NRZCode/ia32-64
1
0
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
ia32-64/x86/vgetexppd.html

210 lines
8.8 KiB
HTML
Raw Permalink Normal View History

Initial commit
2025-07-08 02:23:29 -03:00
<!DOCTYPE html>
<html xmlns="http://www.w3.org/1999/xhtml" xmlns:svg="http://www.w3.org/2000/svg" xmlns:x86="http://www.felixcloutier.com/x86"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"><link rel="stylesheet" type="text/css" href="style.css"></link><title>VGETEXPPD
— Convert Exponents of Packed Double Precision Floating-Point Values to DoublePrecision Floating-Point Values</title></head><body><header><nav><ul><li><a href='index.html'>Index</a></li><li>December 2023</li></ul></nav></header><h1>VGETEXPPD
— Convert Exponents of Packed Double Precision Floating-Point Values to DoublePrecision Floating-Point Values</h1>
<table>
<tr>
<th>Opcode/Instruction</th>
<th>Op/En</th>
<th>64/32 Bit Mode Support</th>
<th>CPUID Feature Flag</th>
<th>Description</th></tr>
<tr>
<td>EVEX.128.66.0F38.W1 42 /r VGETEXPPD xmm1 {k1}{z}, xmm2/m128/m64bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512VL AVX512F</td>
<td>Convert the exponent of packed double precision floating-point values in the source operand to double precision floating-point results representing unbiased integer exponents and stores the results in the destination register.</td></tr>
<tr>
<td>EVEX.256.66.0F38.W1 42 /r VGETEXPPD ymm1 {k1}{z}, ymm2/m256/m64bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512VL AVX512F</td>
<td>Convert the exponent of packed double precision floating-point values in the source operand to double precision floating-point results representing unbiased integer exponents and stores the results in the destination register.</td></tr>
<tr>
<td>EVEX.512.66.0F38.W1 42 /r VGETEXPPD zmm1 {k1}{z}, zmm2/m512/m64bcst{sae}</td>
<td>A</td>
<td>V/V</td>
<td>AVX512F</td>
<td>Convert the exponent of packed double precision floating-point values in the source operand to double precision floating-point results representing unbiased integer exponents and stores the results in the destination under writemask k1.</td></tr></table>
<h2 id="instruction-operand-encoding">Instruction Operand Encoding<a class="anchor" href="#instruction-operand-encoding">
</a></h2>
<table>
<tr>
<th>Op/En</th>
<th>Tuple Type</th>
<th>Operand 1</th>
<th>Operand 2</th>
<th>Operand 3</th>
<th>Operand 4</th></tr>
<tr>
<td>A</td>
<td>Full</td>
<td>ModRM:reg (w)</td>
<td>ModRM:r/m (r)</td>
<td>N/A</td>
<td>N/A</td></tr></table>
<h3 id="description">Description<a class="anchor" href="#description">
</a></h3>
<p>Extracts the biased exponents from the normalized double precision floating-point representation of each qword data element of the source operand (the second operand) as unbiased signed integer value, or convert the denormal representation of input data to unbiased negative integer values. Each integer value of the unbiased exponent is converted to double precision floating-point value and written to the corresponding qword elements of the destination operand (the first operand) as double precision floating-point numbers.</p>
<p>The destination operand is a ZMM/YMM/XMM register and updated under the writemask. The source operand can be a ZMM/YMM/XMM register, a 512/256/128-bit memory location, or a 512/256/128-bit vector broadcasted from a 64-bit memory location.</p>
<p>EVEX.vvvv is reserved and must be 1111b, otherwise instructions will #UD.</p>
<p>Each GETEXP operation converts the exponent value into a floating-point number (permitting input value in denormal representation). Special cases of input values are listed in <a href='vgetexppd.html#tbl-5-15'>Table 5-15</a>.</p>
<p>The formula is:</p>
<p>GETEXP(x) = floor(log<sub>2</sub>(|x|))</p>
<p>Notation <strong>floor(x)</strong> stands for the greatest integer not exceeding real number x.</p>
<figure id="tbl-5-15">
<table>
<tr>
<th>Input Operand</th>
<th>Result</th>
<th>Comments</th></tr>
<tr>
<td>src1 = NaN</td>
<td>QNaN(src1)</td>
<td rowspan="4">If (SRC = SNaN) then #IE If (SRC = denormal) then #DE</td></tr>
<tr>
<td>0 &lt; |src1| &lt; INF</td>
<td>floor(log<sub>2</sub>(|src1|))</td></tr>
<tr>
<td>| src1| = +INF</td>
<td>+INF</td></tr>
<tr>
<td>| src1| = 0</td>
<td>-INF</td></tr></table>
<figcaption><a href='vgetexppd.html#tbl-5-15'>Table 5-15</a>. VGETEXPPD/SD Special Cases</figcaption></figure>
<h3 id="operation">Operation<a class="anchor" href="#operation">
</a></h3>
<pre>NormalizeExpTinyDPFP(SRC[63:0])
{
// Jbit is the hidden integral bit of a floating-point number. In case of denormal number it has the value of ZERO.
Src.Jbit := 0;
Dst.exp := 1;
Dst.fraction := SRC[51:0];
WHILE(Src.Jbit = 0)
{
Src.Jbit := Dst.fraction[51];
// Get the fraction MSB
Dst.fraction := Dst.fraction &lt;&lt; 1 ;
// One bit shift left
Dst.exp-- ;
// Decrement the exponent
}
Dst.fraction := 0;
Dst.sign := 1;
TMP[63:0] := MXCSR.DAZ? 0 : (Dst.sign &lt;&lt; 63) OR (Dst.exp &lt;&lt; 52) OR (Dst.fraction) ;
Return (TMP[63:0]);
}
ConvertExpDPFP(SRC[63:0])
{
Src.sign := 0;
// Zero out sign bit
Src.exp := SRC[62:52];
Src.fraction := SRC[51:0];
// Check for NaN
IF (SRC = NaN)
{
IF ( SRC = SNAN ) SET IE;
Return QNAN(SRC);
}
// Check for +INF
IF (Src = +INF) RETURN (Src);
// check if zero operand
IF ((Src.exp = 0) AND ((Src.fraction = 0) OR (MXCSR.DAZ = 1))) Return (-INF);
}
ELSE // check if denormal operand (notice that MXCSR.DAZ = 0)
{
IF ((Src.exp = 0) AND (Src.fraction != 0))
{
TMP[63:0] := NormalizeExpTinyDPFP(SRC[63:0]) ;
// Get Normalized Exponent
Set #DE
}
ELSE // exponent value is correct
{
TMP[63:0] := (Src.sign &lt;&lt; 63) OR (Src.exp &lt;&lt; 52) OR (Src.fraction) ;
}
TMP := SAR(TMP, 52) ;
// Shift Arithmetic Right
TMP := TMP – 1023;
// Subtract Bias
Return CvtI2D(TMP);
// Convert INT to double precision floating-point number
}
}
</pre>
<h4 id="vgetexppd--evex-encoded-versions-">VGETEXPPD (EVEX encoded versions)<a class="anchor" href="#vgetexppd--evex-encoded-versions-">
</a></h4>
<pre>(KL, VL) = (2, 128), (4, 256), (8, 512)
FOR j := 0 TO KL-1
i := j * 64
IF k1[j] OR *no writemask*
THEN
IF (EVEX.b = 1) AND (SRC *is memory*)
THEN
DEST[i+63:i] :=
ConvertExpDPFP(SRC[63:0])
ELSE
DEST[i+63:i] :=
ConvertExpDPFP(SRC[i+63:i])
FI;
ELSE
IF *merging-masking*
THEN *DEST[i+63:i] remains unchanged*
ELSE ; zeroing-masking
DEST[i+63:i] := 0
FI
FI;
ENDFOR
DEST[MAXVL-1:VL] := 0
</pre>
<h3 id="intel-c-c++-compiler-intrinsic-equivalent">Intel C/C++ Compiler Intrinsic Equivalent<a class="anchor" href="#intel-c-c++-compiler-intrinsic-equivalent">
</a></h3>
<pre>VGETEXPPD __m512d _mm512_getexp_pd(__m512d a);
</pre>
<pre>VGETEXPPD __m512d _mm512_mask_getexp_pd(__m512d s, __mmask8 k, __m512d a);
</pre>
<pre>VGETEXPPD __m512d _mm512_maskz_getexp_pd( __mmask8 k, __m512d a);
</pre>
<pre>VGETEXPPD __m512d _mm512_getexp_round_pd(__m512d a, int sae);
</pre>
<pre>VGETEXPPD __m512d _mm512_mask_getexp_round_pd(__m512d s, __mmask8 k, __m512d a, int sae);
</pre>
<pre>VGETEXPPD __m512d _mm512_maskz_getexp_round_pd( __mmask8 k, __m512d a, int sae);
</pre>
<pre>VGETEXPPD __m256d _mm256_getexp_pd(__m256d a);
</pre>
<pre>VGETEXPPD __m256d _mm256_mask_getexp_pd(__m256d s, __mmask8 k, __m256d a);
</pre>
<pre>VGETEXPPD __m256d _mm256_maskz_getexp_pd( __mmask8 k, __m256d a);
</pre>
<pre>VGETEXPPD __m128d _mm_getexp_pd(__m128d a);
</pre>
<pre>VGETEXPPD __m128d _mm_mask_getexp_pd(__m128d s, __mmask8 k, __m128d a);
</pre>
<pre>VGETEXPPD __m128d _mm_maskz_getexp_pd( __mmask8 k, __m128d a);
</pre>
<h3 class="exceptions" id="simd-floating-point-exceptions">SIMD Floating-Point Exceptions<a class="anchor" href="#simd-floating-point-exceptions">
</a></h3>
<p>Invalid, Denormal.</p>
<h3 class="exceptions" id="other-exceptions">Other Exceptions<a class="anchor" href="#other-exceptions">
</a></h3>
<p>See <span class="not-imported">Table 2-46</span>, “Type E2 Class Exception Conditions.”</p>
<p>Additionally:</p>
<table>
<tr>
<td>#UD</td>
<td>If EVEX.vvvv != 1111B.</td></tr></table><footer><p>
This UNOFFICIAL, mechanically-separated, non-verified reference is provided for convenience, but it may be
inc<span style="opacity: 0.2">omp</span>lete or b<sub>r</sub>oke<sub>n</sub> in various obvious or non-obvious
ways. Refer to <a href="https://software.intel.com/en-us/download/intel-64-and-ia-32-architectures-sdm-combined-volumes-1-2a-2b-2c-2d-3a-3b-3c-3d-and-4">Intel® 64 and IA-32 Architectures Software Developer’s Manual</a> for anything serious.
</p></footer></body></html>
Reference in a new issue Copy permalink