blob: 74988979277a662ba2155ed69bc415316c115da1 [file] [log] [blame] [raw]
<!DOCTYPE html>
<html>
<head>
<meta charset="UTF-8">
<link href="style.css" type="text/css" rel="stylesheet">
<title>RSQRTSS—Compute Reciprocal of Square Root of Scalar Single-Precision Floating-Point Value </title></head>
<body>
<h1>RSQRTSS—Compute Reciprocal of Square Root of Scalar Single-Precision Floating-Point Value</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>
<p>F3 0F 52 /<em>r</em></p>
<p>RSQRTSS<em> xmm1</em>,<em> xmm2/m32</em></p></td>
<td>RM</td>
<td>V/V</td>
<td>SSE</td>
<td>Computes the approximate reciprocal of the square root of the low single-precision floating-point value in <em>xmm2/m32</em> and stores the results in <em>xmm1</em>.</td></tr>
<tr>
<td>VEX.NDS.LIG.F3.0F.WIG 52 /r VRSQRTSS <em>xmm1, xmm2, xmm3/m32</em></td>
<td>RVM</td>
<td>V/V</td>
<td>AVX</td>
<td>Computes the approximate reciprocal of the square root of the low single precision floating-point value in <em>xmm3/m32</em> and stores the results in <em>xmm1</em>. Also, upper single precision floating-point values (bits[127:32]) from <em>xmm2</em> are copied to <em>xmm1</em>[127:32].</td></tr></table>
<h3>Instruction Operand Encoding</h3>
<table>
<tr>
<td>Op/En</td>
<td>Operand 1</td>
<td>Operand 2</td>
<td>Operand 3</td>
<td>Operand 4</td></tr>
<tr>
<td>RM</td>
<td>ModRM:reg (w)</td>
<td>ModRM:r/m (r)</td>
<td>NA</td>
<td>NA</td></tr>
<tr>
<td>RVM</td>
<td>ModRM:reg (w)</td>
<td>VEX.vvvv (r)</td>
<td>ModRM:r/m (r)</td>
<td>NA</td></tr></table>
<h2>Description</h2>
<p>Computes an approximate reciprocal of the square root of the low single-precision floating-point value in the source operand (second operand) stores the single-precision floating-point result in the destination operand. The source operand can be an XMM register or a 32-bit memory location. The destination operand is an XMM register. The three high-order doublewords of the destination operand remain unchanged. See Figure 10-6 in the <em>Intel® 64 and IA-32 Architectures Software Developer’s Manual, Volume 1</em>, for an illustration of a scalar 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 RSQRTSS 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). When a source value is a negative value (other than −0.0), a floating-point indefinite is returned. 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 first source operand and the destination operand are the same. Bits (VLMAX-1:32) of the corresponding YMM destination register remain unchanged.</p>
<p>VEX.128 encoded version: Bits (VLMAX-1:128) of the destination YMM register are zeroed.</p>
<h2>Operation</h2>
<p><strong>RSQRTSS (128-bit Legacy SSE version)</strong></p>
<pre>DEST[31:0] ← APPROXIMATE(1/SQRT(SRC2[31:0]))
DEST[VLMAX-1:32] (Unmodified)</pre>
<p><strong>VRSQRTSS (VEX.128 encoded version)</strong></p>
<pre>DEST[31:0] ← APPROXIMATE(1/SQRT(SRC2[31:0]))
DEST[127:32] ← SRC1[127:32]
DEST[VLMAX-1:128] ← 0</pre>
<h2>Intel C/C++ Compiler Intrinsic Equivalent</h2>
<p>RSQRTSS:</p>
<p>__m128 _mm_rsqrt_ss(__m128 a)</p>
<h2>SIMD Floating-Point Exceptions</h2>
<p>None.</p>
<h2>Other Exceptions</h2>
<p>See Exceptions Type 5.</p></body></html>