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

114 lines
6.9 KiB
HTML
Raw Permalink 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>RCPPS
— Compute Reciprocals of Packed Single Precision Floating-Point Values</title></head><body><header><nav><ul><li><a href='index.html'>Index</a></li><li>December 2023</li></ul></nav></header><h1>RCPPS
— Compute Reciprocals of Packed Single Precision Floating-Point 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>NP 0F 53 /r RCPPS xmm1, xmm2/m128</td>
<td>RM</td>
<td>V/V</td>
<td>SSE</td>
<td>Computes the approximate reciprocals of the packed single precision floating-point values in xmm2/m128 and stores the results in xmm1.</td></tr>
<tr>
<td>VEX.128.0F.WIG 53 /r VRCPPS xmm1, xmm2/m128</td>
<td>RM</td>
<td>V/V</td>
<td>AVX</td>
<td>Computes the approximate reciprocals of packed single precision values in xmm2/mem and stores the results in xmm1.</td></tr>
<tr>
<td>VEX.256.0F.WIG 53 /r VRCPPS ymm1, ymm2/m256</td>
<td>RM</td>
<td>V/V</td>
<td>AVX</td>
<td>Computes the approximate reciprocals of packed single precision values in ymm2/mem and stores the results in ymm1.</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>Operand 1</th>
<th>Operand 2</th>
<th>Operand 3</th>
<th>Operand 4</th></tr>
<tr>
<td>RM</td>
<td>ModRM:reg (w)</td>
<td>ModRM:r/m (r)</td>
<td>N/A</td>
<td>N/A</td></tr></table>
<h2 id="description">Description<a class="anchor" href="#description">
</a></h2>
<p>Performs a SIMD computation of the approximate reciprocals of the four packed single precision floating-point values in the source operand (second operand) stores the packed single precision floating-point results in the destination operand. The source operand can be an XMM register or a 128-bit memory location. The destination operand is an XMM register. See <span class="not-imported">Figure 10-5</span> in the Intel<sup>®</sup> 64 and IA-32 Architectures Software Developers Manual, Volume 1, for an illustration of a SIMD single precision floating-point operation.</p>
<p>The relative error for this approximation is:</p>
<p>|Relative Error| ≤ 1.5 2<sup>12</sup></p>
<p>The RCPPS instruction is not affected by the rounding control bits in the MXCSR register. When a source value is a 0.0, an ∞ of the sign of the source value is returned. A denormal source value is treated as a 0.0 (of the same sign). Tiny results (see Section 4.9.1.5, “Numeric Underflow Exception (#U)” in Intel<sup>®</sup> 64 and IA-32 Architectures Software Developers Manual, Volume 1) are always flushed to 0.0, with the sign of the operand. (Input values greater than or equal to |1.11111111110100000000000B2<sup>125</sup>| are guaranteed to not produce tiny results; input values less than or equal to |1.00000000000110000000001B*2<sup>126</sup>| are guaranteed to produce tiny results, which are in turn flushed to 0.0; and input values in between this range may or may not produce tiny results, depending on the implementation.) When a source value is an SNaN or QNaN, the SNaN is converted to a QNaN or the source QNaN is returned.</p>
<p>In 64-bit mode, using a REX prefix in the form of REX.R permits this instruction to access additional registers (XMM8-XMM15).</p>
<p>128-bit Legacy SSE version: The second source can be an XMM register or an 128-bit memory location. The destination is not distinct from the first source XMM register and the upper bits (MAXVL-1:128) of the corresponding YMM register destination are unmodified.</p>
<p>VEX.128 encoded version: the first source operand is an XMM register or 128-bit memory location. The destination operand is an XMM register. The upper bits (MAXVL-1:128) of the corresponding YMM register destination are zeroed.</p>
<p>VEX.256 encoded version: The first source operand is a YMM register. The second source operand can be a YMM register or a 256-bit memory location. The destination operand is a YMM register.</p>
<p>Note: In VEX-encoded versions, VEX.vvvv is reserved and must be 1111b, otherwise instructions will #UD.</p>
<h2 id="operation">Operation<a class="anchor" href="#operation">
</a></h2>
<h3 id="rcpps--128-bit-legacy-sse-version-">RCPPS (128-bit Legacy SSE Version)<a class="anchor" href="#rcpps--128-bit-legacy-sse-version-">
</a></h3>
<pre>DEST[31:0] := APPROXIMATE(1/SRC[31:0])
DEST[63:32] := APPROXIMATE(1/SRC[63:32])
DEST[95:64] := APPROXIMATE(1/SRC[95:64])
DEST[127:96] := APPROXIMATE(1/SRC[127:96])
DEST[MAXVL-1:128] (Unmodified)
</pre>
<h3 id="vrcpps--vex-128-encoded-version-">VRCPPS (VEX.128 Encoded Version)<a class="anchor" href="#vrcpps--vex-128-encoded-version-">
</a></h3>
<pre>DEST[31:0] := APPROXIMATE(1/SRC[31:0])
DEST[63:32] := APPROXIMATE(1/SRC[63:32])
DEST[95:64] := APPROXIMATE(1/SRC[95:64])
DEST[127:96] := APPROXIMATE(1/SRC[127:96])
DEST[MAXVL-1:128] := 0
</pre>
<h3 id="vrcpps--vex-256-encoded-version-">VRCPPS (VEX.256 Encoded Version)<a class="anchor" href="#vrcpps--vex-256-encoded-version-">
</a></h3>
<pre>DEST[31:0] := APPROXIMATE(1/SRC[31:0])
DEST[63:32] := APPROXIMATE(1/SRC[63:32])
DEST[95:64] := APPROXIMATE(1/SRC[95:64])
DEST[127:96] := APPROXIMATE(1/SRC[127:96])
DEST[159:128] := APPROXIMATE(1/SRC[159:128])
DEST[191:160] := APPROXIMATE(1/SRC[191:160])
DEST[223:192] := APPROXIMATE(1/SRC[223:192])
DEST[255:224] := APPROXIMATE(1/SRC[255:224])
</pre>
<h2 id="intel-c-c++-compiler-intrinsic-equivalent">Intel C/C++ Compiler Intrinsic Equivalent<a class="anchor" href="#intel-c-c++-compiler-intrinsic-equivalent">
</a></h2>
<pre>RCCPS __m128 _mm_rcp_ps(__m128 a)
</pre>
<pre>RCPPS __m256 _mm256_rcp_ps (__m256 a);
</pre>
<h2 class="exceptions" id="simd-floating-point-exceptions">SIMD Floating-Point Exceptions<a class="anchor" href="#simd-floating-point-exceptions">
</a></h2>
<p>None.</p>
<h2 class="exceptions" id="other-exceptions">Other Exceptions<a class="anchor" href="#other-exceptions">
</a></h2>
<p>See <span class="not-imported">Table 2-21</span>, “Type 4 Class Exception Conditions,” additionally:</p>
<table>
<tr>
<td>#UD</td>
<td>If VEX.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 Developers Manual</a> for anything serious.
</p></footer></body></html>