Menu
Home
People
Places
Arts
History
Plants & Animals
Science
Life & Culture
Technology
Reference.org
Single-precision floating-point format
open-in-new
<h2 id="ieee-754-standard-binary32">IEEE 754 standard: binary32</h2> <p>The IEEE 754 standard specifies a <i>binary32</i> as having: </p> <ul><li><a href="/facts/Sign_bit/hLvapzSs">Sign bit</a>: 1 bit</li> <li><a href="/facts/Exponent/pac6QY2g">Exponent</a> width: 8 bits</li> <li><a href="/facts/Significand/JFCVyTz4">Significand</a> <a href="/facts/Precision_(arithmetic)/RUfYtpgo">precision</a>: 24 bits (23 explicitly stored)</li></ul> <p>This gives from 6 to 9 <a href="/facts/Significant_figures/RUfYtpgo">significant decimal digits</a> precision. If a decimal string with at most 6 significant digits is converted to the IEEE 754 single-precision format, giving a <a href="/facts/Normal_number_(computing)/5pvlHVvg">normal number</a>, and then converted back to a decimal string with the same number of digits, the final result should match the original string. If an IEEE 754 single-precision number is converted to a decimal string with at least 9 significant digits, and then converted back to single-precision representation, the final result must match the original number.<a class="footnote-ref" id="fnref:6" href="#fn:6"><sup>6</sup></a> </p><p>The sign bit determines the sign of the number, which is the sign of the significand as well. "1" stands for negative. The exponent field is an 8-bit unsigned integer from 0 to 255, in <a href="/facts/Exponent_bias/u5MIKgKP">biased form</a>: a value of 127 represents the actual exponent zero. Exponents range from −126 to +127 (thus 1 to 254 in the exponent field), because the biased exponent values 0 (all 0s) and 255 (all 1s) are reserved for special numbers (<a href="/facts/Subnormal_number/6qB24xrS">subnormal numbers</a>, <a href="/facts/Signed_zero/uX4d8zzV">signed zeros</a>, <a href="/facts/Infinity/cF506uD7">infinities</a>, and <a href="/facts/NaN/DNnYJ4FU">NaNs</a>). </p><p>The true significand of normal numbers includes 23 fraction bits to the right of the binary point and an <i><a href="/facts/Implicit_leading_bit/eIckahxe">implicit leading bit</a></i> (to the left of the binary point) with value 1. Subnormal numbers and zeros (which are the floating-point numbers smaller in magnitude than the least positive normal number) are represented with the biased exponent value 0, giving the implicit leading bit the value 0. Thus only 23 fraction bits of the <a href="/facts/Significand/JFCVyTz4">significand</a> appear in the memory format, but the total precision is 24 bits (equivalent to log10(224) ≈ 7.225 decimal digits) for normal values; subnormals have gracefully degrading precision down to 1 bit for the smallest non-zero value. </p><p>The bits are laid out as follows: </p><p>The real value assumed by a given 32-bit <i>binary32</i> data with a given <i>sign</i>, biased exponent <i>e</i> (the 8-bit unsigned integer), and a <i>23-bit fraction</i> is </p> ( − 1 ) b 31 × 2 ( b 30 b 29 … b 23 ) 2 − 127 × ( 1. b 22 b 21 … b 0 ) 2 {\displaystyle (-1)^{b_{31}}\times 2^{(b_{30}b_{29}\dots b_{23})_{2}-127}\times (1.b_{22}b_{21}\dots b_{0})_{2}} , <p>which yields </p> value = ( − 1 ) sign × 2 ( E − 127 ) × ( 1 + ∑ i = 1 23 b 23 − i 2 − i ) . {\displaystyle {\text{value}}=(-1)^{\text{sign}}\times 2^{(E-127)}\times \left(1+\sum _{i=1}^{23}b_{23-i}2^{-i}\right).} <p>In this example: </p> <ul><li> sign = b 31 = 0 {\displaystyle {\text{sign}}=b_{31}=0} ,</li> <li> ( − 1 ) sign = ( − 1 ) 0 = + 1 ∈ { − 1 , + 1 } {\displaystyle (-1)^{\text{sign}}=(-1)^{0}=+1\in \{-1,+1\}} ,</li> <li> E = ( b 30 b 29 … b 23 ) 2 = ∑ i = 0 7 b 23 + i 2 + i = 124 ∈ { 1 , … , ( 2 8 − 1 ) − 1 } = { 1 , … , 254 } {\displaystyle E=(b_{30}b_{29}\dots b_{23})_{2}=\sum _{i=0}^{7}b_{23+i}2^{+i}=124\in \{1,\ldots ,(2^{8}-1)-1\}=\{1,\ldots ,254\}} ,</li> <li> 2 ( E − 127 ) = 2 124 − 127 = 2 − 3 ∈ { 2 − 126 , … , 2 127 } {\displaystyle 2^{(E-127)}=2^{124-127}=2^{-3}\in \{2^{-126},\ldots ,2^{127}\}} ,</li> <li> 1. b 22 b 21 . . . b 0 = 1 + ∑ i = 1 23 b 23 − i 2 − i = 1 + 1 ⋅ 2 − 2 = 1.25 ∈ { 1 , 1 + 2 − 23 , … , 2 − 2 − 23 } ⊂ [ 1 ; 2 − 2 − 23 ] ⊂ [ 1 ; 2 ) {\displaystyle 1.b_{22}b_{21}...b_{0}=1+\sum _{i=1}^{23}b_{23-i}2^{-i}=1+1\cdot 2^{-2}=1.25\in \{1,1+2^{-23},\ldots ,2-2^{-23}\}\subset [1;2-2^{-23}]\subset [1;2)} .</li></ul> <p>thus: </p> <ul><li> value = ( + 1 ) × 2 − 3 × 1.25 = + 0.15625 {\displaystyle {\text{value}}=(+1)\times 2^{-3}\times 1.25=+0.15625} .</li></ul> <p>Note: </p> <ul><li> 1 + 2 − 23 ≈ 1.000 000 119 {\displaystyle 1+2^{-23}\approx 1.000\,000\,119} ,</li> <li> 2 − 2 − 23 ≈ 1.999 999 881 {\displaystyle 2-2^{-23}\approx 1.999\,999\,881} ,</li> <li> 2 − 126 ≈ 1.175 494 35 × 10 − 38 {\displaystyle 2^{-126}\approx 1.175\,494\,35\times 10^{-38}} ,</li> <li> 2 + 127 ≈ 1.701 411 83 × 10 + 38 {\displaystyle 2^{+127}\approx 1.701\,411\,83\times 10^{+38}} .</li></ul> <h3>Exponent encoding</h3> <p>The single-precision binary floating-point exponent is encoded using an <a href="/facts/Offset_binary/4B81wxxe">offset-binary</a> representation, with the zero offset being 127; also known as exponent bias in the IEEE 754 standard. </p> <ul><li>Emin = 01H−7FH = −126</li> <li>Emax = FEH−7FH = 127</li> <li><a href="/facts/Exponent_bias/u5MIKgKP">Exponent bias</a> = 7FH = 127</li></ul> <p>Thus, in order to get the true exponent as defined by the offset-binary representation, the offset of 127 has to be subtracted from the stored exponent. </p><p>The stored exponents 00H and FFH are interpreted specially. </p> <table><tbody><tr><th>Exponent</th><th>fraction = 0</th><th>fraction ≠ 0</th><th>Equation</th></tr><tr><td>00H = 000000002</td><td><a href="/facts/Signed_zero/uX4d8zzV">±zero</a></td><td><a href="/facts/Denormal_numbers/6qB24xrS">subnormal number</a></td><td> ( − 1 ) sign × 2 − 126 × 0. fraction {\displaystyle (-1)^{\text{sign}}\times 2^{-126}\times 0.{\text{fraction}}} </td></tr><tr><td>01H, ..., FEH = 000000012, ..., 111111102</td><td colspan="2">normal value</td><td> ( − 1 ) sign × 2 exponent − 127 × 1. fraction {\displaystyle (-1)^{\text{sign}}\times 2^{{\text{exponent}}-127}\times 1.{\text{fraction}}} </td></tr><tr><td>FFH = 111111112</td><td>±<a href="/facts/Infinity/cF506uD7">infinity</a></td><td><a href="/facts/NaN/DNnYJ4FU">NaN</a> (quiet, signalling)</td><td></td></tr></tbody></table> <p>The minimum positive normal value is 2 − 126 ≈ 1.18 × 10 − 38 {\displaystyle 2^{-126}\approx 1.18\times 10^{-38}} and the minimum positive (subnormal) value is 2 − 149 ≈ 1.4 × 10 − 45 {\displaystyle 2^{-149}\approx 1.4\times 10^{-45}} . </p> <h3>Converting decimal to binary32</h3> <p>In general, refer to the IEEE 754 standard itself for the strict conversion (including the rounding behaviour) of a real number into its equivalent binary32 format. </p><p>Here we can show how to convert a base-10 real number into an IEEE 754 binary32 format using the following outline: </p> <ul><li>Consider a real number with an integer and a fraction part such as 12.375</li> <li>Convert and <a href="/facts/Normalized_number/auQAhM4D">normalize</a> the integer part into <a href="/facts/Binary_numeral_system/a6JDSRPs">binary</a></li> <li>Convert the fraction part using the following technique as shown here</li> <li>Add the two results and adjust them to produce a proper final conversion</li></ul> <p>Conversion of the fractional part: Consider 0.375, the fractional part of 12.375. To convert it into a binary fraction, multiply the fraction by 2, take the integer part and repeat with the new fraction by 2 until a fraction of zero is found or until the precision limit is reached which is 23 fraction digits for IEEE 754 binary32 format. </p> 0.375 × 2 = 0.750 = 0 + 0.750 ⇒ b − 1 = 0 {\displaystyle 0.375\times 2=0.750=0+0.750\Rightarrow b_{-1}=0} , the integer part represents the binary fraction digit. Re-multiply 0.750 by 2 to proceed 0.750 × 2 = 1.500 = 1 + 0.500 ⇒ b − 2 = 1 {\displaystyle 0.750\times 2=1.500=1+0.500\Rightarrow b_{-2}=1} 0.500 × 2 = 1.000 = 1 + 0.000 ⇒ b − 3 = 1 {\displaystyle 0.500\times 2=1.000=1+0.000\Rightarrow b_{-3}=1} , fraction = 0.011, terminate <p>We see that ( 0.375 ) 10 {\displaystyle (0.375)_{10}} can be exactly represented in binary as ( 0.011 ) 2 {\displaystyle (0.011)_{2}} . Not all decimal fractions can be represented in a finite digit binary fraction. For example, decimal 0.1 cannot be represented in binary exactly, only approximated. Therefore: </p> ( 12.375 ) 10 = ( 12 ) 10 + ( 0.375 ) 10 = ( 1100 ) 2 + ( 0.011 ) 2 = ( 1100.011 ) 2 {\displaystyle (12.375)_{10}=(12)_{10}+(0.375)_{10}=(1100)_{2}+(0.011)_{2}=(1100.011)_{2}} <p>Since IEEE 754 binary32 format requires real values to be represented in ( 1. x 1 x 2 . . . x 23 ) 2 × 2 e {\displaystyle (1.x_{1}x_{2}...x_{23})_{2}\times 2^{e}} format (see <a href="/facts/Normalized_number/auQAhM4D">Normalized number</a>, <a href="/facts/Denormalized_number/6qB24xrS">Denormalized number</a>), 1100.011 is shifted to the right by 3 digits to become ( 1.100011 ) 2 × 2 3 {\displaystyle (1.100011)_{2}\times 2^{3}} </p><p>Finally we can see that: ( 12.375 ) 10 = ( 1.100011 ) 2 × 2 3 {\displaystyle (12.375)_{10}=(1.100011)_{2}\times 2^{3}} </p><p>From which we deduce: </p> <ul><li>The exponent is 3 (and in the biased form it is therefore ( 127 + 3 ) 10 = ( 130 ) 10 = ( 1000 0010 ) 2 {\displaystyle (127+3)_{10}=(130)_{10}=(1000\ 0010)_{2}} )</li> <li>The fraction is 100011 (looking to the right of the binary point)</li></ul> <p>From these we can form the resulting 32-bit IEEE 754 binary32 format representation of 12.375: </p> ( 12.375 ) 10 = ( 0 10000010 10001100000000000000000 ) 2 = ( 41460000 ) 16 {\displaystyle (12.375)_{10}=(0\ 10000010\ 10001100000000000000000)_{2}=(41460000)_{16}} <p>Note: consider converting 68.123 into IEEE 754 binary32 format: Using the above procedure you expect to get ( 42883EF9 ) 16 {\displaystyle ({\text{42883EF9}})_{16}} with the last 4 bits being 1001. However, due to the default rounding behaviour of IEEE 754 format, what you get is ( 42883EFA ) 16 {\displaystyle ({\text{42883EFA}})_{16}} , whose last 4 bits are 1010. </p><p>Example 1: Consider decimal 1. We can see that: ( 1 ) 10 = ( 1.0 ) 2 × 2 0 {\displaystyle (1)_{10}=(1.0)_{2}\times 2^{0}} </p><p>From which we deduce: </p> <ul><li>The exponent is 0 (and in the biased form it is therefore ( 127 + 0 ) 10 = ( 127 ) 10 = ( 0111 1111 ) 2 {\displaystyle (127+0)_{10}=(127)_{10}=(0111\ 1111)_{2}} </li> <li>The fraction is 0 (looking to the right of the binary point in 1.0 is all 0 = 000...0 {\displaystyle 0=000...0} )</li></ul> <p>From these we can form the resulting 32-bit IEEE 754 binary32 format representation of real number 1: </p> ( 1 ) 10 = ( 0 01111111 00000000000000000000000 ) 2 = ( 3F800000 ) 16 {\displaystyle (1)_{10}=(0\ 01111111\ 00000000000000000000000)_{2}=({\text{3F800000}})_{16}} <p>Example 2: Consider a value 0.25. We can see that: ( 0.25 ) 10 = ( 1.0 ) 2 × 2 − 2 {\displaystyle (0.25)_{10}=(1.0)_{2}\times 2^{-2}} </p><p>From which we deduce: </p> <ul><li>The exponent is −2 (and in the biased form it is ( 127 + ( − 2 ) ) 10 = ( 125 ) 10 = ( 0111 1101 ) 2 {\displaystyle (127+(-2))_{10}=(125)_{10}=(0111\ 1101)_{2}} )</li> <li>The fraction is 0 (looking to the right of binary point in 1.0 is all zeroes)</li></ul> <p>From these we can form the resulting 32-bit IEEE 754 binary32 format representation of real number 0.25: </p> ( 0.25 ) 10 = ( 0 01111101 00000000000000000000000 ) 2 = ( 3E800000 ) 16 {\displaystyle (0.25)_{10}=(0\ 01111101\ 00000000000000000000000)_{2}=({\text{3E800000}})_{16}} <p>Example 3: Consider a value of 0.375. We saw that 0.375 = ( 0.011 ) 2 = ( 1.1 ) 2 × 2 − 2 {\displaystyle 0.375={(0.011)_{2}}={(1.1)_{2}}\times 2^{-2}} </p><p>Hence after determining a representation of 0.375 as ( 1.1 ) 2 × 2 − 2 {\displaystyle {(1.1)_{2}}\times 2^{-2}} we can proceed as above: </p> <ul><li>The exponent is −2 (and in the biased form it is ( 127 + ( − 2 ) ) 10 = ( 125 ) 10 = ( 0111 1101 ) 2 {\displaystyle (127+(-2))_{10}=(125)_{10}=(0111\ 1101)_{2}} )</li> <li>The fraction is 1 (looking to the right of binary point in 1.1 is a single 1 = x 1 {\displaystyle 1=x_{1}} )</li></ul> <p>From these we can form the resulting 32-bit IEEE 754 binary32 format representation of real number 0.375: </p> ( 0.375 ) 10 = ( 0 01111101 10000000000000000000000 ) 2 = ( 3EC00000 ) 16 {\displaystyle (0.375)_{10}=(0\ 01111101\ 10000000000000000000000)_{2}=({\text{3EC00000}})_{16}} <h3>Converting binary32 to decimal</h3> <p>If the binary32 value, 41C80000 in this example, is in hexadecimal we first convert it to binary: </p> 41C8 0000 16 = 0100 0001 1100 1000 0000 0000 0000 0000 2 {\displaystyle {\text{41C8 0000}}_{16}=0100\ 0001\ 1100\ 1000\ 0000\ 0000\ 0000\ 0000_{2}} <p>then we break it down into three parts: sign bit, exponent, and significand. </p> <ul><li>Sign bit: 0 2 {\displaystyle 0_{2}} </li> <li>Exponent: 1000 0011 2 = 83 16 = 131 10 {\displaystyle 1000\ 0011_{2}=83_{16}=131_{10}} </li> <li>Significand: 100 1000 0000 0000 0000 0000 2 = 480000 16 {\displaystyle 100\ 1000\ 0000\ 0000\ 0000\ 0000_{2}=480000_{16}} </li></ul> <p>We then add the implicit 24th bit to the significand: </p> <ul><li>Significand: 1 100 1000 0000 0000 0000 0000 2 = C80000 16 {\displaystyle \mathbf {1} 100\ 1000\ 0000\ 0000\ 0000\ 0000_{2}={\text{C80000}}_{16}} </li></ul> <p>and decode the exponent value by subtracting 127: </p> <ul><li>Raw exponent: 83 16 = 131 10 {\displaystyle 83_{16}=131_{10}} </li> <li>Decoded exponent: 131 − 127 = 4 {\displaystyle 131-127=4} </li></ul> <p>Each of the 24 bits of the significand (including the implicit 24th bit), bit 23 to bit 0, represents a value, starting at 1 and halves for each bit, as follows: </p> bit 23 = 1 bit 22 = 0.5 bit 21 = 0.25 bit 20 = 0.125 bit 19 = 0.0625 bit 18 = 0.03125 bit 17 = 0.015625 . . bit 6 = 0.00000762939453125 bit 5 = 0.000003814697265625 bit 4 = 0.0000019073486328125 bit 3 = 0.00000095367431640625 bit 2 = 0.000000476837158203125 bit 1 = 0.0000002384185791015625 bit 0 = 0.00000011920928955078125 <p>The significand in this example has three bits set: bit 23, bit 22, and bit 19. We can now decode the significand by adding the values represented by these bits. </p> <ul><li>Decoded significand: 1 + 0.5 + 0.0625 = 1.5625 = C80000 / 2 23 {\displaystyle 1+0.5+0.0625=1.5625={\text{C80000}}/2^{23}} </li></ul> <p>Then we need to multiply with the base, 2, to the power of the exponent, to get the final result: </p> 1.5625 × 2 4 = 25 {\displaystyle 1.5625\times 2^{4}=25} <p>Thus </p> 41C8 0000 = 25 {\displaystyle {\text{41C8 0000}}=25} <p>This is equivalent to: </p> n = ( − 1 ) s × ( 1 + m ∗ 2 − 23 ) × 2 x − 127 {\displaystyle n=(-1)^{s}\times (1+m*2^{-23})\times 2^{x-127}} <p>where s is the sign bit, x is the exponent, and m is the significand. </p> <h3>Precision limitations on decimal values (between 1 and 16777216)</h3> <ul><li>Decimals between 1 and 2: fixed interval 2−23 (1+2−23 is the next largest float after 1)</li> <li>Decimals between 2 and 4: fixed interval 2−22</li> <li>Decimals between 4 and 8: fixed interval 2−21</li> <li>...</li> <li>Decimals between 2n and 2n+1: fixed interval 2n-23</li> <li>...</li> <li>Decimals between 222=4194304 and 223=8388608: fixed interval 2−1=0.5</li> <li>Decimals between 223=8388608 and 224=16777216: fixed interval 20=1</li></ul> <h3>Precision limitations on integer values</h3> <ul><li>Integers between 0 and 16777216 can be exactly represented (also applies for negative integers between −16777216 and 0)</li> <li>Integers between 224=16777216 and 225=33554432 round to a multiple of 2 (even number)</li> <li>Integers between 225 and 226 round to a multiple of 4</li> <li>...</li> <li>Integers between 2n and 2n+1 round to a multiple of 2n-23</li> <li>...</li> <li>Integers between 2127 and 2128 round to a multiple of 2104</li> <li>Integers greater than or equal to 2128 are rounded to "infinity".</li></ul> <h3>Notable single-precision cases</h3> <p>These examples are given in bit <i>representation</i>, in <a href="/facts/Hexadecimal/02ohQW22">hexadecimal</a> and <a href="/facts/Binary_number/a6JDSRPs">binary</a>, of the floating-point value. This includes the sign, (biased) exponent, and significand. </p> 0 00000000 000000000000000000000012 = 0000 000116 = 2−126 × 2−23 = 2−149 ≈ 1.4012984643 × 10−45 (smallest positive subnormal number) 0 00000000 111111111111111111111112 = 007f ffff16 = 2−126 × (1 − 2−23) ≈ 1.1754942107 ×10−38 (largest subnormal number) 0 00000001 000000000000000000000002 = 0080 000016 = 2−126 ≈ 1.1754943508 × 10−38 (smallest positive normal number) 0 11111110 111111111111111111111112 = 7f7f ffff16 = 2127 × (2 − 2−23) ≈ 3.4028234664 × 1038 (largest normal number) 0 01111110 111111111111111111111112 = 3f7f ffff16 = 1 − 2−24 ≈ 0.999999940395355225 (largest number less than one) 0 01111111 000000000000000000000002 = 3f80 000016 = 1 (one) 0 01111111 000000000000000000000012 = 3f80 000116 = 1 + 2−23 ≈ 1.00000011920928955 (smallest number larger than one) 1 10000000 000000000000000000000002 = c000 000016 = −2 0 00000000 000000000000000000000002 = 0000 000016 = 0 1 00000000 000000000000000000000002 = 8000 000016 = −0 0 11111111 000000000000000000000002 = 7f80 000016 = infinity 1 11111111 000000000000000000000002 = ff80 000016 = −infinity 0 10000000 100100100001111110110112 = 4049 0fdb16 ≈ 3.14159274101257324 ≈ π ( pi ) 0 01111101 010101010101010101010112 = 3eaa aaab16 ≈ 0.333333343267440796 ≈ 1/3 x 11111111 100000000000000000000012 = ffc0 000116 = qNaN (on x86 and ARM processors) x 11111111 000000000000000000000012 = ff80 000116 = sNaN (on x86 and ARM processors) <p>By default, 1/3 rounds up, instead of down like <a href="/facts/Double_precision/JYyXXYFM">double precision</a>, because of the even number of bits in the significand. The bits of 1/3 beyond the rounding point are 1010... which is more than 1/2 of a <a href="/facts/Unit_in_the_last_place/L2fqbaDb">unit in the last place</a>. </p><p>Encodings of qNaN and sNaN are not specified in <a href="/facts/IEEE_floating_point/dD3e7zAl">IEEE 754</a> and implemented differently on different processors. The <a href="/facts/X86/Kypq5fgu">x86</a> family and the <a href="/facts/ARM_architecture/SALsYA79">ARM</a> family processors use the most significant bit of the significand field to indicate a <a href="/facts/NaN/DNnYJ4FU">quiet NaN</a>. The <a href="/facts/PA-RISC/gjDs1HEh">PA-RISC</a> processors use the bit to indicate a <a href="/facts/NaN/DNnYJ4FU">signaling NaN</a>. </p> <h3>Optimizations</h3> <p>The design of floating-point format allows various optimisations, resulting from the easy generation of a <a href="/facts/Base-2_logarithm/Uy7kAtRm">base-2 logarithm</a> approximation from an integer view of the raw bit pattern. Integer arithmetic and bit-shifting can yield an approximation to <a href="/facts/Reciprocal_square_root/dRG2BtBp">reciprocal square root</a> (<a href="/facts/Fast_inverse_square_root/W3DUIX4u">fast inverse square root</a>), commonly required in <a href="/facts/Computer_graphics/lapBE8wv">computer graphics</a>. </p> <h2 id="see-also">See also</h2> <ul><li><a href="/facts/ISO/IEC_10967/lajN77uw">ISO/IEC 10967</a>, language independent arithmetic</li> <li><a href="/facts/Primitive_data_type/0c898228">Primitive data type</a></li> <li><a href="/facts/Numerical_stability/jRDdo8zV">Numerical stability</a></li> <li><a href="/facts/Scientific_notation/YgyYEmq8">Scientific notation</a></li></ul> <h2 id="external-links">External links</h2> <ul><li><a href="https://evanw.github.io/float-toy/">Live floating-point bit pattern editor</a></li> <li><a href="http://www.h-schmidt.net/FloatConverter/IEEE754.html">Online calculator</a></li> <li><a href="http://www.binaryconvert.com/convert_float.html">Online converter for IEEE 754 numbers with single precision</a></li> <li><a href="https://web.archive.org/web/20091031135212/http://www.mathworks.com/matlabcentral/fileexchange/23173">C source code to convert between IEEE double, single, and half precision</a></li></ul> <h2 id="references">References</h2> <ol> <li id="fn:1"><p>"REAL Statement". scc.ustc.edu.cn. Archived from the original on 2021-02-24. Retrieved 2013-02-28. <a href="https://web.archive.org/web/20210224045812/http://scc.ustc.edu.cn/zlsc/sugon/intel/compiler_f/main_for/lref_for/source_files/rfreals.htm" target="_blank">https://web.archive.org/web/20210224045812/http://scc.ustc.edu.cn/zlsc/sugon/intel/compiler_f/main_for/lref_for/source_files/rfreals.htm</a> <a href="#fnref:1" class="footnote-back-ref">↩</a></p></li> <li id="fn:2"><p>"CLHS: Type SHORT-FLOAT, SINGLE-FLOAT, DOUBLE-FLOAT..." www.lispworks.com. <a href="https://www.lispworks.com/documentation/HyperSpec/Body/t_short_.htm" target="_blank">https://www.lispworks.com/documentation/HyperSpec/Body/t_short_.htm</a> <a href="#fnref:2" class="footnote-back-ref">↩</a></p></li> <li id="fn:3"><p>"Primitive Data Types". Java Documentation. <a href="https://docs.oracle.com/javase/tutorial/java/nutsandbolts/datatypes.html" target="_blank">https://docs.oracle.com/javase/tutorial/java/nutsandbolts/datatypes.html</a> <a href="#fnref:3" class="footnote-back-ref">↩</a></p></li> <li id="fn:4"><p>"6 Predefined Types and Classes". haskell.org. 20 July 2010. <a href="https://www.haskell.org/onlinereport/haskell2010/haskellch6.html#x13-1350006.4" target="_blank">https://www.haskell.org/onlinereport/haskell2010/haskellch6.html#x13-1350006.4</a> <a href="#fnref:4" class="footnote-back-ref">↩</a></p></li> <li id="fn:5"><p>"Float". Apple Developer Documentation. <a href="https://developer.apple.com/documentation/swift/float" target="_blank">https://developer.apple.com/documentation/swift/float</a> <a href="#fnref:5" class="footnote-back-ref">↩</a></p></li> <li id="fn:6"><p>William Kahan (1 October 1997). "Lecture Notes on the Status of IEEE Standard 754 for Binary Floating-Point Arithmetic" (PDF). p. 4. Archived from the original (PDF) on 8 February 2012. <a href="https://web.archive.org/web/20120208075518/http://www.cs.berkeley.edu/~wkahan/ieee754status/IEEE754.PDF" target="_blank">https://web.archive.org/web/20120208075518/http://www.cs.berkeley.edu/~wkahan/ieee754status/IEEE754.PDF</a> <a href="#fnref:6" class="footnote-back-ref">↩</a></p></li> </ol>