Fast and accurate digital signal processing realized with GPGPU technology

An idea of the so-called quasi-maximum accuracy computations for improvement of precision of the floating-point digital signal processing with graphic processing units (GPUs) is presented in this paper. In the presented approach, the increase of the precision of computations does not need any increase of the length of the data words. Special attention has been paid to efficiency and precision of computations. The maximum accuracy has been analyzed and technically realized with no additional costs in hardware and computation time. Streszczenie. W artykule zaproponowano wykorzystanie obliczen zmiennoprzecinkowych o quasi-maksymalnej dokladności do cyfrowego przetwarzania sygnalow za pomocą procesorow graficznych (GPU). W zaprezentowanym rozwiązaniu, zwiekszenie precyzji obliczen nie wymaga zwiekszenia dlugości slow danych. Szczegolną uwage zwrocono na efektywnośc przeprowadzanych obliczen. Idea uzycia obliczen o maksymalnej precyzji zostala technicznie zrealizowana bez dodatkowych kosztow w sprzecie i w czasie obliczeniowym. (Szybkie i dokladne cyfrowe przetwarzanie sygnalow z wykorzystaniem technologii GPGPU) Slowa kluczowe: liczby zmiennoprzecinkowe, GPGPU, CUDA, maksymalna precyzja obliczen, dwa akumulatory, generator cyfrowy

[1]  A. Dabrowski,et al.  The ability to use parallel programing on graphics processor unit in digital signal processing , 2008, New Trends in Audio and Video / Signal Processing Algorithms, Architectures, Arrangements, and Applications SPA 2008.

[2]  Adam Dąbrowski.  Multirate and Multiphase Switched-capacitor Circuits , 1996 .

[3]  Anselmo Lastra,et al.  Physically-based visual simulation on graphics hardware , 2002, HWWS '02.

[4]  Willard L. Miranker,et al.  Computer arithmetic in theory and practice , 1981, Computer science and applied mathematics.

[5]  Daniel L. Rosenband,et al.  A design case study: CPU vs. GPGPU vs. FPGA , 2009, 2009 7th IEEE/ACM International Conference on Formal Methods and Models for Co-Design.

[6]  Adam Dąbrowski,et al.  Extended precision method for accumulation of floating-point numbers in digital signal processors , 2007 .

[7]  Gerd Bohlender,et al.  Floating-Point Computation of Functions with Maximum Accuracy , 1975, IEEE Transactions on Computers.

[8]  A. Dabrowski.  Transmission of effective pseudopower in multirate signal processing , 1989, International Conference on Acoustics, Speech, and Signal Processing,.

[9]  Gerd Bohlender,et al.  Floating-point computation of functions with maximum accuracy , 1977, 1975 IEEE 3rd Symposium on Computer Arithmetic (ARITH).

[10]  Tarek S. Abdelrahman,et al.  hiCUDA: High-Level GPGPU Programming , 2011, IEEE Transactions on Parallel and Distributed Systems.

[11]  Martin Rumpf,et al.  Using Graphics Cards for Quantized FEM Computations , 2001, VIIP.