forked from NRZCode/ia32-64
130 lines
5.6 KiB
HTML
130 lines
5.6 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>VCVTPS2PHX
|
|||
|
|
— Convert Packed Single Precision Floating-Point Values 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>VCVTPS2PHX
|
|||
|
|
— Convert Packed Single Precision Floating-Point Values to Packed FP16 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.MAP5.W0 1D /r VCVTPS2PHX xmm1{k1}{z}, xmm2/m128/m32bcst</td>
|
|||
|
|
<td>A</td>
|
|||
|
|
<td>V/V</td>
|
|||
|
|
<td>AVX512-FP16 AVX512VL</td>
|
|||
|
|
<td>Convert four packed single precision floating-point values in xmm2/m128/m32bcst to packed FP16 values, and store the result in xmm1 subject to writemask k1.</td></tr>
|
|||
|
|
<tr>
|
|||
|
|
<td>EVEX.256.66.MAP5.W0 1D /r VCVTPS2PHX xmm1{k1}{z}, ymm2/m256/m32bcst</td>
|
|||
|
|
<td>A</td>
|
|||
|
|
<td>V/V</td>
|
|||
|
|
<td>AVX512-FP16 AVX512VL</td>
|
|||
|
|
<td>Convert eight packed single precision floating-point values in ymm2/m256/m32bcst to packed FP16 values, and store the result in xmm1 subject to writemask k1.</td></tr>
|
|||
|
|
<tr>
|
|||
|
|
<td>EVEX.512.66.MAP5.W0 1D /r VCVTPS2PHX ymm1{k1}{z}, zmm2/m512/m32bcst {er}</td>
|
|||
|
|
<td>A</td>
|
|||
|
|
<td>V/V</td>
|
|||
|
|
<td>AVX512-FP16</td>
|
|||
|
|
<td>Convert sixteen packed single precision floating-point values in zmm2 /m512/m32bcst to 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 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>This instruction converts packed single precision floating values in the source operand to FP16 values and stores to the destination operand.</p>
|
|||
|
|
<p>The VCVTPS2PHX instruction supports broadcasting.</p>
|
|||
|
|
<p>This instruction uses MXCSR.DAZ for handling FP32 inputs. FP16 outputs can be normal or denormal numbers, and are not conditionally flushed based on MXCSR settings.</p>
|
|||
|
|
<h3 id="operation">Operation<a class="anchor" href="#operation">
|
|||
|
|
¶
|
|||
|
|
</a></h3>
|
|||
|
|
<h4 id="vcvtps2phx-dest--src--avx512_fp16-load-version-with-broadcast-support-">VCVTPS2PHX DEST, SRC (AVX512_FP16 Load Version With Broadcast Support)<a class="anchor" href="#vcvtps2phx-dest--src--avx512_fp16-load-version-with-broadcast-support-">
|
|||
|
|
¶
|
|||
|
|
</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.fp32[0]
|
|||
|
|
ELSE
|
|||
|
|
tsrc := SRC.fp32[j]
|
|||
|
|
DEST.fp16[j] := Convert_fp32_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="flags-affected">Flags Affected<a class="anchor" href="#flags-affected">
|
|||
|
|
¶
|
|||
|
|
</a></h3>
|
|||
|
|
<p>None.</p>
|
|||
|
|
<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>VCVTPS2PHX __m256h _mm512_cvtx_roundps_ph (__m512 a, int rounding);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m256h _mm512_mask_cvtx_roundps_ph (__m256h src, __mmask16 k, __m512 a, int rounding);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m256h _mm512_maskz_cvtx_roundps_ph (__mmask16 k, __m512 a, int rounding);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m128h _mm_cvtxps_ph (__m128 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m128h _mm_mask_cvtxps_ph (__m128h src, __mmask8 k, __m128 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m128h _mm_maskz_cvtxps_ph (__mmask8 k, __m128 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m128h _mm256_cvtxps_ph (__m256 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m128h _mm256_mask_cvtxps_ph (__m128h src, __mmask8 k, __m256 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m128h _mm256_maskz_cvtxps_ph (__mmask8 k, __m256 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m256h _mm512_cvtxps_ph (__m512 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m256h _mm512_mask_cvtxps_ph (__m256h src, __mmask16 k, __m512 a);
|
|||
|
|
</pre>
|
|||
|
|
<pre>VCVTPS2PHX __m256h _mm512_maskz_cvtxps_ph (__mmask16 k, __m512 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, Underflow, Overflow, Precision, Denormal (if MXCSR.DAZ=0).</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>
|
|||
|
|
<p>Additionally:</p>
|
|||
|
|
<table>
|
|||
|
|
<tr>
|
|||
|
|
<td>#UD</td>
|
|||
|
|
<td>If VEX.W=1.</td></tr>
|
|||
|
|
<tr>
|
|||
|
|
<td>#UD</td>
|
|||
|
|
<td>If VEX.vvvv != 1111B or 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>
|