| | <?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"> |
| |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| |
|
| | <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"> |
| | |
| | |
| | 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; |
| | } |
| | |
| | |
| | |
| | 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; |
| | } |
| | |
| | .highlight { background: #dddddd } |
| | |
| | .nowrap { white-space: nowrap } |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | 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="https://www.mpfr.org/faq.html">https://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 positive infinity, <code>MPFR_RNDD</code> for rounding |
| | toward negative 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 <mpf2mpfr.h></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 in any order inside a group: |
| | <samp class="block-code">gcc link.c -Wl,--start-group libgmp.a libmpfr.a -Wl,--end-group</samp> |
| | with the full path to the libraries, or |
| | <samp class="block-code">gcc link.c -Wl,--start-group -l:libgmp.a -l:libmpfr.a -Wl,--end-group</samp> |
| | to use the library path. |
| | 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 π) 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. |
| | Limited <cite>pkg-config</cite> support has been added for |
| | <cite><acronym>MPFR</acronym></cite>�4.0.0; example:</p> |
| | <pre style="margin-left: 2em">cc myprogram.c $(pkg-config --cflags --libs mpfr)</pre></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="https://doi.org/10.1145/1236463.1236468"><acronym title="Association for Computing Machinery">ACM</acronym> |
| | <acronym title="Transactions on Mathematical Software">TOMS</acronym> |
| | paper</a> |
| | and/or the library web page |
| | <a href="https://www.mpfr.org/">https://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> |
| | |
| | |
| |
|
| | </dl> |
| |
|
| | </body> |
| |
|
| | </html> |
| |
|
