Class APPLEBiasedFixedPointImageFormats


  • public final class APPLEBiasedFixedPointImageFormats
    extends java.lang.Object
    Native bindings to the APPLE_biased_fixed_point_image_formats extension.

    This selector may be passed to CreateImage2D in the cl_image_format.image_channel_data_type field. It defines a biased signed 1.14 fixed point storage format, with range [-1, 3). The conversion from float to this fixed point format is defined as follows:

    
     ushort float_to_sfixed14( float x ){
         int i = convert_int_sat_rte( x * 0x1.0p14f );         // scale [-1, 3.0) to [-16384, 3*16384), round to nearest integer
         i = add_sat( i, 0x4000 );                             // apply bias, to convert to [0, 65535) range
         return convert_ushort_sat(i);                         // clamp to destination size
     }

    The inverse conversion is the reverse process. The formats are currently only available on the CPU with the RGBA channel layout.

    • Method Summary

      • Methods inherited from class java.lang.Object

        equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait