80 lines
4.3 KiB
HTML
80 lines
4.3 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>VRCPSH
|
|||
|
|
— Compute Reciprocal of Scalar FP16 Value</title></head><body><header><nav><ul><li><a href='index.html'>Index</a></li><li>December 2023</li></ul></nav></header><h1>VRCPSH
|
|||
|
|
— Compute Reciprocal of Scalar FP16 Value</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.LLIG.66.MAP6.W0 4D /r VRCPSH xmm1{k1}{z}, xmm2, xmm3/m16</td>
|
|||
|
|
<td>A</td>
|
|||
|
|
<td>V/V</td>
|
|||
|
|
<td>AVX512-FP16</td>
|
|||
|
|
<td>Compute the approximate reciprocal of the low FP16 value in xmm3/m16 and store the result in xmm1 subject to writemask k1. Bits 127:16 from xmm2 are copied to xmm1[127:16].</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>Scalar</td>
|
|||
|
|
<td>ModRM:reg (w)</td>
|
|||
|
|
<td>VEX.vvvv (r)</td>
|
|||
|
|
<td>ModRM:r/m (r)</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 reciprocal of the low FP16 value in the second source operand (the third operand) and stores the result in the low word element of the destination operand (the first operand) according to the writemask k1. Bits 127:16 of the XMM register destination are copied from corresponding bits in the first source operand (the second operand). The maximum relative error for this approximation is less than 2<sup>−11</sup> + 2<sup>−14</sup>.</p>
|
|||
|
|
<p>Bits 127:16 of the destination operand are copied from the corresponding bits of the first source operand. Bits MAXVL-1:128 of the destination operand are zeroed. The low FP16 element of the destination is updated according to the writemask.</p>
|
|||
|
|
<p>For special cases, see <a href='vrcpph.html#tbl-5-28'>Table 5-28</a>.</p>
|
|||
|
|
<h3 id="operation">Operation<a class="anchor" href="#operation">
|
|||
|
|
¶
|
|||
|
|
</a></h3>
|
|||
|
|
<h4 id="vrcpsh-dest-k1---src1--src2">VRCPSH dest{k1}, src1, src2<a class="anchor" href="#vrcpsh-dest-k1---src1--src2">
|
|||
|
|
¶
|
|||
|
|
</a></h4>
|
|||
|
|
<pre>IF k1[0] or *no writemask*:
|
|||
|
|
DEST.fp16[0] := APPROXIMATE(1.0 / src2.fp16[0])
|
|||
|
|
ELSE IF *zeroing*:
|
|||
|
|
DEST.fp16[0] := 0
|
|||
|
|
//else DEST.fp16[0] remains unchanged
|
|||
|
|
DEST[127:16] := src1[127:16]
|
|||
|
|
DEST[MAXVL-1:128] := 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>VRCPSH __m128h _mm_mask_rcp_sh (__m128h src, __mmask8 k, __m128h a, __m128h b);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VRCPSH __m128h _mm_maskz_rcp_sh (__mmask8 k, __m128h a, __m128h b);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VRCPSH __m128h _mm_rcp_sh (__m128h a, __m128h b);
|
|||
|
|
</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 2-58</span>, “Type E10 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>
|