Implementation of float-float operators on graphics hardware

TL;DR

This paper presents a method to emulate 44-bit floating-point operations on GPUs, enabling higher precision computations beyond standard single precision, with an analysis of performance and accuracy.

Contribution

It introduces a novel implementation of 44-bit float-float operators on graphics hardware, expanding GPU computational capabilities for higher precision tasks.

Findings

The implementation achieves acceptable performance levels.

The emulation maintains high accuracy for 44-bit floating-point operations.

The approach enables higher precision computations on existing GPU hardware.

Abstract

The Graphic Processing Unit (GPU) has evolved into a powerful and flexible processor. The latest graphic processors provide fully programmable vertex and pixel processing units that support vector operations up to single floating-point precision. This computational power is now being used for general-purpose computations. However, some applications require higher precision than single precision. This paper describes the emulation of a 44-bit floating-point number format and its corresponding operations. An implementation is presented along with performance and accuracy results.