dist/doc/FAQ.html

<?xml version="1.0" encoding="iso-8859-1"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">

<!--
Copyright 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Free Software Foundation, Inc.
Contributed by the AriC and Caramel projects, INRIA.

This file is part of the GNU MPFR Library.

The GNU MPFR Library is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 3 of the License, or (at your
option) any later version.

The GNU MPFR 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 Lesser General Public
License for more details.

You should have received a copy of the GNU Lesser General Public License
along with the GNU MPFR Library; see the file COPYING.LESSER.  If not, see
http://www.gnu.org/licenses/ or write to the Free Software Foundation, Inc.,
51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
-->

<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">

<head><meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1" />
<title>Frequently Asked Questions about GNU MPFR</title>
<style type="text/css">/*<![CDATA[*/
/* Global stylesheet for visual media */

html, body
{
  background: white;
  color: black;
}

div.logo { float: right }
div.logo img { border: 0 }

div.footer img { border: 0 }

dt
{
  margin-top: 2ex;
  margin-bottom: 1ex;
  font-weight: bolder;
}

/* For testing: dd { background: #ddddff } */

table { margin: 0.5ex auto }

li { margin-top: 0.5ex; margin-bottom: 0.5ex }

dd + dd
{
  margin-top: 0;
  margin-bottom: 0;
  padding-top: 0.5ex;
  padding-bottom: 0;
}

li > p, dd > p
{
  margin-top: 0;
  margin-bottom: 0;
  padding-top: 0.5ex;
  padding-bottom: 0.5ex;
}

.block-code code, code.block-code,
.block-code samp, samp.block-code
{
  display: block;
  padding: 0.5ex 0;
  margin-left: 2em;
}

.nowrap { white-space: nowrap }

/*
dl.faq { counter-reset: faq }

dl.faq dt:before
{
  counter-increment: faq;
  content: counter(faq) ". ";
}
*/

dl.faq dt { background: #dddddd }

dl.faq dd
{
  border-left: 4px solid;
  border-color: transparent;
  margin-left: 0em;
  padding-left: 1.5em;
}

dl.faq dt:target + dd
{
  border-left-color: #aaaaaa;
}

var.env { font-style: normal }
/*]]>*/</style>
</head>

<body>

<h1>Frequently Asked Questions about <cite><acronym>GNU</acronym> <acronym>MPFR</acronym></cite></h1>

<p><strong>Important notice: Problems with a particular version of
<cite><acronym>MPFR</acronym></cite> are discussed in the corresponding
bugs page.</strong></p>

<p>The latest version of this <acronym>FAQ</acronym> is available at
<a href="http://www.mpfr.org/faq.html">http://www.mpfr.org/faq.html</a>.
Please look at this version if possible.</p>

<ol>
<li><a href="#mpfr_vs_mpf">What are the differences between
<cite><acronym>MPF</acronym></cite> from <cite><acronym>GMP</acronym></cite>
and <cite><acronym>MPFR</acronym></cite>?</a></li>
<li><a href="#mpf2mpfr">How to convert my program written using
<cite><acronym>MPF</acronym></cite> to
<cite><acronym>MPFR</acronym></cite>?</a></li>
<li><a href="#no_libgmp">At configure time, I get the error: <q>libgmp not found or uses a different ABI.</q></a></li>
<li><a href="#undef_ref1">I get undefined reference to <code>__gmp_get_memory_functions</code>.</a></li>
<li><a href="#undef_ref2">When I link my program with
<cite><acronym>MPFR</acronym></cite>, I get undefined reference
to <code>__gmpXXXX</code>.</a></li>
<li><a href="#crash_high_prec">My program crashes with high precisions.</a></li>
<li><a href="#accuracy">Though I have increased the precision, the results
are not more accurate.</a></li>
<li><a href="#detect_mpfr">How can I detect <cite><acronym>MPFR</acronym></cite>
installation using <cite>autoconf</cite> or <cite>pkg-config</cite>?</a></li>
<li><a href="#cite">How to cite <cite><acronym>MPFR</acronym></cite> in a
scientific publication?</a></li>
<li><a href="#fpic">When I build <cite><acronym>MPFR</acronym></cite>, I get
an error asking me to recompile with <samp>-fPIC</samp>.</a></li>
</ol>

<dl class="faq">

<dt id="mpfr_vs_mpf">1. What are the differences between
<cite><acronym>MPF</acronym></cite> from <cite><acronym>GMP</acronym></cite>
and <cite><acronym>MPFR</acronym></cite>?</dt>

<dd><p>The main differences are:</p>
<ul>
<li><p>The precision of a <cite><acronym>MPFR</acronym></cite> variable
is the <em>exact</em> number of bits used for its mantissa, whereas in
<cite><acronym>MPF</acronym></cite>, the precision requested by the user
is a minimum value (<cite><acronym>MPF</acronym></cite> generally uses a
higher precision). With the additional difference below, this implies that
the <cite><acronym>MPFR</acronym></cite> results do not depend on the
number of bits (16, 32, 64 or more) of the underlying architecture.</p></li>
<li><p>As a consequence, <cite><acronym>MPFR</acronym></cite> uses a
base-2 exponent, whereas in <cite><acronym>MPF</acronym></cite>, this
is a base-2<sup>32</sup> or base-2<sup>64</sup> exponent, depending on
the limb size. For this reason (and other internal ones), the maximum
exponent range in <cite><acronym>MPFR</acronym></cite> is different
(and smaller, if the exponent is represented by the same type as in
<cite><acronym>MPF</acronym></cite>).</p></li>
<li><p><cite><acronym>MPFR</acronym></cite> provides an additional rounding
mode argument to its functions; furthermore, it is guaranteed that the
result of any operation is the nearest possible floating-point value from
the exact result (considering the input variables as exact values), taking
into account the precision of the destination variable and the rounding
mode. <cite><acronym>MPFR</acronym></cite> also says whether the rounded
result is above or below the exact result.</p></li>
<li><p><cite><acronym>MPFR</acronym></cite> supports much more functions
(in particular transcendental functions such as exponentials, logarithms,
trigonometric functions and so on) and special values: signed zeros,
infinities, not-a-number (NaN).</p></li>
</ul></dd>

<dt id="mpf2mpfr">2. How to convert my program written using
<cite><acronym>MPF</acronym></cite> to
<cite><acronym>MPFR</acronym></cite>?</dt>

<dd><p>You need to add <q><code>r</code></q> to the function names, and to
specify the rounding mode (<code>MPFR_RNDN</code> for rounding to nearest,
<code>MPFR_RNDZ</code> for rounding toward zero, <code>MPFR_RNDU</code>
for rounding toward plus infinity, <code>MPFR_RNDD</code> for rounding
toward minus infinity). You can also define macros as follows:
<code class="block-code">#define mpf_add(a, b, c) mpfr_add(a, b, c, MPFR_RNDN)</code></p>
<p>The header file <samp>mpf2mpfr.h</samp> from the
<cite><acronym>MPFR</acronym></cite> distribution automatically
redefines all <cite><acronym>MPF</acronym></cite> functions in this
way, using the default <cite><acronym>MPFR</acronym></cite> rounding
mode. Thus you simply need to add the following line in all your files
using <cite><acronym>MPF</acronym></cite> functions:
<code class="block-code">#include &lt;mpf2mpfr.h&gt;</code>
just after the <samp>gmp.h</samp> and <samp>mpfr.h</samp>
header files. If the program uses <cite><acronym>MPF</acronym></cite>
internals (such as direct access to <code>__mpf_struct</code> members),
additional changes will be needed.</p></dd>

<dt id="no_libgmp">3. At configure time, I get the error: <q>libgmp not found or uses a different ABI.</q></dt>

<dd><p>This test (<samp>checking for __gmpz_init in -lgmp</samp>) comes
after the <samp>gmp.h</samp> detection. The failure occurs either because
the <cite><acronym>GMP</acronym></cite> library could not be found
(as it is not in the provided library search paths) or because the
<cite><acronym>GMP</acronym></cite> library that was found does not have
the expected <acronym title="Application Binary Interface">ABI</acronym>
(<abbr>e.g.</abbr> 32-bit <abbr>vs</abbr> 64-bit). The former problem can be
due to the fact that a static build of <cite><acronym>MPFR</acronym></cite>
was requested while only a shared <cite><acronym>GMP</acronym></cite> library
is installed (or the opposite, but another error can also show up in this
case, see the <a href="#fpic">question about <samp>-fPIC</samp></a>). The
latter problem can have several causes:</p>
<ul>
<li>A wrong libgmp library has been picked up. This can occur if you have
several <cite><acronym>GMP</acronym></cite> versions installed on the
machine and something is wrong with the provided library search paths.</li>
<li>Wrong compiler options (<samp>CFLAGS</samp>) were given. In general, the
presence or absence of the <samp>-m64</samp> compiler option must match the
library <acronym title="Application Binary Interface">ABI</acronym>.</li>
<li>A wrong <samp>gmp.h</samp> file has been picked up (if you have several
<cite><acronym>GMP</acronym></cite> versions installed). Indeed, by default,
<cite><acronym>MPFR</acronym></cite> gets the compiler options from the
<samp>gmp.h</samp> file (with <cite><acronym>GMP</acronym></cite> 4.2.3
or later); this is needed because <cite><acronym>GMP</acronym></cite> does
not necessarily use the default <acronym>ABI</acronym>. The consequence is
that if the <samp>gmp.h</samp> file is associated with a library using a
different <acronym>ABI</acronym>, the <acronym>ABI</acronym>-related options
will be incorrect. Hence the failure.</li>
</ul>
<p>Note: The <samp>config.log</samp> output gives more information
than the error message. In particular, see the output of the test:
<samp>checking for CC and CFLAGS in gmp.h</samp>; it should give you
the default compiler options (from <samp>gmp.h</samp>).</p>

<p>See also the answer to the <a href="#undef_ref1">next question</a>.</p></dd>

<dt id="undef_ref1">4. I get undefined reference to <code>__gmp_get_memory_functions</code>.</dt>

<dd><p>Note: this was mainly a problem when upgrading from
<cite><acronym>GMP</acronym></cite> 4.1.4 to a later version,
but information given below may still be useful in other cases,
when several <cite><acronym>GMP</acronym></cite> libraries are
installed on the same machine.</p>

<p>If you get such an error, in particular when running
<samp>make check</samp>, then this probably means that you are using
the header file from <cite><acronym>GMP</acronym></cite> 4.2.x but the
<cite><acronym>GMP</acronym></cite> 4.1.4 library. This can happen if
several <cite><acronym>GMP</acronym></cite> versions are installed on
your machine (<abbr>e.g.</abbr>, one provided by the system in
<samp>/usr/{include,lib}</samp> and a new one installed by the owner or
administrator of the machine in <samp>/usr/local/{include,lib}</samp>)
and your include and library search paths are inconsistent. On various
<acronym>GNU</acronym>/Linux machines, this is unfortunately the case
by default (<samp>/usr/local/include</samp> is in the default include
search path, but <samp>/usr/local/lib</samp> is <em>not</em> in the
default library search path). Typical errors are:
<samp class="block-code">undefined reference to `__gmp_get_memory_functions'</samp>
in <samp>make check</samp>. The best solution is to add
<samp>/usr/local/include</samp> to your <var class="env">C_INCLUDE_PATH</var>
environment variable and to add <samp>/usr/local/lib</samp> to your
<var class="env">LIBRARY_PATH</var> and <var class="env">LD_LIBRARY_PATH</var>
environment variables (and/or <var class="env">LD_RUN_PATH</var>).
Alternatively, you can use <samp>--with-gmp*</samp> configure options,
<abbr>e.g.</abbr> <samp>--with-gmp=/usr/local</samp>, but <strong>this is
not guaranteed to work</strong> (in particular with <samp>gcc</samp> and
system directories such as <samp>/usr</samp> or <samp>/usr/local</samp>),
and other software that uses <cite><acronym>GMP</acronym></cite> and/or
<cite><acronym>MPFR</acronym></cite> will need correct paths too;
environment variables allow you to set them in a global way.</p>
<p>Other information can be given in the <samp>INSTALL</samp> file and
<samp>ld</samp> manual. Please look at them for more details. See also
the <a href="#undef_ref2">next question</a>.</p></dd>

<dt id="undef_ref2">5. When I link my program with
<cite><acronym>MPFR</acronym></cite>, I get undefined reference
to <code>__gmpXXXX</code>.</dt>

<dd><p>Link your program with <cite><acronym>GMP</acronym></cite>. Assuming
that your program is <samp>foo.c</samp>, you should link it using:
<samp class="block-code">cc link.c -lmpfr -lgmp</samp>
<cite><acronym>MPFR</acronym></cite> library reference (<samp>-lmpfr</samp>)
should be before <cite><acronym>GMP</acronym></cite>'s one
(<samp>-lgmp</samp>). Another solution is, with <acronym>GNU</acronym>
<samp>ld</samp>, to give all the libraries inside a group:
<samp class="block-code">gcc link.c -Wl,--start-group libgmp.a libmpfr.a -Wl,--end-group</samp>
See <samp>INSTALL</samp> file and <samp>ld</samp> manual for more
details.</p>
<p>If you used correct link options, but still get an error, this may mean
that your include and library search paths are inconsistent. Please see the
<a href="#undef_ref1">previous question</a>.</p></dd>

<dt id="crash_high_prec">6. My program crashes with high precisions.</dt>

<dd><p>Your stack size limit may be too small; indeed, by default,
<cite><acronym>GMP</acronym></cite> 4.1.4 and below allocates all
temporary results on the stack, and in very high precisions, this
limit may be reached. You can solve this problem in different ways:</p>
<ul>
<li><p>You can upgrade to <cite><acronym>GMP</acronym></cite> 4.2 (or above),
which now makes temporary allocations on the stack only when they are
small.</p></li>
<li><p>You can increase the stack size limit with the <samp>limit</samp>,
<samp>unlimit</samp> or <samp>ulimit</samp> command, depending on your
shell. This may fail on some systems, where the maximum stack size cannot
be increased above some value.</p></li>
<li><p>You can rebuild both <cite><acronym>GMP</acronym></cite> and
<cite><acronym>MPFR</acronym></cite> to use another allocation method.</p></li>
</ul></dd>

<dt id="accuracy">7. Though I have increased the precision, the results
are not more accurate.</dt>

<dd><p>The reason may be the use of C floating-point numbers. If you want
to store a floating-point constant to a <code>mpfr_t</code>, you should use
<code>mpfr_set_str</code> (or one of the <cite><acronym>MPFR</acronym></cite>
constant functions, such as <code>mpfr_const_pi</code> for &#960;) instead
of <code>mpfr_set_d</code> or <code>mpfr_set_ld</code>. Otherwise the
floating-point constant will be first converted into a reduced-precision
(<abbr>e.g.</abbr>, 53-bit) binary number before
<cite><acronym>MPFR</acronym></cite> can work with it. This is the case
in particular for most exact decimal numbers, such as 0.17, which are
not exactly representable in binary.</p>
<p>Also remember that <cite><acronym>MPFR</acronym></cite> does not track
the accuracy of the results: copying a value <var>x</var> to <var>y</var>
with <code>mpfr_set (y, x, MPFR_RNDN)</code> where the variable <var>y</var>
is more precise than the variable <var>x</var> will not make it more
accurate; the (binary) value will remain unchanged.</p></dd>

<dt id="detect_mpfr">8. How can I detect <cite><acronym>MPFR</acronym></cite>
installation using <cite>autoconf</cite> or <cite>pkg-config</cite>?</dt>

<dd><p>The <cite><acronym>MPFR</acronym></cite> team does not currently
recommend any <cite>autoconf</cite> code, but a section will later
be added to the <cite><acronym>MPFR</acronym></cite> manual. The
<cite><acronym>MPFR</acronym></cite> team does not wish to support
<cite>pkg-config</cite> yet.</p></dd>

<dt id="cite">9. How to cite <cite><acronym>MPFR</acronym></cite> in a
scientific publication?</dt>

<dd><p>To properly cite <cite><acronym>MPFR</acronym></cite> in a scientific
publication, please cite the
<a href="http://doi.acm.org/10.1145/1236463.1236468"><acronym title="Association for Computing Machinery">ACM</acronym>
<acronym title="Transactions on Mathematical Software">TOMS</acronym>
paper</a>
(<a href="http://toms.acm.org/cgi/TOMSbibget.cgi?Fousse:2007:MMP">BibTeX</a>)
and/or the library web page
<a href="http://www.mpfr.org/">http://www.mpfr.org</a>. If your publication
is related to a particular release of <cite><acronym>MPFR</acronym></cite>,
for example if you report timings, please also indicate the release number
for future reference.</p></dd>

<dt id="fpic">10. When I build <cite><acronym>MPFR</acronym></cite>, I get
an error asking me to recompile with <samp>-fPIC</samp>.</dt>

<dd><p>A typical error looks like:</p>
<p><tt>/usr/bin/ld: <em>/path/to/</em>libgmp.a(realloc.o): relocation
R_X86_64_32 against `.rodata.str1.1' can not be used when making a
shared object; recompile with -fPIC<br />
<em>/path/to/</em>libgmp.a: could not read symbols: Bad value<br />
collect2: ld returned 1 exit status</tt></p>
<p>The probable reason is that you tried to build
<cite><acronym>MPFR</acronym></cite> with the shared library enabled (this
is the default), while only a static <cite><acronym>GMP</acronym></cite>
library could be found. To solve this problem, either rebuild and reinstall
<cite><acronym>GMP</acronym></cite> without the <samp>--disable-shared</samp>
configure option, or configure <cite><acronym>MPFR</acronym></cite> with
<samp>--disable-shared</samp>. If you did this and still get the above
error, the cause may be conflicting <cite><acronym>GMP</acronym></cite>
versions installed on your system; please check that your search path
settings are correct.</p>
<p>Additional note about the last sentence: Under <acronym>GNU</acronym>/Linux
(for instance), the linker takes the first library found in the library search
path, whether it is dynamic or static. The default behavior under darwin is
different, but <cite><acronym>MPFR</acronym></cite> will change it.</p></dd>
<!-- Reference concerning darwin: see MPFR_LD_SEARCH_PATHS_FIRST
     in MPFR's configure.{ac,in} and acinclude.m4 -->

</dl>

</body>

</html>