OpenGL Programming Guide: The Official Guide to Learning OpenGL, Version 4.3, Eighth Edition

Reduced-Precision Floating-Point Values

In addition to the normal 32-bit single-precision floating-point values you usually use when you declare a GLfloat in your application, OpenGL supports reduced-precision floating-point representations for storing data more compactly than its 32-bit representation. In many instances, your floating-point data may not require the entire dynamic range of a 32-bit float, and storing or processing data in a reduced-precision format may save memory and increase data transfer rates.

OpenGL supports three reduced-precision floating-point formats: 16-bit (signed) floating-point values, and 10- and 11-bit unsigned floating-point values. Table G-1 describes the bit layout of each representation, and the associated ...