14.06.2015 Views

Power ISA™ Version 2.03 - Power.org

Power ISA™ Version 2.03 - Power.org

Power ISA™ Version 2.03 - Power.org

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Version</strong> <strong>2.03</strong><br />

Vector Reciprocal Estimate<br />

Single-Precision<br />

VX-form<br />

Vector Reciprocal Square Root Estimate<br />

Single-Precision<br />

VX-form<br />

vrefp<br />

VRT,VRB<br />

vrsqrtefp<br />

VRT,VRB<br />

4 VRT /// VRB 266<br />

0 6 11 16 21 31<br />

do i=0 to 127 by 32<br />

VRT i:i+31 ReciprocalEstimateSP( (VRB) i:i+31 )<br />

For each vector element i from 0 to 3, do the following.<br />

The single-precision floating-point estimate of the<br />

reciprocal of single-precision floating-point element<br />

i in VRB is placed into word element i of<br />

VRT.<br />

Unless the single-precision floating-point result of computing<br />

the reciprocal of a value would be a zero, an<br />

infinity, or a QNaN, the estimate has a relative error in<br />

precision no greater than one part in 4096.<br />

Note that results may vary between implementations,<br />

and between different executions on the same implementation.<br />

The result for various special cases of the source value<br />

is given below.<br />

Value<br />

Result<br />

- Infinity -0<br />

- 0 - Infinity<br />

+0 + Infinity<br />

+Infinity +0<br />

NaN<br />

QNaN<br />

Special Registers Altered:<br />

None<br />

4 VRT /// VRB 330<br />

0 6 11 16 21 31<br />

do i=0 to 127 by 32<br />

VRT i:i+31 ReciprocalSquareRootEstimateSP(<br />

(VRB) i:i+31 )<br />

For each vector element i from 0 to 3, do the following.<br />

The single-precision floating-point estimate of the<br />

reciprocal of the square root of single-precision<br />

floating-point element i in VRB is placed into word<br />

element i of VRT.<br />

Let x be any single-precision floating-point value.<br />

Unless the single-precision floating-point result of computing<br />

the reciprocal of the square root of x would be a<br />

zero, an infinity, or a QNaN, the estimate has a relative<br />

error in precision no greater than one part in 4096.<br />

Note that results may vary between implementations,<br />

and between different executions on the same implementation.<br />

The result for various special cases of the source value<br />

is given below.<br />

Value<br />

Result<br />

- Infinity QNaN<br />

< 0 QNaN<br />

- 0 - Infinity<br />

+0 + Infinity<br />

+Infinity +0<br />

NaN<br />

QNaN<br />

Special Registers Altered:<br />

None<br />

192<br />

<strong>Power</strong> ISA -- Book I

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!