Developer’s Guide

to

the PARI library

(version 2.13.3)

The PARI Group

Institut de Mathématiques de Bordeaux, UMR 5251 du CNRS.

Université de Bordeaux, 351 Cours de la Libération
F-33405 TALENCE Cedex, FRANCE
e-mail: pari@math.u-bordeaux.fr

Home Page:
http://pari.math.u-bordeaux.fr/
Copyright c 2000–2020 The PARI Group

Permission is granted to make and distribute verbatim copies of this manual provided the copyright
notice and this permission notice are preserved on all copies.

Permission is granted to copy and distribute modified versions, or translations, of this manual
under the conditions for verbatim copying, provided also that the entire resulting derived work is
distributed under the terms of a permission notice identical to this one.

PARI/GP is Copyright c 2000–2020 The PARI Group

PARI/GP is free software; you can redistribute it and/or modify it under the terms of the GNU
General Public License as published by the Free Software Foundation. It is distributed in the hope
that it will be useful, but WITHOUT ANY WARRANTY WHATSOEVER.
 Table of Contents
Chapter 1: Work in progress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1 The type t_CLOSURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Debugging information in closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 The type t_LIST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.1 Maps as Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Protection of noninterruptible code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.1 Multithread interruptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 F l2 field for small primes l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Public functions useless outside of GP context . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5.1 Conversions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5.2 Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5.3 Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5.4 Control flow statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5.5 Accessors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5.6 Iterators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5.7 Local precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5.8 Functions related to the GP evaluator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5.9 Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 Embedded GP interpretor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Readline interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Constructors called by pari init functions . . . . . . . . . . . . . . . . . . . . . . . . 13
1.9 Destructors called by pari close . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.10 Constructors and destructors used by the pthreads interface . . . . . . . . . . . . . . 14
Chapter 2: Regression tests, benches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 Functions for GP2C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.1 Functions for safe access to components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Chapter 3: Parallelism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1 The PARI multithread interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Technical functions required by MPI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 A complete example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4
 Chapter 1:
Work in progress

This draft documents private internal functions and structures for hard-core PARI developers.
Anything in here is liable to change on short notice. Don’t use anything in the present document,
unless you are implementing new features for the PARI library. Try to fix the interfaces before
using them, or document them in a better way. If you find an undocumented hack somewhere, add
it here.
Hopefully, this will eventually document everything that we buried in paripriv.h or even
more private header files like anal.h. Possibly, even implementation choices! Way to go.

1.1 The type t_CLOSURE.

This type holds closures and functions in compiled form, so is deeply linked to the internals
of the GP compiler and evaluator. The length of this type can be 6, 7 or 8 depending whether the
object is an “inline closure”, a “function” or a “true closure”.
A function is a regular GP function. The GP input line is treated as a function of arity 0.
A true closure is a GP function defined in a nonempty lexical context.
An inline closure is a closure that appears in the code without the preceding -> token. They
are generally attached to the prototype code ’E’ and ’I’. Inline closures can only exist as data of
other closures, see below.
In the following example,
f(a=Euler)=x->sin(x+a);
g=f(Pi/2);
plot(x=0,2*Pi,g(x))
f is a function, g is a true closure and both Euler and g(x) are inline closures.
This type has a second codeword z[1], which is the arity of the function or closure. This is
zero for inline closures. To access it, use
long closure_arity(GEN C)
• z[2] points to a t_STR which holds the opcodes. To access it, use
GEN closure_get_code(GEN C).
const char * closure_codestr(GEN C) returns as an array of char starting at 1.
• z[3] points to a t_VECSMALL which holds the operands of the opcodes. To access it, use
GEN closure_get_oper(GEN C)
• z[4] points to a t_VEC which hold the data referenced by the pushgen opcodes, which can
be t_CLOSURE, and in particular inline closures. To access it, use
GEN closure_get_data(GEN C)

5
 • z[5] points to a t_VEC which hold extra data needed for error-reporting and debugging. See
Section 1.1.1 for details. To access it, use
GEN closure_get_dbg(GEN C)
Additionally, for functions and true closures,
• z[6] usually points to a t_VEC with two components which are t_STR. The first one displays
the list of arguments of the closure without the enclosing parentheses, the second one the GP code
of the function at the right of the -> token. They are used to display the closure, either in implicit
or explicit form. However for closures that were not generated from GP code, z[6] can point to a
t_STR instead. To access it, use
GEN closure_get_text(GEN C)
Additionally, for true closure,
• z[7] points to a t_VEC which holds the values of all lexical variables defined in the scope
the closure was defined. To access it, use
GEN closure_get_frame(GEN C)

1.1.1 Debugging information in closure.

Every t_CLOSURE object z has a component dbg=z[5] which hold extra data needed for error-
reporting and debugging. The object dbg is a t_VEC with 3 components:
dbg[1] is a t_VECSMALL of the same length than z[3]. For each opcode, it holds the position
of the corresponding GP source code in the strings stored in z[6] for function or true closures,
positive indices referring to the second strings, and negative indices referring to the first strings,
the last element being indexed as −1. For inline closures, the string of the parent function or true
closure is used instead.
dbg[2] is a t_VECSMALL that lists opcodes index where new lexical local variables are created.
The value 0 denotes the position before the first offset and variables created by the prototype code
’V’.
dbg[3] is a t_VEC of t_VECSMALLs that give the list of entree* of the lexical local variables
created at a given index in dbg[2].

1.2 The type t_LIST.

This type needs to go through various hoops to support GP’s inconvenient memory model.
Don’t use t_LISTs in pure library mode, reimplement ordinary lists! This dynamic type is imple-
mented by a GEN of length 3: two codewords and a vector containing the actual entries. In a normal
setup (a finished list, ready to be used),
• the vector is malloc’ed, so that it can be realloc’ated without moving the parent GEN.
• all the entries are clones, possibly with cloned subcomponents; they must be deleted with
gunclone_deep, not gunclone.
The following macros are proper lvalues and access the components
long list_nmax(GEN L): current maximal number of elements. This grows as needed.

6
GEN list_data(GEN L): the elements. If v = list data(L), then either v is NULL (empty list) or
l = lg(v) is defined, and the elements are v[1], . . . , v[l-1].
In most gerepile scenarios, the list components are not inspected and a shallow copy of the
malloc’ed vector is made. The functions gclone, copy bin canon are exceptions, and make a full
copy of the list.
The main problem with lists is to avoid memory leaks; in the above setup, a statement like a
= List(1) would already leak memory, since List(1) allocates memory, which is cloned (second
allocation) when assigned to a; and the original list is lost. The solution we implemented is
• to create anonymous lists (from List, gtolist, concat or vecsort) entirely on the stack,
not as described above, and to set list nmax to 0. Such a list is not yet proper and trying to
append elements to it fails:
? listput(List(),1)
*** variable name expected: listput(List(),1)
*** ^----------------
If we had been malloc’ing memory for the List([1,2,3]), it would have leaked already.
• as soon as a list is assigned to a variable (or a component thereof) by the GP evaluator, the
assigned list is converted to the proper format (with list nmax set) previously described.
GEN listcopy(GEN L) return a full copy of the t_LIST L, allocated on the stack (hence list nmax
is 0). Shortcut for gcopy.
GEN mklistcopy(GEN x) returns a list with a single element x, allocated on the stack. Used to
implement most cases of gtolist (except vectors and lists).
A typical low-level construct:
long l;
/* assume L is a t_LIST */
L = list_data(L); /* discard t_LIST wrapper */
l = L? lg(L): 1;
for (i = 1; i < l; i++) output( gel(L, i) );
for (i = 1; i < l; i++) gel(L, i) = gclone( ... );

1.2.1 Maps as Lists.

GP’s maps are implemented on top of t_LISTs so as to benefit from their peculiar memory
models. Lists thus come in two subtypes: t_LIST RAW (actual lists) and t_LIST MAP (a map).
GEN mklist_typ(long t) create a list of subtype t. GEN mklist(void) is an alias for
mklist_typ(t_LIST_RAW);
and GEN mkmap(void) is an alias for
mklist_typ(t_LIST_MAP);
long list_typ(GEN L) return the list subtype, either t_LIST RAW or t_LIST MAP.
void listpop(GEN L, long index) as listpop0, assuming that L is a t_LIST RAW.
GEN listput(GEN list, GEN object, long index) as listput0, assuming that L is a
t_LIST RAW.

7
GEN mapdomain(GEN T) vector of keys of the map T .

GEN mapdomain_shallow(GEN T) shallow version of mapdomain.

GEN maptomat(GEN T) convert a map to a factorization matrix.

GEN maptomat_shallow(GEN T) shallow version of maptomat.

1.3 Protection of noninterruptible code.

GP allows the user to interrupt a computation by issuing SIGINT (usually by entering control-
C) or SIGALRM (usually using alarm()). To avoid such interruption to occurs in section of code
which are not reentrant (in particular malloc and free) the following mechanism is provided:

BLOCK_SIGINT_START() Start a noninterruptible block code. Block both SIGINT and SIGARLM.

BLOCK_SIGALRM_START() Start a noninterruptible block code. Block only SIGARLM. This is used
in the SIGINT handler itself to delay an eventual pending alarm.

BLOCK_SIGINT_END() End a noninterruptible block code

The above macros make use of the following global variables:

PARI_SIGINT_block: set to 1 (resp. 2) by BLOCK SIGINT START (resp. BLOCK SIGALRM START).

PARI_SIGINT_pending: Either 0 (no signal was blocked), SIGINT (SIGINT was blocked) or
SIGALRM (SIGALRM was blocked). This need to be set by the signal handler.

Within a block, an automatic variable int block is defined which records the value of
PARI SIGINT block when entering the block.

1.3.1 Multithread interruptions.

To support multithreaded programs, BLOCK SIGINT START and BLOCK SIGALRM START call
MT SIGINT BLOCK(block), and BLOCK SIGINT END calls MT SIGINT UNBLOCK(block).

MT_SIGINT_BLOCK and MT_SIGINT_UNBLOCK are defined by the multithread engine. They can
calls the following public functions defined by the multithread engine.

void mt_sigint_block(void)

void mt_sigint_unblock(void)

In practice this mechanism is used by the POSIX thread engine to protect against asychronous
cancellation.

8
1.4 Fl2 field for small primes l.
Let l > 2 be a prime ulong. A Fl2 is an element of the finite field Fl2 represented (currently) by
a Flx of degree at most 1 modulo a polynomial of the form x2 − D for some non square 0 ≤ D < p.
Below pi denotes the pseudo inverse of p, see Fl mul pre
int Fl2_equal1(GEN x) return 1 if x = 1, else return 0.
GEN Fl2_mul_pre(GEN x, GEN y, ulong D, ulong p, ulong pi) return xy.
GEN Fl2_sqr_pre(GEN x, ulong D, ulong p, ulong pi) return x2 .
GEN Fl2_inv_pre(GEN x, ulong D, ulong p, ulong pi) return x−1 .
GEN Fl2_pow_pre(GEN x, GEN n, ulong D, ulong p, ulong pi) return xn .
GEN Fl2_sqrtn_pre(GEN a, GEN n, ulong D, ulong p, ulong pi, GEN *zeta) n-th root, as
Flxq sqrtn.
GEN Fl2_norm_pre(GEN x, GEN n, ulong D, ulong p, ulong pi) return the norm of x.
GEN Flx_Fl2_eval_pre(GEN P, GEN x, ulong D, ulong p, ulong pi) return P (x).

1.5 Public functions useless outside of GP context.

These functions implement GP functionality for which the C language or other libpari routines
provide a better equivalent; or which are so tied to the gp interpreter as to be virtually useless in
libpari. Some may be generated by gp2c. We document them here for completeness.

1.5.1 Conversions.
GEN toser_i(GEN x) internal shallow function, used to implement automatic conversions to power
series in GP (as in cos(x)). Converts a t_POL or a t_RFRAC to a t_SER in the same variable
and precision precdl (the global variable corresponding to seriesprecision). Returns x itself
for a t_SER, and NULL for other argument types. The fact that it uses a global variable makes it
awkward whenever you’re not implementing a new transcendental function in GP. Use RgX_to_ser
or rfrac_to_ser for a fast clean alternative to gtoser.
GEN listinit(GEN x) a t_LIST (from List or Map) may exist in two different forms due to GP
memory model:
• an ordinary read-only copy on the PARI stack (as produced by gtolist or gtomap) to which
one may not assign elements (listput will fail) unless the list is empty.
• a feature-complete GP list using (malloc’ed) blocks to allow dynamic insertions. An empty
list is automaticaly promoted to this status on first insertion.
The listinit function creates a copy of existing t_SER x and makes sure it is of the second
kind. Variants of this are automatically called by gp when assigning a t_LIST to a GP variable;
the mecanism avoid memory leaks when creating a constant list, e.g. List([1,2,3]) (read-only),
without assigning it to a variable. Whereas after L = List([1,2,3]) (GP list), we keep a pointer
to the object and may delete it when L goes out of scope.
This libpari function allows gp2c to simulate this process by generating listinit calls at
appropriate places.

9
1.5.2 Output.
void print0(GEN g, long flag) internal function underlying the print GP function. Prints
the entries of the t_VEC g, one by one, without any separator; entries of type t_STR are printed
without enclosing quotes. flagis one of f_RAW, f_PRETTYMAT or f_TEX, using the current default
output context.
void out_print0(PariOUT *out, const char *sep, GEN g, long flag) as print0, using
output context out and separator sep between successive entries (no separator if NULL).
void printsep(const char *s, GEN g, long flag) out_print0 on pariOut followed by a
newline.
void printsep1(const char *s, GEN g, long flag) out_print0 on pariOut.
char* pari_sprint0(const char *s, GEN g, long flag) displays s, then print0(g, flag).
void print(GEN g) equivalent to print0(g, f RAW), followed by a \n then an fflush.
void printp(GEN g) equivalent to print0(g, f PRETTYMAT), followed by a \n then an fflush.
void print1(GEN g) as above, without the \n. Use pari_printf or output instead.
void printtex(GEN g) equivalent to print0(g, t TEX), followed by a \n then an fflush. Use
GENtoTeXstr and pari_printf instead.
void write0(const char *s, GEN g)
void write1(const char *s, GEN g) use fprintf
void writetex(const char *s, GEN g) use GENtoTeXstr and fprintf.
void printf0(GEN fmt, GEN args) use pari_printf.
GEN strprintf(GEN fmt, GEN args) use pari_sprintf.

1.5.3 Input.
gp’s input is read from the stream pari_infile, which is changed using
FILE* switchin(const char *name)
Note that this function is quite complicated, maintaining stacks of files to allow smooth error
recovery and gp interaction. You will be better off using gp_read_file.

1.5.4 Control flow statements.

GEN break0(long n). Use the C control statement break. Since break(2) is invalid in C, either
rework your code or use goto.
GEN next0(long n). Use the C control statement continue. Since continue(2) is invalid in C,
either rework your code or use goto.
GEN return0(GEN x). Use return!
void error0(GEN g). Use pari err(e USER,)
void warning0(GEN g). Use pari warn(e USER,)

10
1.5.5 Accessors.

GEN vecslice0(GEN A, long a, long b) implements A[a..b].

GEN matslice0(GEN A, long a, long b, long c, long d) implements A[a..b, c..d].

1.5.6 Iterators.

GEN apply0(GEN f, GEN A) gp wrapper calling genapply, where f is a t_CLOSURE, applied to A.

Use genapply or a standard C loop.

GEN select0(GEN f, GEN A) gp wrapper calling genselect, where f is a t_CLOSURE selecting

from A. Use genselect or a standard C loop.

GEN vecapply(void *E, GEN (*f)(void* E, GEN x), GEN x) implements [a(x)|x<-b].

GEN veccatapply(void *E, GEN (*f)(void* E, GEN x), GEN x) implements concat([a(x)|x<-
b]) which used to implement [a0(x,y)|x<-b;y<-c(b)] which is equal to concat([[a0(x,y)|y<-
c(b)]|x<-b]).

GEN vecselect(void *E, long (*f)(void* E, GEN x), GEN A) implements [x<-b,c(x)].

GEN vecselapply(void *Epred, long (*pred)(void* E, GEN x), void *Efun, GEN
(*fun)(void* E, GEN x), GEN A) implements [a(x)|x<-b,c(x)].

1.5.7 Local precision.

These functions allow to change realprecision locally when calling the GP interpretor.

void push_localprec(long p) set the local precision to p.

void push_localbitprec(long b) set the local precision to b bits.

void pop_localprec(void) reset the local precision to the previous value.

long get_localprec(void) returns the current local precision.

long get_localbitprec(void) returns the current local precision in bits.

void localprec(long p) trivial wrapper around push localprec (sanity checks and convert from
decimal digits to a number of codewords). Use push localprec.

void localbitprec(long p) trivial wrapper around push localbitprec (sanity checks). Use
push localbitprec.

These two function are used to implement getlocalprec and getlocalbitprec for the GP
interpreter and essentially return their argument (the current dynamic precision, respectively in
bits or as a prec word count):

long getlocalbitprec(long bit)

long getlocalprec(long prec)

11
1.5.8 Functions related to the GP evaluator.

The prototype code C instructs the GP compiler to save the current lexical context (pairs made
of a lexical variable name and its value) in a GEN, called pack in the sequel. This pack can be used
to evaluate expressions in the corresponding lexical context, providing it is current.

GEN localvars_read_str(const char *s, GEN pack) evaluate the string s in the lexical context
given by pack. Used by geval_gp in GP to implement the behavior below:

? my(z=3);eval("z=z^2");z
%1 = 9

long localvars_find(GEN pack, entree *ep) does pack contain a pair whose variable corre-
sponds to ep? If so, where is the corresponding value? (returns an offset on the value stack).

1.5.9 Miscellaneous.

char* os_getenv(const char *s) either calls getenv, or directly return NULL if the libc does
not provide it. Use getenv.

sighandler_t os_signal(int sig, pari_sighandler_t fun) after a

typedef void (*pari_sighandler_t)(int);

(private type, not exported). Installs signal handler fun for signal sig, using sigaction with flag
SA_NODEFER. If sigaction is not available use signal. If even the latter is not available, just return
SIG_IGN. Use sigaction.

1.6 Embedded GP interpretor.

These function provide a simplified interface to embed a GP interpretor in a program.

void gp_embedded_init(long rsize, long vsize) Initialize the GP interpretor (like pari init
does) with parisize=rsize rsize and parisizemax=vsize.

char * gp_embedded(const char *s) Evaluate the string s with GP and return the result as a
string, in a format similar to what GP displays (with the history index). The resulting string is
allocated on the PARI stack, so subsequent call to gp embedded will destroy it.

12
1.7 Readline interface.
Code which wants to use libpari readline (such as the Jupyter notebook) needs to do the
following:
#include <readline.h>
#include <paripriv.h>
pari_rl_interface S;
...
pari_use_readline(S);
The variable S, as initialized above, encapsulates the libpari readline interface. (And allow us to
move gp’s readline code to libpari without introducing a mandatory dependency on readline in
libpari.) The following functions then become available:
char** pari_completion_matches(pari_rl_interface *pS, const char *s, long pos, long
*wordpos) given a command string s, where the cursor is at index pos, return an array of completion
matches.
If wordpos is not NULL, set *wordpos to the index for the start of the expression we complete.
char** pari_completion(pari_rl_interface *pS, char *text, int start, int end) the
low-level completer called by pari_completion_matches. The following wrapper
char**
gp_completion(char *text, int START, int END)
{ return pari_completion(&S, text, START, END);)
is a valid value for rl_attempted_completion_function.

1.8 Constructors called by pari init functions.

void pari_init_buffers()
void pari_init_compiler()
void pari_init_defaults()
void pari_init_evaluator()
void pari_init_files()
void pari_init_floats()
void pari_init_graphics()
void pari_init_homedir()
void pari_init_parser()
void pari_init_paths()
void pari_init_primetab()
void pari_init_rand()
void pari_init_seadata()

13
1.9 Destructors called by pari close.
void pari_close_compiler()
void pari_close_evaluator()
void pari_close_files()
void pari_close_floats()
void pari_close_homedir()
void pari_close_mf()
void pari_close_parser()
void pari_close_paths()
void pari_close_primes()

1.10 Constructors and destructors used by the pthreads interface.

• Called by pari_thread_close
void pari_thread_close_files()

14
 Chapter 2:
Regression tests, benches

This chapter documents how to write an automated test module, say fun, so that make test-fun
executes the statements in the fun module and times them, compares the output to a template,
and prints an error message if they do not match.
• Pick a new name for your test, say fun, and write down a GP script named fun. Make sure
it produces some useful output and tests adequately a set of routines.
• The script should not be too long: one minute runs should be enough. Try to break your
script into independent easily reproducible tests, this way regressions are easier to debug; e.g.
include setrand(1) statement before a randomized computation. The expected output may be
different on 32-bit and 64-bit machines but should otherwise be platform-independent. If possible,
the output shouldn’t even depend on sizeof(long); using a realprecision that exists on both
32-bit and 64-bit architectures, e.g. \p 38 is a good first step. You can use sizebyte(0)==16 to
detect a 64-bit architecture and sizebyte(0)==8 for 32-bit.
• Dump your script into src/test/in/ and run Configure.
• make test-fun now runs the new test, producing a [BUG] error message and a .dif file in
the relevant object directory Oxxx. In fact, we compared the output to a nonexisting template, so
this must fail.
• Now
patch -p1 < Oxxx/fun.dif
generates a template output in the right place src/test/32/fun, for instance on a 32-bit machine.
• If different output is expected on 32-bit and 64-bit machines, run the test on a 64-bit machine
and patch again, thereby producing src/test/64/fun. If, on the contrary, the output must be
the same (preferred behavior!), make sure the output template land in the src/test/32/ directory
which provides a default template when the 64-bit output file is missing; in particular move the file
from src/test/64/ to src/test/32/ if the test was run on a 64-bit machine.
• You can now re-run the test to check for regressions: no [BUG] is expected this time! Of
course you can at any time add some checks, and iterate the test / patch phases. In particular,
each time a bug in the fun module is fixed, it is a good idea to add a minimal test case to the test
suite.
• By default, your new test is now included in make test-all. If it is particularly annoying,
e.g. opens tons of graphical windows as make test-ploth or just much longer than the recom-
mended minute, you may edit config/get tests and add the fun test to the list of excluded tests,
in the test extra out variable.
• You can run a subset of existing tests by using the following idiom:
cd Oxxx # call from relevant build directory
make TESTS="lfuntype lfun gamma" test-all
will run the lfuntype, lfun and gamma tests. This produces a combined output whereas the
alternative

15
 make test-lfuntype test-lfun test-gamma
would not.
• By default, the test is run on both the gp-sta and gp-dyn binaries, making it twice as slow.
If the test is somewhat long, it can be annoying; you can restrict to one binary only using the
statest-all or dyntest-all targets. Both accept the TESTS argument seen above.
make test-lfuntype test-lfun gamma
would not.
• Finally, the get tests script also defines the recipe for make bench timings, via the variable
test basic. A test is included as fun or fun n, where n is an integer ≤ 1000; the latter means
that the timing is weighted by a factor n/1000. (This was introduced a long time ago, when the
nfields bench was so much slower than the others that it hid slowdowns elsewhere.)

2.1 Functions for GP2C.

2.1.1 Functions for safe access to components.

These functions return the address of the requested component after checking it is actually
valid. This is used by GP2C -C.
GEN* safegel(GEN x, long l), safe version of gel(x,l) for t_VEC, t_COL and t_MAT.
long* safeel(GEN x, long l), safe version of x[l] for t_VECSMALL.
GEN* safelistel(GEN x, long l) safe access to t_LIST component.
GEN* safegcoeff(GEN x, long a, long b) safe version of gcoeff(x,a, b) for t_MAT.

16
 Chapter 3:
Parallelism

PARI provides an abstraction, herafter called the MT engine, for doing parallel computations. The
exact same high level routines are used whether the underlying communication protocol is POSIX
threads or MPI and they behave differently depending on how libpari was configured, specifically
on Configure’s --mt option. Sequential computation is also supported (no --mt argument) which
is helpful for debugging newly written parallel code. The final section in this chapter comments a
complete example.

3.1 The PARI multithread interface.

void mt_queue_start(struct pari_mt *pt, GEN worker) Let worker be a t_CLOSURE object of
arity 1. Initialize the opaque structure pt to evaluate worker in parallel, using nbthreads threads.
This allocates data in various ways, e.g., on the PARI stack or as malloc’ed objects: you may not
collect garbage on the PARI stack starting from an earlier avma point until the parallel computation
is over, it could destroy something in pt. All ressources allocated outside the PARI stack are freed
by mt queue end.
void mt_queue_start_lim(struct pari_mt *pt, GEN worker, long lim) as mt queue start,
where lim is an upper bound on the number of tasks to perform. Concretely the number of threads
is the minimum of lim and nbthreads. The values 0 and 1 of lim are special:
• 0: no limit, equivalent to mt queue start (use nbthreads threads).
• 1: no parallelism, evaluate the tasks sequentially.
void mt_queue_submit(struct pari_mt *pt, long taskid, GEN task) submit task to be
evaluated by worker; use task = NULL if no further task needs to be submitted. The parameter
taskid is attached to the task but not used in any way by the worker or the MT engine, it will be
returned to you by mt queue get together with the result for the task, allowing to match up results
and submitted tasks if desired. For instance, if the tasks (t1 , . . . , tm ) are known in advance, stored
in a vector, and you want to recover the evaluation results in the same order as in that vector, you
may use consecutive integers 1, . . . , m as taskids. If you do not care about the ordering, on the
other hand, you can just use taskid = 0 for all tasks.
The taskid parameter is ignored when task is NULL. It is forbidden to call this function twice
without an intervening mt queue get.
GEN mt_queue_get(struct pari_mt *pt, long *taskid, long *pending) return NULL until
mt queue submit has submitted tasks for the required number (nbthreads) of threads; then return
the result of the evaluation by worker of one of the previously submitted tasks, in random order. Set
pending to the number of remaining pending tasks: if this is 0 then no more tasks are pending and
it is safe to call mt_queue_end. Set *taskid to the value attached to this task by mt queue submit,
unless the taskid pointer is NULL. It is forbidden to call this function twice without an intervening
mt queue submit.
void mt_queue_end(struct pari_mt *pt) end the parallel execution and free ressources attached
to the opaque pari mt structure. For instance malloc’ed data; in the pthreads interface, it would

17
destroy mutex locks, condition variables, etc. This must be called once there are no longer pending
tasks to avoid leaking ressources; but not before all tasks have been processed else crashes will
occur.
long mt_nbthreads(void) return the effective number of parallel threads that would be started
by mt_queue_start if it has been called in place of mt_nbthreads.

3.2 Technical functions required by MPI.

The functions in this section are needed when writing complex independent programs in
order to support the MPI MT engine, as more flexible complement/variants of pari init and
pari close.
void mt_broadcast(GEN code): do nothing unless the MPI threading engine is in use. In that
case, evaluates the closure code on all secondary nodes. This can be used to change the state of
all MPI child nodes, e.g., in gpinstall run in the main thread, which allows all nodes to use the
new function.
void pari_mt_init(void) when using MPI, it is often necessary to run initialization code on the
child nodes after PARI is initialized. This is done by calling successively:
• pari_init_opts with the flag INIT_noIMTm: this initializes PARI, but not the MT engine;
• the required initialization code;
• pari_mt_init to initialize the MT engine. Note that under MPI, this function returns on
the master node but enters slave mode on the child nodes. Thus it is no longer possible to run
initialization code on the child nodes.
void pari_mt_close(void) when using MPI, calling pari_close terminates the MPI execution
environment and it will not be possible to restart it. If this is undesirable, call pari_close_opts
with the flag INIT_noIMTm instead of pari close: this closes PARI without terminating the MPI
execution environment. You may later call pari mt close to terminate it. It is an error for a
program to end without terminating the MPI execution environment.

3.3 A complete example.

We now proceed to an example exhibiting complex features of this interface, in particular
showing how to generate a valid worker. Explanations and details follow.
#include <pari/pari.h>
GEN
Cworker(GEN d, long kind) { return kind? det(d): Z_factor(d); }
int
main(void)
{
long i, taskid, pending;
GEN M,N1,N2, in,out, done;
struct pari_mt pt;
entree ep = {"_worker",0,(void*)Cworker,20,"GL",""};
/* initialize PARI, postponing parallelism initialization */

18
 pari_init_opts(8000000,500000, INIT_JMPm|INIT_SIGm|INIT_DFTm|INIT_noIMTm);
pari_add_function(&ep); /* add Cworker function to gp */
pari_mt_init(); /* ... THEN initialize parallelism */
/* Create inputs and room for output in main PARI stack */
N1 = addis(int2n(256), 1); /* 2^256 + 1 */
N2 = subis(int2n(193), 1); /* 2^193 - 1 */
M = mathilbert(80);
in = mkvec3(mkvec2(N1,gen_0), mkvec2(N2,gen_0), mkvec2(M,gen_1));
out = cgetg(4,t_VEC);
/* Initialize parallel evaluation of Cworker */
mt_queue_start(&pt, strtofunction("_worker"));
for (i = 1; i <= 3 || pending; i++)
{ /* submit job (in) and get result (out) */
mt_queue_submit(&pt, i, i<=3? gel(in,i): NULL);
done = mt_queue_get(&pt, &taskid, &pending);
if (done) gel(out,taskid) = done;
}
mt_queue_end(&pt); /* end parallelism */
output(out); pari_close(); return 0;
}
We start from some arbitrary C function Cworker and create an entree summarizing all that
GP would need to know about it, in particular
• a GP name worker; the leading is not necessary, we use it as a namespace mechanism
grouping private functions;
• the name of the C function;
• and its prototype, see install for an introduction to Prototype Codes.
The other three arguments (0, 20 and "") are required in an entree but not useful in our simple
context: they are respectively a valence (0 means “nothing special”), a help section (20 is customary
for internal functions which need to be exported for technical reasons, see ?20), and a help text
(no help).
Then we initialize the MT engine; doing things in this order with a two part initialization
ensures that nodes have access to our Cworker. We convert the ep data to a t_CLOSURE using
strtofunction, which provides a valid worker to mt queue start. This creates a parallel evalua-
tion queue mt, and we proceed to submit all tasks, recording all results. Results are stored in the
right order by making good use of the taskid label, although we have no control over when each
result is returned. We finally free all ressources attached to the mt structure. If needed, we could
have collected all garbage on the PARI stack using gerepilecopy on the out array and gone on
working instead of quitting.
Note the argument passing convention for Cworker: the task consists of a single vector con-
taining all arguments as GENs, which are interpreted according to the function prototype, here GL
so the first argument is left as is and the second one is converted to a long integer. In more com-
plicated situations, this second (and possibly further) argument could provide arbitrary evaluation
contexts. In this example, we just used it as a flag to indicate the kind of evaluation expected on
the data: integer factorization (0) or matrix determinant (1).
Note also that

19
 gel(out, taskid) = mt_queue_get(&mt, &taskid, &pending);
instead of our use of a temporary done would have undefined behaviour (taskid may be uninitial-
ized in the left hand side).

20
 Index

SomeWord refers to PARI-GP concepts. get_localbitprec . . . . . . . . . . . . . 11

SomeWord is a PARI-GP keyword. get_localprec . . . . . . . . . . . . . . . 11
SomeWord is a generic index entry. geval_gp . . . . . . . . . . . . . . . . . . 11
gpinstall . . . . . . . . . . . . . . . . . 18
A gp_embedded . . . . . . . . . . . . . . . . 12
apply0 . . . . . . . . . . . . . . . . . . . 10 gp_embedded_init . . . . . . . . . . . . . 12
gp_read_file . . . . . . . . . . . . . . . 10
B gunclone . . . . . . . . . . . . . . . . . . . 6
gunclone_deep . . . . . . . . . . . . . . . . 6
BLOCK_SIGALRM_START . . . . . . . . . . . . 8
BLOCK_SIGINT_END . . . . . . . . . . . . . . 8 I
BLOCK_SIGINT_START . . . . . . . . . . . . 8
break0 . . . . . . . . . . . . . . . . . . . 10 INIT_noIMTm . . . . . . . . . . . . . . . . 18

C L

closure . . . . . . . . . . . . . . . . . . . . . 5 list . . . . . . . . . . . . . . . . . . . . . . . 6
closure_arity . . . . . . . . . . . . . . . . 5 listcopy . . . . . . . . . . . . . . . . . . . 7
closure_codestr . . . . . . . . . . . . . . 5 listinit . . . . . . . . . . . . . . . . . . . 9
closure_get_code . . . . . . . . . . . . . . 5 listpop . . . . . . . . . . . . . . . . . . . . 7
closure_get_data . . . . . . . . . . . . . . 5 listput . . . . . . . . . . . . . . . . . . . . 7
closure_get_dbg . . . . . . . . . . . . . . 5 list_data . . . . . . . . . . . . . . . . . . 6
closure_get_frame . . . . . . . . . . . . . 6 list_nmax . . . . . . . . . . . . . . . . . . 6
closure_get_oper . . . . . . . . . . . . . . 5 list_typ . . . . . . . . . . . . . . . . . . . 7
closure_get_text . . . . . . . . . . . . . . 6 localbitprec . . . . . . . . . . . . . . . 11
localprec . . . . . . . . . . . . . . . . . 11
E localvars_find . . . . . . . . . . . . . . 11
localvars_read_str . . . . . . . . . . . 11
error0 . . . . . . . . . . . . . . . . . . . 10
M
F
mapdomain . . . . . . . . . . . . . . . . . . 7
Fl2_equal1 . . . . . . . . . . . . . . . . . . 8 mapdomain_shallow . . . . . . . . . . . . . 7
Fl2_inv_pre . . . . . . . . . . . . . . . . . 8 maptomat . . . . . . . . . . . . . . . . . . . 7
Fl2_mul_pre . . . . . . . . . . . . . . . . . 8 maptomat_shallow . . . . . . . . . . . . . . 8
Fl2_norm_pre . . . . . . . . . . . . . . . . 8 matslice0 . . . . . . . . . . . . . . . . . 10
Fl2_pow_pre . . . . . . . . . . . . . . . . . 8 mklist . . . . . . . . . . . . . . . . . . . . 7
Fl2_sqrtn_pre . . . . . . . . . . . . . . . . 8 mklistcopy . . . . . . . . . . . . . . . . . . 7
Fl2_sqr_pre . . . . . . . . . . . . . . . . . 8 mklist_typ . . . . . . . . . . . . . . . . . . 7
Flx_Fl2_eval_pre . . . . . . . . . . . . . . 9 mkmap . . . . . . . . . . . . . . . . . . . . . 7
f_PRETTYMAT . . . . . . . . . . . . . . . . . 9 mt_broadcast . . . . . . . . . . . . . . . 18
f_RAW . . . . . . . . . . . . . . . . . . . . . 9 mt_nbthreads . . . . . . . . . . . . . 17, 18
f_TEX . . . . . . . . . . . . . . . . . . . . . 9 mt_queue_end . . . . . . . . . . . . . . . 17
mt_queue_get . . . . . . . . . . . . . . . 17
G
mt_queue_start . . . . . . . . . . . . 17, 18
genapply . . . . . . . . . . . . . . . . . . 11 mt_queue_start_lim . . . . . . . . . . . 17
genselect . . . . . . . . . . . . . . . . . 11 mt_queue_submit . . . . . . . . . . . . . 17
GENtoTeXstr . . . . . . . . . . . . . . . . 10 MT_SIGINT_BLOCK . . . . . . . . . . . . . . 8
getenv . . . . . . . . . . . . . . . . . . . 12 mt_sigint_block . . . . . . . . . . . . . . 8
getlocalbitprec . . . . . . . . . . . . . 11 MT_SIGINT_UNBLOCK . . . . . . . . . . . . . 8
getlocalprec . . . . . . . . . . . . . . . 11 mt_sigint_unblock . . . . . . . . . . . . . 8

21
 N pari_thread_close_files . . . . . . . . 14
pop_localprec . . . . . . . . . . . . . . . 11
next0 . . . . . . . . . . . . . . . . . . . . 10 print . . . . . . . . . . . . . . . . . . . . 10
print0 . . . . . . . . . . . . . . . . . . 9, 10
O
print1 . . . . . . . . . . . . . . . . . . . 10
os_getenv . . . . . . . . . . . . . . . . . 12 printf0 . . . . . . . . . . . . . . . . . . . 10
os_signal . . . . . . . . . . . . . . . . . 12 printp . . . . . . . . . . . . . . . . . . . 10
output . . . . . . . . . . . . . . . . . . . 10 printsep . . . . . . . . . . . . . . . . . . 10
out_print0 . . . . . . . . . . . . . . . . . 10 printsep1 . . . . . . . . . . . . . . . . . 10
printtex . . . . . . . . . . . . . . . . . . 10
P push_localbitprec . . . . . . . . . . . . 11
push_localprec . . . . . . . . . . . . . . 11
pariOut . . . . . . . . . . . . . . . . . . . 10
pari_close . . . . . . . . . . . . . . . . . 18 R
pari_close_compiler . . . . . . . . . . . 13
pari_close_evaluator . . . . . . . . . . 13 return0 . . . . . . . . . . . . . . . . . . . 10
pari_close_files . . . . . . . . . . . . . 13 rfrac_to_ser . . . . . . . . . . . . . . . . 9
pari_close_floats . . . . . . . . . . . . 14 RgX_to_ser . . . . . . . . . . . . . . . . . . 9
pari_close_homedir . . . . . . . . . . . 14 rl_attempted_completion_function . . 13
pari_close_mf . . . . . . . . . . . . . . . 14
S
pari_close_opts . . . . . . . . . . . . . 18
pari_close_parser . . . . . . . . . . . . 14 safeel . . . . . . . . . . . . . . . . . . . 16
pari_close_paths . . . . . . . . . . . . . 14 safegcoeff . . . . . . . . . . . . . . . . . 16
pari_close_primes . . . . . . . . . . . . 14 safegel . . . . . . . . . . . . . . . . . . . 16
pari_completion . . . . . . . . . . . . . 12 safelistel . . . . . . . . . . . . . . . . . 16
pari_completion_matches . . . . . . 12, 13 SA_NODEFER . . . . . . . . . . . . . . . . . 12
pari_infile . . . . . . . . . . . . . . . . 10 select0 . . . . . . . . . . . . . . . . . . . 11
pari_init_buffers . . . . . . . . . . . . 13 sigaction . . . . . . . . . . . . . . . . . 12
pari_init_compiler . . . . . . . . . . . 13 signal . . . . . . . . . . . . . . . . . . . 12
pari_init_defaults . . . . . . . . . . . 13 SIG_IGN . . . . . . . . . . . . . . . . . . . 12
pari_init_evaluator . . . . . . . . . . . 13 strprintf . . . . . . . . . . . . . . . . . 10
pari_init_files . . . . . . . . . . . . . 13 switchin . . . . . . . . . . . . . . . . . . 10
pari_init_floats . . . . . . . . . . . . . 13
pari_init_graphics . . . . . . . . . . . 13 T
pari_init_homedir . . . . . . . . . . . . 13 toser_i . . . . . . . . . . . . . . . . . . . . 9
pari_init_opts . . . . . . . . . . . . . . 18 t_CLOSURE . . . . . . . . . . . . . . . . . . 5
pari_init_parser . . . . . . . . . . . . . 13 t_LIST . . . . . . . . . . . . . . . . . . . . 6
pari_init_paths . . . . . . . . . . . . . 13
pari_init_primetab . . . . . . . . . . . 13 V
pari_init_rand . . . . . . . . . . . . . . 13
pari_init_seadata . . . . . . . . . . . . 13 vecapply . . . . . . . . . . . . . . . . . . 11
pari_mt_close . . . . . . . . . . . . . . . 18 veccatapply . . . . . . . . . . . . . . . . 11
pari_mt_init . . . . . . . . . . . . . . . 18 vecselapply . . . . . . . . . . . . . . . . 11
pari_printf . . . . . . . . . . . . . . . . 10 vecselect . . . . . . . . . . . . . . . . . 11
PARI_SIGINT_block . . . . . . . . . . . . . 8 vecslice0 . . . . . . . . . . . . . . . . . 10
PARI_SIGINT_pending . . . . . . . . . . . . 8
W
pari_sprint0 . . . . . . . . . . . . . . . 10
pari_sprintf . . . . . . . . . . . . . . . 10 warning0 . . . . . . . . . . . . . . . . . . 10
pari_thread_close . . . . . . . . . . . . 14 write0 . . . . . . . . . . . . . . . . . . . 10

22
write1 . . . . . . . . . . . . . . . . . . . 10
writetex . . . . . . . . . . . . . . . . . . 10

23