Porting pari/gp routines into a C program??
 

This will be detailed; possibly much more information than is needed, but I'm getting frustrated (again).

Background: I'm trying to learn how to use some pari/gp routines within some c programs, via an IDE called Code::Blocks. Unfortunately, I don't understand some of the declarations for the inner workings of pari/gp. I didn't think I needed to, to construct a simple executable. Apparently, I was wrong (yet, again).

I have the following function, called cycle02.gp, that looks for aliquot cycles:
[code]
findcycle(a,b)=
{
for(e=a,b,
i=e;
for(c=1,10,
i=sigma(i)-i;
if(i<e,
break()
);
if(i==e,
print(i," ",c);
break()
)
);
)
}
[/code]I can load this into gp and run it:
[code]
[user@comp pari]$ gp2c-run -g cycle02.gp
GP/PARI CALCULATOR Version 2.3.5 (released)
amd64 running linux (x86-64/GMP-4.3.2 kernel) 64-bit version
compiled: Feb 8 2011, gcc-4.6.0 20110205 (Red Hat 4.6.0-0.6) (GCC)
(readline v6.2 enabled [was v6.1 in Configure], extended help available)

Copyright (C) 2000-2006 The PARI Group

PARI/GP is free software, covered by the GNU General Public License, and
comes WITHOUT ANY WARRANTY WHATSOEVER.

Type ? for help, \q to quit.
Type ?12 for how to get moral (and possibly technical) support.

parisize = 8000000, primelimit = 500000
? findcycle(10000,20000)
10744 2
12285 2
12496 5
17296 2
?
[/code]gp2c creates the following file cycle02.gp.c:
[code]
/*-*- compile-command: "/usr/bin/gcc -c -o cycle02.gp.o -O3 -Wall -fno-strict-aliasing -fomit-frame-pointer -fPIC -I/usr/include cycle02.gp.c && /usr/bin/gcc -o cycle02.gp.so -shared -O3 -Wall -fno-strict-aliasing -fomit-frame-pointer -fPIC -Wl,-shared cycle02.gp.o -lc -ldl -lm -L/usr/lib64 -lgmp -L/usr/lib64 -lpari"; -*-*/
#include <pari/pari.h>
/*
GP;install("init_cycle02","v","init_cycle02","./cycle02.gp.so");
GP;install("findcycle","vD0,G,D0,G,","findcycle","./cycle02.gp.so");
*/
void init_cycle02(void);
void findcycle(GEN a, GEN b);
/*End of prototype*/

void
init_cycle02(void) /* void */
{
pari_sp ltop = avma;
avma = ltop;
return;
}

void
findcycle(GEN a, GEN b) /* void */
{
pari_sp ltop = avma;
GEN i = pol_x[fetch_user_var("i")];
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
GEN e = gen_0;
for (e = gcopy(a); gcmp(e, b) <= 0; e = gaddgs(e, 1))
{
i = e;
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
long c;
for (c = 1; c <= 10; ++c)
{
i = gsub(gsumdiv(i), i);
if (gcmp(i, e) < 0)
break;
if (gequal(i, e))
{
pariprintf("%Z %ld\n", i, c);
break;
}
if (low_stack(st_lim, stack_lim(btop, 1)))
i = gerepilecopy(btop, i);
}
}
if (low_stack(st_lim, stack_lim(btop, 1)))
gerepileall(btop, 2, &e, &i);
}
}
avma = ltop;
return;
}
[/code]I took the above routines and tried to build a basic program in c using Code::Blocks:
[code]
#include <stdio.h>
#include <stdlib.h>
#include "pari.h"

void init_cycle02(void);
void findcycle(GEN a, GEN b);


int main(int argc, char *argv[])
{
GEN j, k;

j=(GEN)100000;
k=(GEN)200000;

findcycle(j, k);
return 0;
}

void
init_cycle02(void) /* void */
{
pari_sp ltop = avma;
avma = ltop;
return;
}

void
findcycle(GEN a, GEN b) /* void */
{
pari_sp ltop = avma;
GEN i = pol_x[fetch_user_var("i")];
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
GEN e = gen_0;
for (e = gcopy(a); gcmp(e, b) <= 0; e = gaddgs(e, 1))
{
i = e;
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
long c;
for (c = 1; c <= 10; ++c)
{
i = gsub(gsumdiv(i), i);
if (gcmp(i, e) < 0)
break;
if (gequal(i, e))
{
pariprintf("%Z %ld\n", i, c);
break;
}
if (low_stack(st_lim, stack_lim(btop, 1)))
i = gerepilecopy(btop, i);
}
}
if (low_stack(st_lim, stack_lim(btop, 1)))
gerepileall(btop, 2, &e, &i);
}
}
avma = ltop;
return;
}
[/code]and, then I compiled and ran the program:
[code]
Process returned -1 (0xFFFFFFFF) execution time: 0.004 s
Press ENTER to continue.
[/code]So, then I ran the debugger:
[code]
Program received signal SIGSEGV, Segmentation fault.
In fetch_named_var () (usr/lib64/libpari-gmp.so.2)
Debugger finished with status 0
[/code]I am obviously missing something...

Would anyone like to point me in a direction where I can find the missing details on how to use pari/gp functions within c programs?

Thanks!

Presumably it's trying to access the global variable i, but there are no global variables since you're not in gp. Try changing

i=e;

into

my(i=e);

in the GP script to make the variable local, then see if it works. Also don't forget to use the -g option when running gp2c.

[QUOTE=CRGreathouse;316232]Presumably it's trying to access the global variable i, but there are no global variables since you're not in gp. Try changing

i=e;

into

my(i=e);

in the GP script to make the variable local, then see if it works. Also don't forget to use the -g option when running gp2c.[/QUOTE]
Thanks for the help, but I got no detectable change. I do normally run the -g option, but simply skipped it for the small sample. I did use it this time.

I should be able to do what I'm try to, shouldn't I?

Maybe I'll try something more simple...

If/when this does work, will the c version have a noticeable difference in speed over running it via gp2c-run?

I tried again with a version that's about as simple as I can get. I used the following gp routine:
[code]
pTest(a,b)=
{
for(e=a,b,
print("e is ",e)
)
}
[/code]Terminal result:
[code]
[user@comp pari]$ gp2c-run -g pariTest.gp
GP/PARI CALCULATOR Version 2.3.5 (released)

...

Type ? for help, \q to quit.
Type ?12 for how to get moral (and possibly technical) support.

parisize = 8000000, primelimit = 500000
? pTest(15,32)
e is 15
e is 16
...
e is 31
e is 32
?
[/code]The PariTest.gp.c file:
[code]
/*-*- compile-command: "/usr/bin/gcc -c -o pariTest.gp.o -O3 -Wall -fno-strict-aliasing -fomit-frame-pointer -fPIC -I/usr/include pariTest.gp.c && /usr/bin/gcc -o pariTest.gp.so -shared -O3 -Wall -fno-strict-aliasing -fomit-frame-pointer -fPIC -Wl,-shared pariTest.gp.o -lc -ldl -lm -L/usr/lib64 -lgmp -L/usr/lib64 -lpari"; -*-*/
#include <pari/pari.h>
/*
GP;install("init_pariTest","v","init_pariTest","./pariTest.gp.so");
GP;install("pTest","vD0,G,D0,G,","pTest","./pariTest.gp.so");
*/
void init_pariTest(void);
void pTest(GEN a, GEN b);
/*End of prototype*/

void
init_pariTest(void) /* void */
{
pari_sp ltop = avma;
avma = ltop;
return;
}

void
pTest(GEN a, GEN b) /* void */
{
pari_sp ltop = avma;
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
GEN e = gen_0;
for (e = gcopy(a); gcmp(e, b) <= 0; e = gaddgs(e, 1))
{
pariprintf("e is %Z\n", e);
if (low_stack(st_lim, stack_lim(btop, 1)))
e = gerepilecopy(btop, e);
}
}
avma = ltop;
return;
}
[/code]Debugger message:
[code]
Program received signal SIGSEGV, Segmentation fault.
In gcopy () (usr/lib64/libpari-gmp.so.2)
[/code]
Thanks for any assistance...

EdH, One problem I see with your program is that there is no pari_init() at the beginning. You need something like:

[CODE] pari_init( 1e9, 1e8 );[/CODE]before you execute any pari code.

It's also good form to call:

[CODE] pari_close();
[/CODE]at the end.

Another problem is the casts:

[CODE] j=(GEN)100000;
k=(GEN)200000;
[/CODE]You are assigning integers into pointers here. I really don't know much about pari, but the proper coding techniques are to be found in the "libpari.dvi" documentation.

Cheers

[QUOTE=jcrombie;316254]Another problem is the casts:

[CODE] j=(GEN)100000;
k=(GEN)200000;
[/CODE]You are assigning integers into pointers here. I really don't know much about pari, but the proper coding techniques are to be found in the "libpari.dvi" documentation.

Cheers[/QUOTE]


Ah, good catch -- I was just looking at the code that gp2c generated. Those should be
[code]j = utoipos(100000);
k = utoipos(200000);[/code]

Thank you both for the help, but it still resists my efforts:
[code]
Program received signal SIGSEGV, Segmentation fault.
In pariputc () (/usr/lib64/libpari-gmp.so.2)
[/code]I had wondered what I was supposed to do with the created init_pariTest function, also. I tried simply inserting it at the beginning of main, but it had no effect, either. Does it belong somewhere?

Off to research libpari.dvi docs...

Would you post the code you're using now, with the various modifications?

This is my latest very basic testing setup from start to finish.

pariTest.gp:
[code]
pTest(a,b)=
{
for(e=a,b,
print("e is ",e)
)
}
[/code]terminal session:
[code]
[user@comp pari]$ gp2c-run -g pariTest.gp
GP/PARI CALCULATOR Version 2.3.5 (released)
amd64 running linux (x86-64/GMP-4.3.2 kernel) 64-bit version
compiled: Feb 8 2011, gcc-4.6.0 20110205 (Red Hat 4.6.0-0.6) (GCC)
(readline v6.2 enabled [was v6.1 in Configure], extended help available)

Copyright (C) 2000-2006 The PARI Group

PARI/GP is free software, covered by the GNU General Public License, and
comes WITHOUT ANY WARRANTY WHATSOEVER.

Type ? for help, \q to quit.
Type ?12 for how to get moral (and possibly technical) support.

parisize = 8000000, primelimit = 500000
? pTest(11,17)
e is 11
e is 12
e is 13
e is 14
e is 15
e is 16
e is 17
?
[/code]generated pariTest.gp.c
[code]
/*-*- compile-command: "/usr/bin/gcc -c -o pariTest.gp.o -O3 -Wall -fno-strict-aliasing -fomit-frame-pointer -fPIC -I/usr/include pariTest.gp.c && /usr/bin/gcc -o pariTest.gp.so -shared -O3 -Wall -fno-strict-aliasing -fomit-frame-pointer -fPIC -Wl,-shared pariTest.gp.o -lc -ldl -lm -L/usr/lib64 -lgmp -L/usr/lib64 -lpari"; -*-*/
#include <pari/pari.h>
/*
GP;install("init_pariTest","v","init_pariTest","./pariTest.gp.so");
GP;install("pTest","vD0,G,D0,G,","pTest","./pariTest.gp.so");
*/
void init_pariTest(void);
void pTest(GEN a, GEN b);
/*End of prototype*/

void
init_pariTest(void) /* void */
{
pari_sp ltop = avma;
avma = ltop;
return;
}

void
pTest(GEN a, GEN b) /* void */
{
pari_sp ltop = avma;
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
GEN e = gen_0;
for (e = gcopy(a); gcmp(e, b) <= 0; e = gaddgs(e, 1))
{
pariprintf("e is %Z\n", e);
if (low_stack(st_lim, stack_lim(btop, 1)))
e = gerepilecopy(btop, e);
}
}
avma = ltop;
return;
}
[/code]c program:
[code]
#include <stdio.h>
#include <stdlib.h>
#include "pari.h"

void init_pariTest(void);
void pTest(GEN a, GEN b);


int main()
{
GEN c, d;

c=utoipos(11);
d=utoipos(17);

pari_init(1e9,1e8);
init_pariTest();

pTest(c,d);

pari_close();

printf("\nFinished!\n");
return 0;
}

void
init_pariTest(void) /* void */
{
pari_sp ltop = avma;
avma = ltop;
return;
}

void
pTest(GEN a, GEN b) /* void */
{
pari_sp ltop = avma;
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
GEN e = gen_0;
for (e = gcopy(a); gcmp(e, b) <= 0; e = gaddgs(e, 1))
{
pariprintf("e is %Z\n", e);
if (low_stack(st_lim, stack_lim(btop, 1)))
e = gerepilecopy(btop, e);
}
}
avma = ltop;
return;
}
[/code]c program session output:
[code]
Process returned -1 (0xFFFFFFFF) execution time : 0.004 s
Press ENTER to continue.
[/code]debugger output:
[code]
Building to ensure sources are up-to-date
Build succeeded
Selecting target:
Release
Adding source dir: /home/user/Programming/pariTest/
Adding source dir: /home/user/Programming/pariTest/
Adding file: bin/Release/pariTest
Starting debugger:
done
Registered new type: wxString
Registered new type: STL String
Registered new type: STL Vector
Setting breakpoints
(no debugging symbols found)...done.
Debugger name and version: GNU gdb (GDB) Fedora (7.3.50.20110722-16.fc16)
Program received signal SIGSEGV, Segmentation fault.
In pariputc () (/usr/lib64/libpari-gmp.so.2)
[/code]My program console gives me the following message:
[code]
warning: GDB: Failed to set controlling terminal: Operation not permitted
[/code]Call stack:
[code]
#0 0x36e27f53fd pariputc() (/usr/lib64/libpari-gmp.so.2:??)
#1 0x36e27fe74a pari_err() (/usr/lib64/libpari-gmp.so.2:??)
#2 ( 0x0000000000400976 in ??() (??:??)
#3 ( 0x000000000040099f in ??() (??:??)
#4 ( 0x00000000004009db in ??() (??:??)
#5 ( 0x0000000000400a15 in ??() (??:??)
#6 0x36e122169d __libc_start_main() (/lib64/libc.so.6:??)
#7 ( 0x0000000000400839 in ??() (??:??)
#8 ( 0x00007fffffffe708 in ??() (??:??)
#9 ( 0x000000000000001c in ??() (??:??)
#10 ( 0x0000000000000001 in ??() (??:??)
#11 ( 0x00007fffffffea03 in ??() (??:??)
#12 ( 0x0000000000000000 in ??() (??:??)
[/code]Thanks for any help. I have the aforementioned documents and am looking into them. I haven't tried constructing this in the referenced Makefile manner, since I've never programmed with makefiles. I have started looking into them, though...

I took the output of gp2c -g and added a short main function to produce
[CODE]/*-*- compile-command: "cc -c -o ptest.gp.o -g -O3 -Wall -fomit-frame-pointer -fno-strict-aliasing -fPIC -I"/usr/include" ptest.gp.c && cc -o ptest.gp.so -shared -g -O3 -Wall -fomit-frame-pointer -fno-strict-aliasing -fPIC -Wl,-shared ptest.gp.o -lc -lm -L/usr/lib -lpari"; -*-*/
#include <pari/pari.h>
/*
GP;install("init_ptest","v","init_ptest","./ptest.gp.so");
GP;install("pTest","vD0,G,D0,G,","pTest","./ptest.gp.so");
*/
void init_ptest(void);
void pTest(GEN a, GEN b);
/*End of prototype*/

void
init_ptest(void) /* void */
{
pari_sp ltop = avma;
avma = ltop;
return;
}

void
pTest(GEN a, GEN b) /* void */
{
pari_sp ltop = avma;
{
pari_sp btop = avma, st_lim = stack_lim(btop, 1);
GEN e = gen_0;
for (e = gcopy(a); gcmp(e, b) <= 0; e = gaddgs(e, 1))
{
pari_printf("e is %Ps\n", e);
if (low_stack(st_lim, stack_lim(btop, 1)))
e = gerepilecopy(btop, e);
}
}
avma = ltop;
return;
}

int main() {
long a, b;
GEN x, y;
pari_init(1000000,2);
init_ptest();
printf("a = ");
scanf("%ld",&a);
printf("b = ");
scanf("%ld",&b);
x = stoi(a);
y = stoi(b);
pTest(x, y);
pari_close();
return 0;
}
[/CODE]

This I compiled with
[CODE]gcc -o ptest -g -O3 -Wall -fomit-frame-pointer -fno-strict-aliasing -fPIC -I"/usr/include" ptest.gp.c -lc -lm -L/usr/lib -lpari
[/CODE]

This runs as expected.

[QUOTE=EdH;316242]If/when this does work, will the c version have a noticeable difference in speed over running it via gp2c-run?[/QUOTE]
I doubt it. The idea of gp2c-run is to combine the speed of the c compiled function with the ease of use of the gp calculator.

Thanks frmky,

I must have something else wrong. Both my IDE and command line give me the same trouble:

With pari_init(1000000000,2); or smaller:
[code]
[user@comp pariTest]$ ./ptest
a = 11
b = 17

*** the PARI stack overflows !
current stack size: 100000000 (95.367 Mbytes)
[hint] you can increase GP stack with allocatemem()

*** Error in the PARI system. End of program.
[/code]With pari_init(10000000000,2);:
[code]
[user@comp pariTest]$ ./ptest
a = 11
b = 17

*** segmentation fault: bug in PARI or calling program.
*** Error in the PARI system. End of program.
[/code]Must be a flaw in my system.:sad:

Off to try another system...

The test program works on my system as well so it must be something with yours...

There's a minor printing issue with the first machine, but the second one is frustrating. The second one, is my main system and I totally removed/reinstalled from scratch, both pari and gp2c.

machine math06 (Ubuntu 32-bit, repository versions of pari and gp2c):
[code]
math06@math06:~/Mathwork/pari$ whereis pari
pari: /usr/lib/pari /usr/include/pari /usr/share/pari /usr/share/man/man1/pari.1.gz
math06@math06:~/Mathwork/pari$ whereis gp2c
gp2c: /usr/bin/gp2c /usr/share/gp2c /usr/share/man/man1/gp2c.1.gz
math06@math06:~/Mathwork/pari$ gcc -o ptest -g -O3 -Wall -fomit-frame-pointer -fno-strict-aliasing -fPIC -I"/usr/include" pariTest.gp.c -lc -lm -L/usr/lib -lpari
pariTest.gp.c: In function 'pTest':
pariTest.gp.c:28:7: warning: implicit declaration of function 'pari_printf'
pariTest.gp.c: In function 'main':
pariTest.gp.c:43:8: warning: ignoring return value of 'scanf', declared with attribute warn_unused_result
pariTest.gp.c:45:8: warning: ignoring return value of 'scanf', declared with attribute warn_unused_result
/tmp/ccgpmm0C.o: In function `pTest':
/home/math06/Mathwork/pari/pariTest.gp.c:28: undefined reference to `pari_printf'[B]***[/B]
collect2: ld returned 1 exit status
math06@math06:~/Mathwork/pari$ gcc -o ptest -g -O3 -Wall -fomit-frame-pointer -fno-strict-aliasing -fPIC -I"/usr/include" pariTest.gp.c -lc -lm -L/usr/lib -lpari
pariTest.gp.c: In function 'main':
pariTest.gp.c:43:8: warning: ignoring return value of 'scanf', declared with attribute warn_unused_result
pariTest.gp.c:45:8: warning: ignoring return value of 'scanf', declared with attribute warn_unused_result
math06@math06:~/Mathwork/pari$ ./ptest
a = 11
b = 17
e is %Ps
e is %Ps
e is %Ps
e is %Ps
e is %Ps
e is %Ps
e is %Ps
[/code][B]***[/B]I had to change pari_print() to pariprint().

machine comp (Fedora 64-bit, manual installation of pari and gp2c):
[code]
[user@comp examples]$ whereis pari
pari: /usr/local/lib/pari
[user@comp examples]$ whereis libpari
libpari: /usr/local/lib/libpari.so
[user@comp examples]$ whereis libpari-gmp.so.3
libpari-gmp.so: /usr/local/lib/libpari-gmp.so.3
[user@comp examples]$ gcc -o ptest -g -O3 -Wall -fomit-frame-pointer -fno-strict-aliasing -fPIC -I"/usr/local/lib" ptest.gp.c -lc -lm -L/usr/local/lib -lpari
[user@comp examples]$ ./ptest
./ptest: error while loading shared libraries: libpari-gmp.so.3: cannot open shared object file: No such file or directory
[user@comp examples]$ gcc -o ptest -g -O3 -Wall -fomit-frame-pointer -fno-strict-aliasing -fPIC -I"/usr/local/lib" ptest.gp.c -lc -lm -L"/usr/local/lib" -lpari
[user@comp examples]$ ./ptest
./ptest: error while loading shared libraries: libpari-gmp.so.3: cannot open shared object file: No such file or directory
[/code]:sad:

The problem on the 64-bit Fedora computer sounds like an ldconfig issue. See [url]http://www.brunolinux.com/02-The_Terminal/ldconfig.html[/url].

[QUOTE=frmky;316355]The problem on the 64-bit Fedora computer sounds like an ldconfig issue. See [URL]http://www.brunolinux.com/02-The_Terminal/ldconfig.html[/URL].[/QUOTE]

That was what I was thinking too.

I would try just typing: "sudo ldconfig" first without editing any system files and see if that works.

Also, you can type: "ldd ptest" to check if all shared libraries are going to be found at runtime without having to run the program.

Thanks guys! That did the trick, although it was very convoluted on this Fedora system. sudo anything doesn't work and you can't do any GUI type file management or text editing as the superuser. Also the ld.so.conf file only contains a single line to include all the files in the ld.so.conf.d directory. So, instead of trying to edit it, I wrote usr-local-lib.conf elsewhere (as user) and cp'd that into the ld.so.conf.d directory using superuser.

Prior to usr-local-lib.conf creation:
[code]
[user@comp examples]$ ldd ptest
linux-vdso.so.1 => (0x00007fff0f95b000)
libc.so.6 => /lib64/libc.so.6 (0x00000036e1200000)
libm.so.6 => /lib64/libm.so.6 (0x00000036e2200000)
/lib64/ld-linux-x86-64.so.2 (0x00000036e0e00000)
libpari-gmp.so.3 => not found
[/code]After usr-local-lib.conf creation:
[code]
[user@comp examples]$ ldd ptest
linux-vdso.so.1 => (0x00007fff005cb000)
libc.so.6 => /lib64/libc.so.6 (0x00000036e1200000)
libm.so.6 => /lib64/libm.so.6 (0x00000036e2200000)
libpari-gmp.so.3 => /usr/local/lib/libpari-gmp.so.3 (0x00007f64ded57000)
/lib64/ld-linux-x86-64.so.2 (0x00000036e0e00000)
libgmp.so.3 => /usr/lib64/libgmp.so.3 (0x00000036e3200000)
[user@comp examples]$ ./ptest
a = 11
b = 17
e is 11
e is 12
e is 13
e is 14
e is 15
e is 16
e is 17
[/code]BTW, ldconfig ran under superuser, but made no difference until I manually added the file to the directory.

Now, I'll see what happens when I modify it more into what I'm after and eventually see if I can get it all working in my IDE.

Thanks much, everyone...

[QUOTE=frmky;316304]I doubt it. The idea of gp2c-run is to combine the speed of the c compiled function with the ease of use of the gp calculator.[/QUOTE]
They do indeed seem to run at extremely close to the same speed.

I still haven't figured out how to compile within my IDE. But, I'm going to study makefiles for a while, anyway, and set the IDE aside for these things...

Thanks for all...