140 lines
5.7 KiB
HTML
140 lines
5.7 KiB
HTML
<!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>VRSQRTPH
|
||
— Compute Reciprocals of Square Roots of Packed FP16 Values</title></head><body><header><nav><ul><li><a href='index.html'>Index</a></li><li>December 2023</li></ul></nav></header><h1>VRSQRTPH
|
||
— Compute Reciprocals of Square Roots of Packed FP16 Values</h1>
|
||
|
||
<table>
|
||
<tr>
|
||
<th> Instruction En bit Mode Flag
|
||
Support Instruction En bit Mode Flag
|
||
Support 64/32 CPUID Feature Instruction En bit Mode Flag CPUID Feature Instruction En bit Mode Flag Op/ 64/32 CPUID Feature Instruction En bit Mode Flag 64/32 CPUID Feature Instruction En bit Mode Flag CPUID Feature Instruction En bit Mode Flag Op/ 64/32 CPUID Feature </th>
|
||
<th></th>
|
||
<th>Support</th>
|
||
<th></th>
|
||
<th>Description</th></tr>
|
||
<tr>
|
||
<td>EVEX.128.66.MAP6.W0 4E /r VRSQRTPH xmm1{k1}{z}, xmm2/m128/m16bcst</td>
|
||
<td>A</td>
|
||
<td>V/V</td>
|
||
<td>AVX512-FP16 AVX512VL</td>
|
||
<td>Compute the approximate reciprocals of the square roots of packed FP16 values in xmm2/m128/m16bcst and store the result in xmm1 subject to writemask k1.</td></tr>
|
||
<tr>
|
||
<td>EVEX.256.66.MAP6.W0 4E /r VRSQRTPH ymm1{k1}{z}, ymm2/m256/m16bcst</td>
|
||
<td>A</td>
|
||
<td>V/V</td>
|
||
<td>AVX512-FP16 AVX512VL</td>
|
||
<td>Compute the approximate reciprocals of the square roots of packed FP16 values in ymm2/m256/m16bcst and store the result in ymm1 subject to writemask k1.</td></tr>
|
||
<tr>
|
||
<td>EVEX.512.66.MAP6.W0 4E /r VRSQRTPH zmm1{k1}{z}, zmm2/m512/m16bcst</td>
|
||
<td>A</td>
|
||
<td>V/V</td>
|
||
<td>AVX512-FP16</td>
|
||
<td>Compute the approximate reciprocals of the square roots of packed FP16 values in zmm2/m512/m16bcst and store the result in zmm1 subject to 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</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>This instruction performs a SIMD computation of the approximate reciprocals square-root of 8/16/32 packed FP16 floating-point values in the source operand (the second operand) and stores the packed FP16 floating-point results in the destination operand.</p>
|
||
<p>The maximum relative error for this approximation is less than 2<sup>−11</sup> + 2<sup>−14</sup>. For special cases, see <a href='vrsqrtph.html#tbl-5-38'>Table 5-38</a>.</p>
|
||
<p>The destination elements are updated according to the writemask.</p>
|
||
<figure id="tbl-5-38">
|
||
<table>
|
||
<tr>
|
||
<th>Input value</th>
|
||
<th>Reset Value</th>
|
||
<th>Comments</th></tr>
|
||
<tr>
|
||
<td>Any denormal</td>
|
||
<td>Normal</td>
|
||
<td>Cannot generate overflow</td></tr>
|
||
<tr>
|
||
<td>X = 2<em><sup>−2n</sup></em></td>
|
||
<td><sub>2</sub><em><sup>n</sup></em></td>
|
||
<td></td></tr>
|
||
<tr>
|
||
<td>X<0</td>
|
||
<td>QNaN_Indefinite</td>
|
||
<td>Including −∞</td></tr>
|
||
<tr>
|
||
<td>X = −0</td>
|
||
<td>−∞</td>
|
||
<td></td></tr>
|
||
<tr>
|
||
<td>X = +0</td>
|
||
<td>+∞</td>
|
||
<td></td></tr>
|
||
<tr>
|
||
<td>X = +∞</td>
|
||
<td>+0</td>
|
||
<td></td></tr></table>
|
||
<figcaption><a href='vrsqrtph.html#tbl-5-38'>Table 5-38</a>. VRSQRTPH/VRSQRTSH Special Cases</figcaption></figure>
|
||
<h3 id="operation">Operation<a class="anchor" href="#operation">
|
||
¶
|
||
</a></h3>
|
||
<h4 id="vrsqrtph-dest-k1---src">VRSQRTPH dest{k1}, src<a class="anchor" href="#vrsqrtph-dest-k1---src">
|
||
¶
|
||
</a></h4>
|
||
<pre>VL = 128, 256 or 512
|
||
KL := VL/16
|
||
FOR i := 0 to KL-1:
|
||
IF k1[i] or *no writemask*:
|
||
IF SRC is memory and (EVEX.b = 1):
|
||
tsrc := src.fp16[0]
|
||
ELSE:
|
||
tsrc := src.fp16[i]
|
||
DEST.fp16[i] := APPROXIMATE(1.0 / SQRT(tsrc) )
|
||
ELSE IF *zeroing*:
|
||
DEST.fp16[i] := 0
|
||
//else DEST.fp16[i] remains unchanged
|
||
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>VRSQRTPH __m128h _mm_mask_rsqrt_ph (__m128h src, __mmask8 k, __m128h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m128h _mm_maskz_rsqrt_ph (__mmask8 k, __m128h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m128h _mm_rsqrt_ph (__m128h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m256h _mm256_mask_rsqrt_ph (__m256h src, __mmask16 k, __m256h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m256h _mm256_maskz_rsqrt_ph (__mmask16 k, __m256h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m256h _mm256_rsqrt_ph (__m256h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m512h _mm512_mask_rsqrt_ph (__m512h src, __mmask32 k, __m512h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m512h _mm512_maskz_rsqrt_ph (__mmask32 k, __m512h a);
|
||
</pre>
|
||
<pre>VRSQRTPH __m512h _mm512_rsqrt_ph (__m512h 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>None.</p>
|
||
<h3 class="exceptions" id="other-exceptions">Other Exceptions<a class="anchor" href="#other-exceptions">
|
||
¶
|
||
</a></h3>
|
||
<p>EVEX-encoded instruction, see <span class="not-imported">Table 2-49</span>, “Type E4 Class Exception Conditions.”</p><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>
|