiplication (2.7)
AVEVAR  calculate mean and variance of a data set (14.2)
BADLUK  Friday the 13th when the moon is full (1.1)
BALANC  balance a nonsymmetric matrix (11.5)
BANBKS  band diagonal systems, backsubstitution (2.4)
BANDEC  band diagonal systems, decomposition (2.4)
BANMUL  multiply vector by band diagonal matrix (2.4)
BCUCOF  construct two-dimensional bicubic (3.6)
BCUINT  two-dimensional bicubic interpolation (3.6)
BESCHB  Chebyshev expansion used by {BESSJY} (6.7)
BESSI   modified Bessel function I of integer order (6.6)
BESSI0  modified Bessel function I_0 (6.6)
BESSI1  modified Bessel function I_1 (6.6)
BESSIK  modified Bessel functions of fractional order (6.7)
BESSJ   Bessel function J of general integer order (6.5)
BESSJ0  Bessel function J_0 (6.5)
BESSJ1  Bessel function J_1 (6.5)
BESSJY  Bessel functions of fractional order (6.7)
BESSK   modified Bessel function K of integer order (6.6)
BESSK0  modified Bessel function K_0 (6.6)
BESSK1  modified Bessel function K_1 (6.6)
BESSY   Bessel function Y of general integer order (6.5)
BESSY0  Bessel function Y_0 (6.5)
BESSY1  Bessel function Y_1 (6.5)
BETA    beta function (6.1)
BETACF  continued fraction used by {BETAI} (6.4)
BETAI   incomplete beta function (6.4)
BICO    binomial coefficients function (6.1)
BKSUB   backsubstitution, used by {SOLVDE} (17.3)
BNLDEV  binomial distributed random deviates (7.3)
BRENT   find minimum of a function by Brent's method (10.2)
BROYDN  secant method for systems of equations (9.7)
BSSTEP  integrate ODEs, Bulirsch-Stoer step (16.4)
CALDAT  calendar date from Julian day number (1.1)
CHDER   derivative of a function already Chebyshev fitted (5.9)
CHEBEV  Chebyshev polynomial evaluation (5.8)
CHEBFT  fit a Chebyshev polynomial to a function (5.8)
CHEBPC  polynomial coefficients from a Chebyshev fit (5.10)
CHINT   integrate a function already Chebyshev fitted (5.9)
CHIXY   used by {FITEXY} to calculate a chi^2 (15.3)
CHOLDC  Cholesky decomposition (2.9)
CHOLSL  Cholesky backsubstitution (2.9)
CHSONE  chi-square test for difference between data and model (14.3)
CHSTWO  chi-square test for difference between two data sets (14.3)
CISI    cosine and sine integrals Ci and Si (6.9)