Gpc

gpc@gnu.de

5400 discussions

Re: random number generator
by Frank Heckenbach 18 Nov '98

18 Nov '98

Robert P. Ewing wrote: > I've seen some discussion about random number generators here > recently. I came across a quite good one (this being something > I use extensively in my work), and thought I'd share it. It is > written up in ACM Transactions on Modeling and Computer Simulation > 8(1) 3-30, 1990. Supposedly it has a period of -1 + 2^19937. > The original was in C; this translation returns the same values > as the original. It is called the Mersenne Twister. Fine. If we want to use it in GPC, we have to be sure about copyright and patent issues. Is this algorithm free of patents? > This library is free software; you can redistribute it and/or > modify it under the terms of the GNU Library General Public > License [...] Was this the license of the original C code, too? > LI = __longlong__ integer; > UL = __unsigned__ LI; {define 64-bit unsigned integer} BTW, UL can be defined in a more readable way as `LongCard' now -- or `Cardinal (64)' if you want it to be exactly 64 bits wide, regardless of the word size of the target platform. Frank -- Frank Heckenbach, frank(a)fjf.gnu.de http://fjf.gnu.de/ PGP and GPG keys: http://fjf.gnu.de/plan

1 0

Curious bug with gpc 19980830
by sam.littlewood＠quantel.com 16 Nov '98

16 Nov '98

Has anyone a fix for this rather curious bug? - it seems subranges, packed records and VAR arguments conspire to make the compiler throw a bogus error!? Thanks for any help. SamL -------8<--------- program foo; { Using a subrange type in a packed recrod seems to break passing the type by reference } type t1 = 0..100; {Has to be subrange} r1 = packed record { Has to be packed } f1 : t1; end; procedure proc1(var a1: t1); { Has to be var } external; procedure proc2(var a1: t1); { Also has to be var } begin proc1(a1); end; begin end. -------8<--------- Reading specs from /usr/local/lib/gcc-lib/i586-pc-linux-gnu/2.8.1/specs gpc version 19980830, based on gcc-2.8.1 /usr/local/lib/gcc-lib/i586-pc-linux-gnu/2.8.1/gpc-cpp -lang-pascal -v -nocharescape -undef -D__GNUC__=2 -D__GNUC_MINOR__=8 -D__GPC__=2 -D__GPC_MINOR__=0 -D__GPC_RELEASE__=19980830 -D__ELF__ -Dunix -Dlinux -D__ELF__ -D__unix__ -D__linux__ -D__unix -D__linux -Asystem(posix) -D__BITS_LITTLE_ENDIAN__=1 -D__BYTES_LITTLE_ENDIAN__=1 -D__WORDS_LITTLE_ENDIAN__=1 -Di386 -Di586 -Asystem(unix) -Acpu(i386) -Amachine(i386) -D__i386 -D__i586 -Asystem(unix) -Acpu(i386) -Amachine(i386) foo.pas /tmp/cca16201.i GNU CPP version 2.8.1 (i386 GNU/Linux with ELF) #include "..." search starts here: #include <...> search starts here: /usr/local/include /usr/local/i586-pc-linux-gnu/include /usr/local/lib/gcc-lib/i586-pc-linux-gnu/2.8.1/include /usr/include End of search list. /usr/local/lib/gcc-lib/i586-pc-linux-gnu/2.8.1/gpc1 /tmp/cca16201.i -quiet -dumpbase foo.pas -version -famtmpfile=foo.gpc -o /tmp/cca16201.s GNU Pascal version 2.8.1 (i586-pc-linux-gnu) compiled by GNU C version 2.7.2.3. foo.pas: In function `Proc2': foo.pas:19: type mismatch in argument 1 of `Proc1' foo.pas:19: passing arg 1 of `Proc1' from incompatible pointer type

2 1

random number generator
by Robert P. Ewing 15 Nov '98

15 Nov '98

Hi all. I've seen some discussion about random number generators here recently. I came across a quite good one (this being something I use extensively in my work), and thought I'd share it. It is written up in ACM Transactions on Modeling and Computer Simulation 8(1) 3-30, 1990. Supposedly it has a period of -1 + 2^19937. The original was in C; this translation returns the same values as the original. It is called the Mersenne Twister. enjoy. Toby (* A Pascal program for MT19937: Real number version([0,1)-interval). Written in C 4/6/1998 by Takuji Nishimura. Translated to GNU Pascal 11/2/1998 by Toby Ewing. genrand() generates one pseudorandom real number (double) which is uniformly distributed in [0,1)-interval, for each call. sgenrand(seed) sets initial values to the working area of 624 words. Before genrand(), sgenrand(seed) must be called once. (The seed is any 32-bit integer except for 0). Coded by Takuji Nishimura, considering the suggestions by Topher Cooper and Marc Rieffel in July-Aug. 1997. This library is free software; you can redistribute it and/or modify it under the terms of the GNU Library General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library General Public License for more details. You should have received a copy of the GNU Library General Public License along with this library; if not, write to the Free Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA Copyright (C) 1997 Makoto Matsumoto and Takuji Nishimura. When you use this, send an email to: matumoto(a)math.keio.ac.jp with an appropriate reference to your work. *) Program MersenneTwister; Const N = 624; {Period parameters} M = 397; type LI = __longlong__ integer; UL = __unsigned__ LI; {define 64-bit unsigned integer} mtary = array[0..N] of UL; magary = array[0..1] of UL; var j : integer; mti : integer value N+1; {mti = 1 means mt[N] is not initialized} mt : mtary; {the array for the state vector} Procedure sgenrand(seed : UL); { Set initial seeds to mt[N] using the generator Line 25 of Table 1 in [KNUTH 1981, The Art of Computer Programming Vol. 2 (2nd Ed.), pp 102]. } begin mt[0] := seed AND 16#ffffffff; for mti := 1 to N do mt[mti] := (69069 * mt[mti-1]) AND 16#ffffffff; end; Function genrand : real; var y : UL; kk : __short__ integer; Matrix_A : UL value 16#9908b0df; {constant vector a} Up_Mask : UL value 16#80000000; {most significant w-r bits} Low_Mask : UL value 16#7fffffff; {least significant r bits} T_Mask_B : UL value 16#9d2c5680; {Tempering parameters} T_Mask_C : UL value 16#efc60000; mag01 : magary value (0, Matrix_A); begin if (mti >= N) then begin {generate N words at one time} if (mti = N+1) then sgenrand(4357);{If sgenrand hasn't been called, use default seed.} for kk := 0 to N-M do begin y := (mt[kk] AND Up_Mask) OR (mt[kk+1] AND Low_Mask); mt[kk] := mt[kk+M] XOR (y shr 1) XOR mag01[y AND 16#00000001]; end; for kk := N-M to N-1 do begin y := (mt[kk] AND Up_Mask) OR (mt[kk+1] AND Low_Mask); mt[kk] := mt[kk+(M-N)] XOR (y shr 1) XOR mag01[y AND 16#00000001]; end; y := (mt[N-1] AND Up_Mask) OR (mt[0] AND Low_Mask); mt[N-1] := mt[M-1] XOR (y shr 1) XOR mag01[y AND 16#00000001]; mti := 0; end; y := mt[mti]; inc(mti); y := y XOR (y shr 11); y := y XOR ((y shl 7) AND T_Mask_B); y := y XOR ((y shl 15) AND T_Mask_C); y := y XOR (y shr 18); genrand := real(y) * 2.3283064365386963e-10; end; begin {Output first 1000 generated numbers} sgenrand(4357); {any integer > 0 works as a seed} for j := 1 to 1000 do begin write(genrand:10:8); if ((j mod 8) = 0) then writeln else write(' '); end; end.

1 0

Re: Curious bug with gpc 19980830
by Frank Heckenbach 13 Nov '98

13 Nov '98

sam.littlewood(a)quantel.com wrote: > Has anyone a fix for this rather curious bug? - it seems subranges, packed > records and VAR arguments conspire to make the compiler throw a bogus > error!? I don't have a fix (apart from the obvious workarounds such are using the base type instead of the subtype or a non-packed record, if these are options), but I added it to the list of known bugs (sam8.pas). -- Frank Heckenbach, frank(a)fjf.gnu.de http://fjf.gnu.de/ PGP and GPG keys: http://fjf.gnu.de/plan

1 0

Re: What is the latest ver of gpc?...
by Frank Heckenbach 12 Nov '98

12 Nov '98

Orlando Llanes wrote: > PROGRAM Tmp; > > var > F : File Of Byte; > V : Char; > > BEGIN > Assign( F, 'tmp.txt' ); > Reset( F ); > BlockRead( F, V, 1 ); > Close( F ); > END. BlockRead and BlockWrite only work with untyped files (simply `File'), not typed files (`File of Byte'), just like in BP. So try: PROGRAM Tmp; var F : File; V : Char; BEGIN Assign( F, 'tmp.txt' ); Reset( F, 1 ); BlockRead( F, V, 1 ); Close( F ); END. (The 1 in the `Reset' is needed because of compatibility to a stupid CPM legacy in TP.)-: -- Frank Heckenbach, frank(a)fjf.gnu.de http://fjf.gnu.de/ PGP and GPG keys: http://fjf.gnu.de/plan

3 2

Re: What is the latest ver of gpc?...
by Frank Heckenbach 12 Nov '98

12 Nov '98

Orlando Llanes wrote: > > BlockRead and BlockWrite only work with untyped files (simply `File'), > > not typed files (`File of Byte'), just like in BP. So try: > > I get a bunch of weird parsing errors if I only use FILE, besides, the > problem was not that the file was not being read, the problem is that > BlockRead is not being recognized as a built-in routine. Well, BlockRead will be rejected for typed files at compile time, not at run time... Anyway, please be more specific with your descriptions: what errors do you get with the original and the modified test program? Are you sure you're using the correct GPC -- try `gpc -v', and especially check if gpc1.exe (the actual compiler) is current. (ISTR, there were some directory changes under DJGPP recently that might cause an older previously installed version to be used...) -- Frank Heckenbach, frank(a)fjf.gnu.de http://fjf.gnu.de/ PGP and GPG keys: http://fjf.gnu.de/plan

1 0

Re: odd bug with constants
by Frank Heckenbach 11 Nov '98

11 Nov '98

Robert P. Ewing wrote: > I've been moving an old program over to GPC, and ran across > the following. Apparently the '**' operator works fine on > variables, but it crashes if I put a hard-coded numerical > value in for the base. The old y^x = exp(y ln x) > workaround from Turbo seems to work fine. I don't know whether > this is a new or a known bug. This seems to be a bug in how the arguments to some operations on LongReal's (aka Extended) are passed, and real constants are assumed Extended by default. I added it to the bug list. Thanks for the report. > I note that if I write 'real(10.0)' rather than '10.0', the > program runs just fine... This would indeed be a useful workaround for now. (But AFAIR, TP doesn't like this -- if that's of any concern to you...) Frank -- Frank Heckenbach, frank(a)fjf.gnu.de http://fjf.gnu.de/ PGP and GPG keys: http://fjf.gnu.de/plan

1 0

odd bug with constants
by Robert P. Ewing 11 Nov '98

11 Nov '98

Hi, all. I've been moving an old program over to GPC, and ran across the following. Apparently the '**' operator works fine on variables, but it crashes if I put a hard-coded numerical value in for the base. The old y^x = exp(y ln x) workaround from Turbo seems to work fine. I don't know whether this is a new or a known bug. Toby Ewing ewing(a)iastate.edu 515/294-7856 http://www.agron.iastate.edu/soilphysics/ewing.html Soil Scientist, Iowa State University --------------------------------------------------- program StarStar(input, output); var y : real; x : real value 10.0; begin writeln('via exp(x ln y):'); {all these are fine} y := 5.557; write('variables, y > 0 '); writeln('exp(y ln x) = ', exp(y * ln(x))); y := -5.557; write('variables, y < 0 '); writeln('exp(y ln x) = ', exp(y * ln(x))); y := 5.557; write('constants, y > 0 '); writeln('exp(y ln x) = ', exp(y * ln(10.0))); y := -5.557; write('constants, y < 0 '); writeln('exp(y ln x) = ', exp(y * ln(10.0))); readln; writeln('via y ** x:'); {now for problems} y := 5.557; write('variables, y > 0 '); writeln(' x ** y = ', (x ** y)); y := -5.557; write('variables, y < 0 '); writeln(' x ** y = ', (x ** y)); y := 5.557; write('constants, y > 0 '); writeln(' x ** y = ', (10.0 ** y)); { <--- crashes here } y := -5.557; write('constants, y < 0 '); writeln(' x ** y = ', (10.0 ** y)); { <-- or, if you prefer, here } end. I note that if I write 'real(10.0)' rather than '10.0', the program runs just fine...

1 0

Re: ncurses
by Frank Heckenbach 11 Nov '98

11 Nov '98

Matthew Wiggins wrote: > I was tring to use ncurses in a pascal program. and I defined a packed > record of the WINDOW type, but how do I import the stdscr etc etc so I can > use the functions? The simple answer is var StdScr : asmname 'stdscr' WINDOW; but please read on... > I guess some people have done it. I'm not sure... Yes, I've done it (when writing a BP compatible CRT unit), and the way I did it was via a C layer. So I made a file crtc.c which #include's <ncurses.h> and defines some routines on top of that. These routines are declared as "C;" in crt.pas (the Pascal unit). Some other routines in crt.pas are implemented in Pascal, but they only call the routines from crtc.c, not directly those from ncurses. The reason why I did it this way is because that guarantees some portability. E.g., the WINDOW type might change in different releases of ncurses, or in different curses libraries. -- In fact, there is a library called PDCurses which is widely ncurses compatible, but, not surprisingly, the internal structure of its WINDOW type is quite different. In the CRT unit, however, I can just drop in PDCurses for ncurses, and it works, because the only routines that need to know about the internals (and if only about the size of the WINDOW type) are written in C and can read the ncurses.h header. Of course, the better solution, as I pointed out in some other mails to this list, would be an automatic C header to Pascal translator which is quite a difficult thing to write. Anyone wants to try? -- Frank Heckenbach, frank(a)fjf.gnu.de http://fjf.gnu.de/ PGP and GPG keys: http://fjf.gnu.de/plan

1 0

Re: Ellipsis formal parameter & CRT module questions...
by Frank Heckenbach 11 Nov '98

11 Nov '98

kc7rad(a)vegasnet.net wrote: > 2) According to http://fjf.gnu.de/gpc-todo_6.html#SEC6 (Implemented > features) there is a portable ncurses CRT unit. Is this built into the RTL > (I couldn't find it on first look) or is it a seperate unit (or module)? It hasn't been published yet, but will be soon... -- Frank Heckenbach, frank(a)fjf.gnu.de http://fjf.gnu.de/ PGP and GPG keys: http://fjf.gnu.de/plan

1 0

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

Gpc