2.2.1. Half-precision floating-point value conversion

The half-precision floating-point value conversion adds support for both IEEE and the common graphic representation, referred to as alternative-half-precision representation, of 16-bit floating-point values. This provides a smaller memory footprint for applications requiring large numbers of lower-precision floating-point values to be stored, while avoiding the overhead of conversion in software.

Additional VFP and Advanced SIMD instructions enable conversion of both individual values and vectors of values to and from single-precision floating-point representation. These values can then be processed using the rest of the VFP and Advanced SIMD instructions.

See Floating-Point Status and Control Register for how to select IEEE or alternative half-precision modes.

Copyright © 2008-2010 ARM. All rights reserved.ARM DDI 0409F
Non-ConfidentialID050110