Descripción
|
|
---|---|
In this work we present an FPGA implementation of a single-precision °oating-point arith- metic powering unit. Our powering unit is based on an indirect method that transforms xy into a chain of operations involving a logarithm, a multiplication, an exponential function and dedicated logic for the case of a negative base. This approach allows to use the full input range for the base and exponent without limiting the range of the exponent as in direct methods. A tailored hardware implementation is exploited to increase the accuracy of the unit reducing the relative errors of the operations while high performance is obtained taking advantage of the FPGA capabilities for parallel architectures. A careful design of the pipeline stages of the involved operators allows a clock cycle of 201.3 MHz on a Xilinx Virtex-4 FPGA. | |
Internacional
|
Si |
Nombre congreso
|
8th Conference on Real Numbers and Computers |
Tipo de participación
|
960 |
Lugar del congreso
|
Santiago de Compostela (Spain) |
Revisores
|
Si |
ISBN o ISSN
|
978-84-691-4381-0 |
DOI
|
|
Fecha inicio congreso
|
07/07/2008 |
Fecha fin congreso
|
09/07/2008 |
Desde la página
|
68 |
Hasta la página
|
77 |
Título de las actas
|
Proceedings 8th Conference on Real Numbers and Computers (RNC8) |