We have developed a technology – a mathematical tool which implements a new format of the real numbers representation and algorithms of processing them in general purpose multicore processors: a multi-module modular numerical system (the residue number system) is used for representation of number mantissas, while the number exponents are represented in a positional form.
This method is utilized and implemented for the first time ever. It makes it possible to perform arithmetic operations with operands – mantissas with paralleling to signed-digits.
The method can be utilized for solving high time complexity problems which are quite demanding to precision of the solutions. In this case, the time spent on direct and inverse transformation is incomparably little relative to the computation time. A great number of problems from various areas of science and technology can be addressed using this method.
The examples of such areas of application are: numerical analytics, physics, computational finance, computational fluid dynamic, structural mechanics calculations, automated engineering systems, astronomy, electronic design automation, modeling, physics simulation, weather and other types of simulations.
Our technology opens up new possibilities for acceleration and increasing accuracy of your computations, in addition to helping to avoid errors in calculations and to lower requirements to hardware. As a result, you will have powerful tools for developing your own projects or products which will help you save your budgets.
We have developed a new technology of high-precision numbers representation in computer systems. This know-how helps to make high-accuracy computations (4x double or higher) without any significant loss of productivity on any general purpose multicore processors.
xPrecision HPDP library can be easily integrated into your scientific software, hardware or other programmes aimed to provide highly accurate computations with 4x double precision or higher.