forked from NRZCode/ia32-64
120 lines
5.7 KiB
HTML
120 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>VCVTDQ2PH
|
||
— Convert Packed Signed Doubleword Integers to 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>VCVTDQ2PH
|
||
— Convert Packed Signed Doubleword Integers to 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.NP.MAP5.W0 5B /r VCVTDQ2PH xmm1{k1}{z}, xmm2/m128/m32bcst</td>
|
||
<td>A</td>
|
||
<td>V/V</td>
|
||
<td>AVX512-FP16 AVX512VL</td>
|
||
<td>Convert four packed signed doubleword integers from xmm2/m128/m32bcst to four packed FP16 values, and store the result in xmm1 subject to writemask k1.</td></tr>
|
||
<tr>
|
||
<td>EVEX.256.NP.MAP5.W0 5B /r VCVTDQ2PH xmm1{k1}{z}, ymm2/m256/m32bcst</td>
|
||
<td>A</td>
|
||
<td>V/V</td>
|
||
<td>AVX512-FP16 AVX512VL</td>
|
||
<td>Convert eight packed signed doubleword integers from ymm2/m256/m32bcst to eight packed FP16 values, and store the result in xmm1 subject to writemask k1.</td></tr>
|
||
<tr>
|
||
<td>EVEX.512.NP.MAP5.W0 5B /r VCVTDQ2PH ymm1{k1}{z}, zmm2/m512/m32bcst {er}</td>
|
||
<td>A</td>
|
||
<td>V/V</td>
|
||
<td>AVX512-FP16</td>
|
||
<td>Convert sixteen packed signed doubleword integers from zmm2/m512/m32bcst to sixteen packed FP16 values, and store the result in ymm1 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 converts four, eight, or sixteen packed signed doubleword integers in the source operand to four, eight, or sixteen packed FP16 values in the destination operand.</p>
|
||
<p>EVEX encoded versions: The source operand can be a ZMM/YMM/XMM register, a 512/256/128-bit memory location or a 512/256/128-bit vector broadcast from a 32-bit memory location. The destination operand is a YMM/XMM register conditionally updated with writemask k1.</p>
|
||
<p>EVEX.vvvv is reserved and must be 1111b, otherwise instructions will #UD.</p>
|
||
<p>If the result of the convert operation is overflow and MXCSR.OM=0 then a SIMD exception will be raised with OE=1, PE=1.</p>
|
||
<h3 id="operation">Operation<a class="anchor" href="#operation">
|
||
¶
|
||
</a></h3>
|
||
<h4 id="vcvtdq2ph-dest--src">VCVTDQ2PH DEST, SRC<a class="anchor" href="#vcvtdq2ph-dest--src">
|
||
¶
|
||
</a></h4>
|
||
<pre>VL = 128, 256 or 512
|
||
KL := VL / 32
|
||
IF *SRC is a register* and (VL = 512) AND (EVEX.b = 1):
|
||
SET_RM(EVEX.RC)
|
||
ELSE:
|
||
SET_RM(MXCSR.RC)
|
||
FOR j := 0 TO KL-1:
|
||
IF k1[j] OR *no writemask*:
|
||
IF *SRC is memory* and EVEX.b = 1:
|
||
tsrc := SRC.dword[0]
|
||
ELSE
|
||
tsrc := SRC.dword[j]
|
||
DEST.fp16[j] := Convert_integer32_to_fp16(tsrc)
|
||
ELSE IF *zeroing*:
|
||
DEST.fp16[j] := 0
|
||
// else dest.fp16[j] remains unchanged
|
||
DEST[MAXVL-1:VL/2] := 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>VCVTDQ2PH __m256h _mm512_cvt_roundepi32_ph (__m512i a, int rounding);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m256h _mm512_mask_cvt_roundepi32_ph (__m256h src, __mmask16 k, __m512i a, int rounding);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m256h _mm512_maskz_cvt_roundepi32_ph (__mmask16 k, __m512i a, int rounding);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m128h _mm_cvtepi32_ph (__m128i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m128h _mm_mask_cvtepi32_ph (__m128h src, __mmask8 k, __m128i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m128h _mm_maskz_cvtepi32_ph (__mmask8 k, __m128i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m128h _mm256_cvtepi32_ph (__m256i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m128h _mm256_mask_cvtepi32_ph (__m128h src, __mmask8 k, __m256i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m128h _mm256_maskz_cvtepi32_ph (__mmask8 k, __m256i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m256h _mm512_cvtepi32_ph (__m512i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m256h _mm512_mask_cvtepi32_ph (__m256h src, __mmask16 k, __m512i a);
|
||
</pre>
|
||
<pre>VCVTDQ2PH __m256h _mm512_maskz_cvtepi32_ph (__mmask16 k, __m512i 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>Overflow, Precision.</p>
|
||
<h3 class="exceptions" id="other-exceptions">Other Exceptions<a class="anchor" href="#other-exceptions">
|
||
¶
|
||
</a></h3>
|
||
<p>EVEX-encoded instructions, see <span class="not-imported">Table 2-46</span>, “Type E2 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>
|