xref: /llvm-project/polly/lib/External/isl/imath/doc.md.in (revision 658eb9e14264d48888ade0e3daf0b648f76c3f0e)
1# User Documentation for the IMath Library
2
3Author: [M. J. Fromberger](https://github.com/creachadair)
4
5## Installation
6
71. Edit Makefile to select compiler and options.  The default is to use gcc.
8   You may want to change CC to `clang` instead of `gcc` (and on macOS that
9   what you will get anyway), but you should be able to use the default GCC
10   settings for either.
11
12   By default, the Makefile assumes you can use 64-bit integer types, even
13   though they were not standard in ANSI C90. If you cannot, add
14   `-DUSE_32BIT_WORDS` to the compiler options.
15
162. Type `make` or `make test` to build the test driver and run the unit tests.
17   None of these should fail.  If they do, see below for how you can report
18   bugs.
19
20   To build with debugging enabled (and optimization disabled), run `make
21   DEBUG=Y`.  This sets the preprocessor macro `DEBUG` to 1, and several other
22   things (see Makefile for details).
23
24To use the library in your code, include "imath.h" wherever you intend to use
25the library's routines.  The integer library is just a single source file, so
26you can compile it into your project in whatever way makes sense.  If you wish
27to use rational arithmetic, you will also need to include "imrat.h".
28
29## Background
30
31The basic types defined by the imath library are `mpz_t`, an arbitrary
32precision signed integer, and `mpq_t`, an arbitrary precision signed rational
33number.  The type `mp_int` is a pointer to an `mpz_t`, and `mp_rat` is a
34pointer to an `mpq_t`.
35
36Most of the functions in the imath library return a value of type `mp_result`.
37This is a signed integer type which can be used to convey status information
38and also return small values.  Any negative value is considered to be a status
39message.  The following constants are defined for processing these:
40
41| Status      | Description                                  |
42| ----------- | -------------------------------------------- |
43| `MP_OK`     | operation successful, all is well (= 0)      |
44| `MP_FALSE`  | boolean false (= `MP_OK`)                    |
45| `MP_TRUE`   | boolean true                                 |
46| `MP_MEMORY` | out of memory                                |
47| `MP_RANGE`  | parameter out of range                       |
48| `MP_UNDEF`  | result is undefined (e.g., division by zero) |
49| `MP_TRUNC`  | output value was truncated                   |
50| `MP_BADARG` | an invalid parameter was passed              |
51
52If you obtain a zero or negative value of an `mp_result`, you can use the
53`mp_error_string()` routine to obtain a pointer to a brief human-readable
54string describing the error.  These strings are statically allocated, so they
55need not be freed by the caller; the same strings are re-used from call to
56call.
57
58Unless otherwise noted, it is legal to use the same parameter for both inputs
59and output with most of the functions in this library.  For example, you can
60add a number to itself and replace the original by writing:
61
62    mp_int_add(a, a, a);  /* a = a + a */
63
64Any cases in which this is not legal will be noted in the function summaries
65below (if you discover that this is not so, please report it as a bug; I will
66fix either the function or the documentation :)
67
68## The IMath API
69
70Each of the API functions is documented here.  The general format of the
71entries is:
72
73> ------------
74> <pre>
75> return_type function_name(parameters ...)
76> </pre>
77>  -  English description.
78
79Unless otherwise noted, any API function that returns `mp_result` may be
80expected to return `MP_OK`, `MP_BADARG`, or `MP_MEMORY`.  Other return values
81should be documented in the description.  Please let me know if you discover
82this is not the case.
83
84The following macros are defined in "imath.h", to define the sizes of the
85various data types used in the library:
86
87| Constant        | Description
88| --------------- | ----------------------------------------
89| `MP_DIGIT_BIT`  | the number of bits in a single `mpz_t` digit.
90| `MP_WORD_BIT`   | the number of bits in a `mpz_t` word.
91| `MP_SMALL_MIN`  | the minimum value representable by an `mp_small`.
92| `MP_SMALL_MAX`  | the maximum value representable by an `mp_small`.
93| `MP_USMALL_MAX` | the maximum value representable by an `mp_usmall`.
94| `MP_MIN_RADIX`  | the minimum radix accepted for base conversion.
95| `MP_MAX_RADIX`  | the maximum radix accepted for base conversion.
96
97#### Initialization
98
99An `mp_int` must be initialized before use. By default, an `mp_int` is
100initialized with a certain minimum amount of storage for digits, and the
101storage is expanded automatically as needed.  To initialize an `mp_int`, use
102the following functions:
103
104{{insert "imath.h"
105  mp_int_init mp_int_alloc mp_int_init_size
106  mp_int_init_copy
107  mp_int_init_value
108}}
109
110#### Cleanup
111
112When you are finished with an `mp_int`, you must free the memory it uses:
113
114{{insert "imath.h" mp_int_clear mp_int_free}}
115
116#### Setting Values
117
118To set an `mp_int` which has already been initialized to a small integer value,
119use:
120
121{{insert "imath.h" mp_int_set_value mp_int_set_uvalue}}
122
123To copy one initialized `mp_int` to another, use:
124
125{{insert "imath.h" mp_int_copy}}
126
127### Arithmetic Functions
128
129{{insert "imath.h"
130  mp_int_is_odd mp_int_is_even
131  mp_int_zero
132  mp_int_abs
133  mp_int_neg
134  mp_int_add mp_int_add_value
135  mp_int_sub mp_int_sub_value
136  mp_int_mul mp_int_mul_value mp_int_mul_pow2
137  mp_int_sqr
138  mp_int_root mp_int_sqrt
139  mp_int_div mp_int_div_value mp_int_div_pow2
140  mp_int_mod mp_int_mod_value
141  mp_int_expt mp_int_expt_value mp_int_expt_full
142}}
143
144### Comparison Functions
145
146Unless otherwise specified, comparison between values `x` and `y` returns a
147**comparator**, an integer value < 0 if `x` is less than `y`, 0 if `x` is equal
148to `y`, and > 0 if `x` is greater than `y`.
149
150{{insert "imath.h"
151  mp_int_compare mp_int_compare_unsigned mp_int_compare_zero
152  mp_int_compare_value mp_int_compare_uvalue
153  mp_int_divisible_value mp_int_is_pow2
154}}
155
156### Modular Operations
157
158{{insert "imath.h"
159  mp_int_exptmod mp_int_exptmod_evalue mp_int_exptmod_bvalue
160  mp_int_exptmod_known mp_int_redux_const
161  mp_int_invmod
162  mp_int_gcd mp_int_egcd mp_int_lcm
163}}
164
165### Conversion of Values
166
167{{insert "imath.h"
168  mp_int_to_int mp_int_to_uint
169  mp_int_to_string mp_int_string_len
170  mp_int_read_string mp_int_read_cstring
171  mp_int_count_bits
172  mp_int_to_binary mp_int_read_binary mp_int_binary_len
173  mp_int_to_unsigned mp_int_read_unsigned mp_int_unsigned_len
174}}
175
176### Other Functions
177
178Ordinarily, integer multiplication and squaring are done using the simple
179quadratic "schoolbook" algorithm.  However, for sufficiently large values,
180there is a more efficient algorithm usually attributed to Karatsuba and Ofman
181that is usually faster.  See Knuth Vol. 2 for more details about how this
182algorithm works.
183
184The breakpoint between the "normal" and the recursive algorithm is controlled
185by a static digit threshold defined in `imath.c`. Values with fewer significant
186digits use the standard algorithm.  This value can be modified by calling
187`mp_int_multiply_threshold(n)`.  The `imtimer` program and the
188`findthreshold.py` script (Python) can help you find a suitable value for for
189your particular platform.
190
191{{insert "imath.h" mp_error_string}}
192
193## Rational Arithmetic
194
195{{insert "imrat.h"}}
196
197## Representation Details
198
199> NOTE: You do not need to read this section to use IMath.  This is provided
200> for the benefit of developers wishing to extend or modify the internals of
201> the library.
202
203IMath uses a signed magnitude representation for arbitrary precision integers.
204The magnitude is represented as an array of radix-R digits in increasing order
205of significance; the value of R is chosen to be half the size of the largest
206available unsigned integer type, so typically 16 or 32 bits.  Digits are
207represented as mp_digit, which must be an unsigned integral type.
208
209Digit arrays are allocated using `malloc(3)` and `realloc(3)`.  Because this
210can be an expensive operation, the library takes pains to avoid allocation as
211much as possible.  For this reason, the `mpz_t` structure distinguishes between
212how many digits are allocated and how many digits are actually consumed by the
213representation.  The fields of an `mpz_t` are:
214
215    mp_digit    single;  /* single-digit value (see note) */
216    mp_digit   *digits;  /* array of digits               */
217    mp_size     alloc;   /* how many digits are allocated */
218    mp_size     used;    /* how many digits are in use    */
219    mp_sign     sign;    /* the sign of the value         */
220
221The elements of `digits` at indices less than `used` are the significant
222figures of the value; the elements at indices greater than or equal to `used`
223are undefined (and may contain garbage).  At all times, `used` must be at least
2241 and at most `alloc`.
225
226To avoid interaction with the memory allocator, single-digit values are stored
227directly in the `mpz_t` structure, in the `single` field.  The semantics of
228access are the same as the more general case.
229
230The number of digits allocated for an `mpz_t` is referred to in the library
231documentation as its "precision".  Operations that affect an `mpz_t` cause
232precision to increase as needed.  In any case, all allocations are measured in
233digits, and rounded up to the nearest `mp_word` boundary.  There is a default
234minimum precision stored as a static constant default_precision (`imath.c`).
235This value can be set using `mp_int_default_precision(n)`.
236
237Note that the allocated size of an `mpz_t` can only grow; the library never
238reallocates in order to decrease the size.  A simple way to do so explicitly is
239to use `mp_int_init_copy()`, as in:
240
241```
242mpz_t big, new;
243
244/* ... */
245mp_int_init_copy(&new, &big);
246mp_int_swap(&new, &big);
247mp_int_clear(&new);
248```
249
250The value of `sign` is 0 for positive values and zero, 1 for negative values.
251Constants `MP_ZPOS` and `MP_NEG` are defined for these; no other sign values
252are used.
253
254If you are adding to this library, you should be careful to preserve the
255convention that inputs and outputs can overlap, as described above.  So, for
256example, `mp_int_add(a, a, a)` is legal.  Often, this means you must maintain
257one or more temporary mpz_t structures for intermediate values.  The private
258macros `DECLARE_TEMP(N)`, `CLEANUP_TEMP()`, and `TEMP(K)` can be used to
259maintain a conventional structure like this:
260
261```c
262{
263  /* Declare how many temp values you need.
264	 Use TEMP(i) to access the ith value (0-indexed). */
265  DECLARE_TEMP(8);
266  ...
267
268  /* Perform actions that must return MP_OK or fail. */
269  REQUIRE(mp_int_copy(x, TEMP(1)));
270  ...
271  REQUIRE(mp_int_expt(TEMP(1), TEMP(2), TEMP(3)));
272  ...
273
274  /* You can also use REQUIRE directly for more complex cases. */
275  if (some_difficult_question(TEMP(3)) != answer(x)) {
276	REQUIRE(MP_RANGE);  /* falls through to cleanup (below) */
277  }
278
279  /* Ensure temporary values are cleaned up at exit.
280
281     If control reaches here via a REQUIRE failure, the code below
282	 the cleanup will not be executed.
283   */
284  CLEANUP_TEMP();
285  return MP_OK;
286}
287```
288
289Under the covers, these macros are just maintaining an array of `mpz_t` values,
290and a jump label to handle cleanup. You may only have one `DECLARE_TEMP` and
291its corresponding `CLEANUP_TEMP` per function body.
292
293"Small" integer values are represented by the types `mp_small` and `mp_usmall`,
294which are mapped to appropriately-sized types on the host system.  The default
295for `mp_small` is "long" and the default for `mp_usmall` is "unsigned long".
296You may change these, provided you insure that `mp_small` is signed and
297`mp_usmall` is unsigned.  You will also need to adjust the size macros:
298
299    MP_SMALL_MIN, MP_SMALL_MAX
300    MP_USMALL_MIN, MP_USMALL_MAX
301
302... which are defined in `<imath.h>`, if you change these.
303
304Rational numbers are represented using a pair of arbitrary precision integers,
305with the convention that the sign of the numerator is the sign of the rational
306value, and that the result of any rational operation is always represented in
307lowest terms.  The canonical representation for rational zero is 0/1.  See
308"imrat.h".
309
310## Testing and Reporting of Bugs
311
312Test vectors are included in the `tests/` subdirectory of the imath
313distribution.  When you run `make test`, it builds the `imtest` program and
314runs all available test vectors.  If any tests fail, you will get a line like
315this:
316
317    x    y    FAILED      v
318
319Here, _x_ is the line number of the test which failed, _y_ is index of the test
320within the file, and _v_ is the value(s) actually computed.  The name of the
321file is printed at the beginning of each test, so you can find out what test
322vector failed by executing the following (with x, y, and v replaced by the
323above values, and where "foo.t" is the name of the test file that was being
324processed at the time):
325
326    % tail +x tests/foo.t | head -1
327
328None of the tests should fail (but see [Note 2](#note2)); if any do, it
329probably indicates a bug in the library (or at the very least, some assumption
330I made which I shouldn't have).  Please [file an
331issue](https://github.com/creachadair/imath/issues/new), including the `FAILED`
332test line(s), as well as the output of the above `tail` command (so I know what
333inputs caused the failure).
334
335If you build with the preprocessor symbol `DEBUG` defined as a positive
336integer, the digit allocators (`s_alloc`, `s_realloc`) fill all new buffers
337with the value `0xdeadbeefabad1dea`, or as much of it as will fit in a digit,
338so that you can more easily catch uninitialized reads in the debugger.
339
340## Notes
341
3421. <a name="note1"></a>You can generally use the same variables for both input
343   and output.  One exception is that you may not use the same variable for
344   both the quotient and the remainder of `mp_int_div()`.
345
3462. <a name="note2"></a>Many of the tests for this library were written under
347   the assumption that the `mp_small` type is 32 bits or more.  If you compile
348   with a smaller type, you may see `MP_RANGE` errors in some of the tests that
349   otherwise pass (due to conversion failures).  Also, the pi generator (pi.c)
350   will not work correctly if `mp_small` is too short, as its algorithm for arc
351   tangent is fairly simple-minded.
352
353## Contacts
354
355The IMath library was written by Michael J. Fromberger.
356
357If you discover any bugs or testing failures, please [open an
358issue](https://github.com/creachadair/imath/issues/new).  Please be sure to
359include a complete description of what went wrong, and if possible, a test
360vector for `imtest` and/or a minimal test program that will demonstrate the bug
361on your system.  Please also let me know what hardware, operating system, and
362compiler you're using.
363
364## Acknowledgements
365
366The algorithms used in this library came from Vol. 2 of Donald Knuth's "The Art
367of Computer Programming" (Seminumerical Algorithms).  Thanks to Nelson Bolyard,
368Bryan Olson, Tom St. Denis, Tushar Udeshi, and Eric Silva for excellent
369feedback on earlier versions of this code.  Special thanks to Jonathan Shapiro
370for some very helpful design advice, as well as feedback and some clever ideas
371for improving performance in some common use cases.
372
373## License and Disclaimers
374
375IMath is Copyright 2002-2009 Michael J. Fromberger
376You may use it subject to the following Licensing Terms:
377
378Permission is hereby granted, free of charge, to any person obtaining a copy of
379this software and associated documentation files (the "Software"), to deal in
380the Software without restriction, including without limitation the rights to
381use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
382of the Software, and to permit persons to whom the Software is furnished to do
383so, subject to the following conditions:
384
385The above copyright notice and this permission notice shall be included in all
386copies or substantial portions of the Software.
387
388THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
389IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
390FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
391AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
392LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
393OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
394SOFTWARE.
395