ia32-64/x86/vfmsubadd132ph.vfmsubadd213ph.vfmsubadd231ph.html
2025-07-08 02:23:29 -03:00

282 lines
13 KiB
HTML
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<!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>VFMSUBADD132PH/VFMSUBADD213PH/VFMSUBADD231PH
— Fused Multiply-AlternatingSubtract/Add 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>VFMSUBADD132PH/VFMSUBADD213PH/VFMSUBADD231PH
— Fused Multiply-AlternatingSubtract/Add 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 97 /r VFMSUBADD132PH xmm1{k1}{z}, xmm2, xmm3/m128/m16bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16 AVX512VL</td>
<td>Multiply packed FP16 values from xmm1 and xmm3/m128/m16bcst, subtract/add elements in xmm2, and store the result in xmm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.256.66.MAP6.W0 97 /r VFMSUBADD132PH ymm1{k1}{z}, ymm2, ymm3/m256/m16bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16 AVX512VL</td>
<td>Multiply packed FP16 values from ymm1 and ymm3/m256/m16bcst, subtract/add elements in ymm2, and store the result in ymm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.512.66.MAP6.W0 97 /r VFMSUBADD132PH zmm1{k1}{z}, zmm2, zmm3/m512/m16bcst {er}</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16</td>
<td>Multiply packed FP16 values from zmm1 and zmm3/m512/m16bcst, subtract/add elements in zmm2, and store the result in zmm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.128.66.MAP6.W0 A7 /r VFMSUBADD213PH xmm1{k1}{z}, xmm2, xmm3/m128/m16bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16 AVX512VL</td>
<td>Multiply packed FP16 values from xmm1 and xmm2, subtract/add elements in xmm3/m128/m16bcst, and store the result in xmm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.256.66.MAP6.W0 A7 /r VFMSUBADD213PH ymm1{k1}{z}, ymm2, ymm3/m256/m16bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16 AVX512VL</td>
<td>Multiply packed FP16 values from ymm1 and ymm2, subtract/add elements in ymm3/m256/m16bcst, and store the result in ymm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.512.66.MAP6.W0 A7 /r VFMSUBADD213PH zmm1{k1}{z}, zmm2, zmm3/m512/m16bcst {er}</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16</td>
<td>Multiply packed FP16 values from zmm1 and zmm2, subtract/add elements in zmm3/m512/m16bcst, and store the result in zmm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.128.66.MAP6.W0 B7 /r VFMSUBADD231PH xmm1{k1}{z}, xmm2, xmm3/m128/m16bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16 AVX512VL</td>
<td>Multiply packed FP16 values from xmm2 and xmm3/m128/m16bcst, subtract/add elements in xmm1, and store the result in xmm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.256.66.MAP6.W0 B7 /r VFMSUBADD231PH ymm1{k1}{z}, ymm2, ymm3/m256/m16bcst</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16 AVX512VL</td>
<td>Multiply packed FP16 values from ymm2 and ymm3/m256/m16bcst, subtract/add elements in ymm1, and store the result in ymm1 subject to writemask k1.</td></tr>
<tr>
<td>EVEX.512.66.MAP6.W0 B7 /r VFMSUBADD231PH zmm1{k1}{z}, zmm2, zmm3/m512/m16bcst {er}</td>
<td>A</td>
<td>V/V</td>
<td>AVX512-FP16</td>
<td>Multiply packed FP16 values from zmm2 and zmm3/m512/m16bcst, subtract/add elements in zmm1, 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 (r, 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 packed multiply-add (even elements) or multiply-subtract (odd elements) computation on FP16 values using three source operands and writes the results in the destination operand. The destination operand is also the first source operand. The notation “132”, “213” and “231” indicate the use of the operands in A * B ± C, where each digit corresponds to the operand number, with the destination being operand 1; see <a href='vfmsubadd132ph.vfmsubadd213ph.vfmsubadd231ph.html#tbl-5-8'>Table 5-8</a>.</p>
<p>The destination elements are updated according to the writemask.</p>
<figure id="tbl-5-8">
<table>
<tr>
<th>Notation</th>
<th>Odd Elements</th>
<th>Even Elements</th></tr>
<tr>
<td>132</td>
<td>dest = dest*src3-src2</td>
<td>dest = dest*src3+src2</td></tr>
<tr>
<td>231</td>
<td>dest = src2*src3-dest</td>
<td>dest = src2*src3+dest</td></tr>
<tr>
<td>213</td>
<td>dest = src2*dest-src3</td>
<td>dest = src2*dest+src3</td></tr></table>
<figcaption><a href='vfmsubadd132ph.vfmsubadd213ph.vfmsubadd231ph.html#tbl-5-8'>Table 5-8</a>. VFMSUBADD[132,213,231]PH Notation for Odd and Even Elements</figcaption></figure>
<h3 id="operation">Operation<a class="anchor" href="#operation">
</a></h3>
<h4 id="vfmsubadd132ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-register">VFMSUBADD132PH DEST, SRC2, SRC3 (EVEX encoded versions) when src3 operand is a register<a class="anchor" href="#vfmsubadd132ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-register">
</a></h4>
<pre>VL = 128, 256 or 512
KL := VL/16
IF (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 *j is even*:
DEST.fp16[j] := RoundFPControl(DEST.fp16[j]*SRC3.fp16[j] + SRC2.fp16[j])
ELSE:
DEST.fp16[j] := RoundFPControl(DEST.fp16[j]*SRC3.fp16[j] - SRC2.fp16[j])
ELSE IF *zeroing*:
DEST.fp16[j] := 0
// else dest.fp16[j] remains unchanged
DEST[MAXVL-1:VL] := 0
</pre>
<h4 id="vfmsubadd132ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-memory-source">VFMSUBADD132PH DEST, SRC2, SRC3 (EVEX encoded versions) when src3 operand is a memory source<a class="anchor" href="#vfmsubadd132ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-memory-source">
</a></h4>
<pre>VL = 128, 256 or 512
KL := VL/16
FOR j := 0 TO KL-1:
IF k1[j] OR *no writemask*:
IF EVEX.b = 1:
t3 := SRC3.fp16[0]
ELSE:
t3 := SRC3.fp16[j]
IF *j is even*:
DEST.fp16[j] := RoundFPControl(DEST.fp16[j] * t3 + SRC2.fp16[j])
ELSE:
DEST.fp16[j] := RoundFPControl(DEST.fp16[j] * t3 - SRC2.fp16[j])
ELSE IF *zeroing*:
DEST.fp16[j] := 0
// else dest.fp16[j] remains unchanged
DEST[MAXVL-1:VL] := 0:
</pre>
<h4 id="vfmsubadd213ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-register">VFMSUBADD213PH DEST, SRC2, SRC3 (EVEX encoded versions) when src3 operand is a register<a class="anchor" href="#vfmsubadd213ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-register">
</a></h4>
<pre>VL = 128, 256 or 512
KL := VL/16
IF (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 *j is even*:
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j]*DEST.fp16[j] + SRC3.fp16[j])
ELSE
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j]*DEST.fp16[j] - SRC3.fp16[j])
ELSE IF *zeroing*:
DEST.fp16[j] := 0
// else dest.fp16[j] remains unchanged
DEST[MAXVL-1:VL] := 0
</pre>
<h4 id="vfmsubadd213ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-memory-source">VFMSUBADD213PH DEST, SRC2, SRC3 (EVEX encoded versions) when src3 operand is a memory source<a class="anchor" href="#vfmsubadd213ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-memory-source">
</a></h4>
<pre>VL = 128, 256 or 512
KL := VL/16
FOR j := 0 TO KL-1:
IF k1[j] OR *no writemask*:
IF EVEX.b = 1:
t3 := SRC3.fp16[0]
ELSE:
t3 := SRC3.fp16[j]
IF *j is even*:
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j] * DEST.fp16[j] + t3 )
ELSE:
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j] * DEST.fp16[j] - t3 )
ELSE IF *zeroing*:
DEST.fp16[j] := 0
// else dest.fp16[j] remains unchanged
DEST[MAXVL-1:VL] := 0:
</pre>
<h4 id="vfmsubadd231ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-register">VFMSUBADD231PH DEST, SRC2, SRC3 (EVEX encoded versions) when src3 operand is a register<a class="anchor" href="#vfmsubadd231ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-register">
</a></h4>
<pre>VL = 128, 256 or 512
KL := VL/16
IF (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 *j is even:
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j]*SRC3.fp16[j] + DEST.fp16[j])
ELSE:
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j]*SRC3.fp16[j] - DEST.fp16[j])
ELSE IF *zeroing*:
DEST.fp16[j] := 0
// else dest.fp16[j] remains unchanged
DEST[MAXVL-1:VL] := 0
</pre>
<h4 id="vfmsubadd231ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-memory-source">VFMSUBADD231PH DEST, SRC2, SRC3 (EVEX encoded versions) when src3 operand is a memory source<a class="anchor" href="#vfmsubadd231ph-dest--src2--src3--evex-encoded-versions--when-src3-operand-is-a-memory-source">
</a></h4>
<pre>VL = 128, 256 or 512
KL := VL/16
FOR j := 0 TO KL-1:
IF k1[j] OR *no writemask*:
IF EVEX.b = 1:
t3 := SRC3.fp16[0]
ELSE:
t3 := SRC3.fp16[j]
IF *j is even*:
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j] * t3 + DEST.fp16[j] )
ELSE:
DEST.fp16[j] := RoundFPControl(SRC2.fp16[j] * t3 - DEST.fp16[j] )
ELSE IF *zeroing*:
DEST.fp16[j] := 0
// else dest.fp16[j] 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>VFMSUBADD132PH, VFMSUBADD213PH, and VFMSUBADD231PH: __m128h _mm_fmsubadd_ph (__m128h a, __m128h b, __m128h c);
</pre>
<pre>__m128h _mm_mask_fmsubadd_ph (__m128h a, __mmask8 k, __m128h b, __m128h c);
</pre>
<pre>__m128h _mm_mask3_fmsubadd_ph (__m128h a, __m128h b, __m128h c, __mmask8 k);
</pre>
<pre>__m128h _mm_maskz_fmsubadd_ph (__mmask8 k, __m128h a, __m128h b, __m128h c);
</pre>
<pre>__m256h _mm256_fmsubadd_ph (__m256h a, __m256h b, __m256h c);
</pre>
<pre>__m256h _mm256_mask_fmsubadd_ph (__m256h a, __mmask16 k, __m256h b, __m256h c);
</pre>
<pre>__m256h _mm256_mask3_fmsubadd_ph (__m256h a, __m256h b, __m256h c, __mmask16 k);
</pre>
<pre>__m256h _mm256_maskz_fmsubadd_ph (__mmask16 k, __m256h a, __m256h b, __m256h c);
</pre>
<pre>__m512h _mm512_fmsubadd_ph (__m512h a, __m512h b, __m512h c);
</pre>
<pre>__m512h _mm512_mask_fmsubadd_ph (__m512h a, __mmask32 k, __m512h b, __m512h c);
</pre>
<pre>__m512h _mm512_mask3_fmsubadd_ph (__m512h a, __m512h b, __m512h c, __mmask32 k);
</pre>
<pre>__m512h _mm512_maskz_fmsubadd_ph (__mmask32 k, __m512h a, __m512h b, __m512h c);
</pre>
<pre>__m512h _mm512_fmsubadd_round_ph (__m512h a, __m512h b, __m512h c, const int rounding);
</pre>
<pre>__m512h _mm512_mask_fmsubadd_round_ph (__m512h a, __mmask32 k, __m512h b, __m512h c, const int rounding);
</pre>
<pre>__m512h _mm512_mask3_fmsubadd_round_ph (__m512h a, __m512h b, __m512h c, __mmask32 k, const int rounding);
</pre>
<pre>__m512h _mm512_maskz_fmsubadd_round_ph (__mmask32 k, __m512h a, __m512h b, __m512h c, const int rounding);
</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.</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 Developers Manual</a> for anything serious.
</p></footer></body></html>