diff options
Diffstat (limited to 'devdocs/c/numeric%2Fmath%2Ffma.html')
| -rw-r--r-- | devdocs/c/numeric%2Fmath%2Ffma.html | 94 |
1 files changed, 94 insertions, 0 deletions
diff --git a/devdocs/c/numeric%2Fmath%2Ffma.html b/devdocs/c/numeric%2Fmath%2Ffma.html new file mode 100644 index 00000000..ce600c3f --- /dev/null +++ b/devdocs/c/numeric%2Fmath%2Ffma.html @@ -0,0 +1,94 @@ + <h1 id="firstHeading" class="firstHeading">fma, fmaf, fmal</h1> <table class="t-dcl-begin"> <tr class="t-dsc-header"> <th> Defined in header <code><math.h></code> </th> <th> </th> <th> </th> </tr> <tr class="t-dcl t-since-c99"> <td> <pre data-language="c">float fmaf( float x, float y, float z );</pre> +</td> <td> (1) </td> <td> <span class="t-mark-rev t-since-c99">(since C99)</span> </td> </tr> <tr class="t-dcl t-since-c99"> <td> <pre data-language="c">double fma( double x, double y, double z );</pre> +</td> <td> (2) </td> <td> <span class="t-mark-rev t-since-c99">(since C99)</span> </td> </tr> <tr class="t-dcl t-since-c99"> <td> <pre data-language="c">long double fmal( long double x, long double y, long double z );</pre> +</td> <td> (3) </td> <td> <span class="t-mark-rev t-since-c99">(since C99)</span> </td> </tr> <tr class="t-dcl t-since-c99"> <td> <pre data-language="c">#define FP_FAST_FMA /* implementation-defined */</pre> +</td> <td> (4) </td> <td> <span class="t-mark-rev t-since-c99">(since C99)</span> </td> </tr> <tr class="t-dcl t-since-c99"> <td> <pre data-language="c">#define FP_FAST_FMAF /* implementation-defined */</pre> +</td> <td> (5) </td> <td> <span class="t-mark-rev t-since-c99">(since C99)</span> </td> </tr> <tr class="t-dcl t-since-c99"> <td> <pre data-language="c">#define FP_FAST_FMAL /* implementation-defined */</pre> +</td> <td> (6) </td> <td> <span class="t-mark-rev t-since-c99">(since C99)</span> </td> </tr> <tr class="t-dsc-header"> <th> Defined in header <code><tgmath.h></code> </th> <th> </th> <th> </th> </tr> <tr class="t-dcl t-since-c99"> <td> <pre data-language="c">#define fma( x, y, z )</pre> +</td> <td> (7) </td> <td> <span class="t-mark-rev t-since-c99">(since C99)</span> </td> </tr> </table> <div class="t-li1"> +<span class="t-li">1-3)</span> Computes <code>(x*y) + z</code> as if to infinite precision and rounded only once to fit the result type.</div> <div class="t-li1"> +<span class="t-li">4-6)</span> If the macro constants <code>FP_FAST_FMA</code>, <code>FP_FAST_FMAF</code>, or <code>FP_FAST_FMAL</code> are defined, the corresponding function <code>fmaf</code>, <code>fma</code>, or <code>fmal</code> evaluates faster (in addition to being more precise) than the expression <code>x*y+z</code> for <code>float</code>, <code>double</code>, and <code>long double</code> arguments, respectively. If defined, these macros evaluate to integer <code>1</code>.</div> <div class="t-li1"> +<span class="t-li">7)</span> Type-generic macro: If any argument has type <code>long double</code>, <code>fmal</code> is called. Otherwise, if any argument has integer type or has type <code>double</code>, <code>fma</code> is called. Otherwise, <code>fmaf</code> is called.</div> <h3 id="Parameters"> Parameters</h3> <table class="t-par-begin"> <tr class="t-par"> <td> x, y, z </td> <td> - </td> <td> floating point values </td> +</tr> +</table> <h3 id="Return_value"> Return value</h3> <p>If successful, returns the value of <code>(x*y) + z</code> as if calculated to infinite precision and rounded once to fit the result type (or, alternatively, calculated as a single ternary floating-point operation).</p> +<p>If a range error due to overflow occurs, <code><a href="huge_val" title="c/numeric/math/HUGE VAL">±HUGE_VAL</a></code>, <code>±HUGE_VALF</code>, or <code>±HUGE_VALL</code> is returned.</p> +<p>If a range error due to underflow occurs, the correct value (after rounding) is returned.</p> +<h3 id="Error_handling"> Error handling</h3> <p>Errors are reported as specified in <a href="math_errhandling" title="c/numeric/math/math errhandling"><code>math_errhandling</code></a>.</p> +<p>If the implementation supports IEEE floating-point arithmetic (IEC 60559),</p> +<ul> +<li> If x is zero and y is infinite or if x is infinite and y is zero, and z is not a NaN, then NaN is returned and <code><a href="../fenv/fe_exceptions" title="c/numeric/fenv/FE exceptions">FE_INVALID</a></code> is raised </li> +<li> If x is zero and y is infinite or if x is infinite and y is zero, and z is a NaN, then NaN is returned and <code><a href="../fenv/fe_exceptions" title="c/numeric/fenv/FE exceptions">FE_INVALID</a></code> may be raised </li> +<li> If <code>x*y</code> is an exact infinity and z is an infinity with the opposite sign, NaN is returned and <code><a href="../fenv/fe_exceptions" title="c/numeric/fenv/FE exceptions">FE_INVALID</a></code> is raised </li> +<li> If x or y are NaN, NaN is returned </li> +<li> If z is NaN, and <code>x*y</code> aren't 0*Inf or Inf*0, then NaN is returned (without <code><a href="../fenv/fe_exceptions" title="c/numeric/fenv/FE exceptions">FE_INVALID</a></code>). </li> +</ul> <h3 id="Notes"> Notes</h3> <p>This operation is commonly implemented in hardware as <a href="https://en.wikipedia.org/wiki/Multiply%E2%80%93accumulate_operation" class="extiw" title="enwiki:Multiply–accumulate operation">fused multiply-add</a> CPU instruction. If supported by hardware, the appropriate <code>FP_FAST_FMA*</code> macros are expected to be defined, but many implementations make use of the CPU instruction even when the macros are not defined.</p> +<p><a rel="nofollow" class="external text" href="http://pubs.opengroup.org/onlinepubs/9699919799/functions/fma.html">POSIX specifies</a> that the situation where the value <code>x*y</code> is invalid and z is a NaN is a domain error.</p> +<p>Due to its infinite intermediate precision, <code>fma</code> is a common building block of other correctly-rounded mathematical operations, such as <code><a href="sqrt" title="c/numeric/math/sqrt">sqrt</a></code> or even the division (where not provided by the CPU, e.g. Itanium).</p> +<p>As with all floating-point expressions, the expression <code>(x*y) + z</code> may be compiled as a fused mutiply-add unless the <a href="../../preprocessor/impl" title="c/preprocessor/impl"><code> #pragma</code></a> <code>STDC FP_CONTRACT</code> is off.</p> +<h3 id="Example"> Example</h3> <div class="t-example"> <div class="c source-c"><pre data-language="c">#include <stdio.h> +#include <math.h> +#include <float.h> +#include <fenv.h> +#pragma STDC FENV_ACCESS ON +int main(void) +{ + // demo the difference between fma and built-in operators + double in = 0.1; + printf("0.1 double is %.23f (%a)\n", in, in); + printf("0.1*10 is 1.0000000000000000555112 (0x8.0000000000002p-3)," + " or 1.0 if rounded to double\n"); + double expr_result = 0.1 * 10 - 1; + printf("0.1 * 10 - 1 = %g : 1 subtracted after " + "intermediate rounding to 1.0\n", expr_result); + double fma_result = fma(0.1, 10, -1); + printf("fma(0.1, 10, -1) = %g (%a)\n", fma_result, fma_result); + + // fma use in double-double arithmetic + printf("\nin double-double arithmetic, 0.1 * 10 is representable as "); + double high = 0.1 * 10; + double low = fma(0.1, 10, -high); + printf("%g + %g\n\n", high, low); + + //error handling + feclearexcept(FE_ALL_EXCEPT); + printf("fma(+Inf, 10, -Inf) = %f\n", fma(INFINITY, 10, -INFINITY)); + if(fetestexcept(FE_INVALID)) puts(" FE_INVALID raised"); +}</pre></div> <p>Possible output:</p> +<div class="text source-text"><pre data-language="c">0.1 double is 0.10000000000000000555112 (0x1.999999999999ap-4) +0.1*10 is 1.0000000000000000555112 (0x8.0000000000002p-3), or 1.0 if rounded to double +0.1 * 10 - 1 = 0 : 1 subtracted after intermediate rounding to 1.0 +fma(0.1, 10, -1) = 5.55112e-17 (0x1p-54) + +in double-double arithmetic, 0.1 * 10 is representable as 1 + 5.55112e-17 + +fma(+Inf, 10, -Inf) = -nan + FE_INVALID raised</pre></div> </div> <h3 id="References"> References</h3> <ul> +<li> C17 standard (ISO/IEC 9899:2018): </li> +<ul> +<li> 7.12.13.1 The fma functions (p: 188-189) </li> +<li> 7.25 Type-generic math <tgmath.h> (p: 272-273) </li> +<li> F.10.10.1 The fma functions (p: 386) </li> +</ul> +<li> C11 standard (ISO/IEC 9899:2011): </li> +<ul> +<li> 7.12.13.1 The fma functions (p: 258) </li> +<li> 7.25 Type-generic math <tgmath.h> (p: 373-375) </li> +<li> F.10.10.1 The fma functions (p: 530) </li> +</ul> +<li> C99 standard (ISO/IEC 9899:1999): </li> +<ul> +<li> 7.12.13.1 The fma functions (p: 239) </li> +<li> 7.22 Type-generic math <tgmath.h> (p: 335-337) </li> +<li> F.9.10.1 The fma functions (p: 466) </li> +</ul> +</ul> <h3 id="See_also"> See also</h3> <table class="t-dsc-begin"> <tr class="t-dsc"> <td> <div><a href="remainder" title="c/numeric/math/remainder"> <span class="t-lines"><span>remainder</span><span>remainderf</span><span>remainderl</span></span></a></div> +<div><span class="t-lines"><span><span class="t-mark-rev t-since-c99">(C99)</span></span><span><span class="t-mark-rev t-since-c99">(C99)</span></span><span><span class="t-mark-rev t-since-c99">(C99)</span></span></span></div> </td> <td> computes signed remainder of the floating-point division operation <br> <span class="t-mark">(function)</span> </td> +</tr> <tr class="t-dsc"> <td> <div><a href="remquo" title="c/numeric/math/remquo"> <span class="t-lines"><span>remquo</span><span>remquof</span><span>remquol</span></span></a></div> +<div><span class="t-lines"><span><span class="t-mark-rev t-since-c99">(C99)</span></span><span><span class="t-mark-rev t-since-c99">(C99)</span></span><span><span class="t-mark-rev t-since-c99">(C99)</span></span></span></div> </td> <td> computes signed remainder as well as the three last bits of the division operation <br> <span class="t-mark">(function)</span> </td> +</tr> <tr class="t-dsc"> <td colspan="2"> <span><a href="https://en.cppreference.com/w/cpp/numeric/math/fma" title="cpp/numeric/math/fma">C++ documentation</a></span> for <code>fma</code> </td> +</tr> </table> <div class="_attribution"> + <p class="_attribution-p"> + © cppreference.com<br>Licensed under the Creative Commons Attribution-ShareAlike Unported License v3.0.<br> + <a href="https://en.cppreference.com/w/c/numeric/math/fma" class="_attribution-link">https://en.cppreference.com/w/c/numeric/math/fma</a> + </p> +</div> |