ABSTRACT
COPACOBANA ............................................................... 239 11.4 COPACOBANA 5000 ................................................................................. 242
11.4.1 Direction toward New Applications ........................................... 242 11.4.2 Requirements .................................................................................. 242 11.4.3 Architecture of COPACOBANA 5000 ......................................... 243
11.4.3.1 Bus Concept and Backplane .......................................... 243 11.4.3.2 FPGA Module .................................................................. 244 11.4.3.3 Interface Controller ......................................................... 246 11.4.3.4 Power Supply and Cooling Mechanism ...................... 246 11.4.3.5 Application Development .............................................. 247
11.5 Applications in Bioinformatics ................................................................ 248 11.5.1 Sequence Alignment ..................................................................... 249
11.5.1.1 Smith-Waterman Alignment ........................................ 249 11.5.1.2 Hardware Implementation ............................................250 11.5.1.3 Performance on COPACOBANA 5000 ......................... 251
11.1.1 History of Complexity
Several complexity measures have been used to evaluate the quality of algorithms running of different computer architectures. The dominating measures for sequential computers used to be the required computing time T and memory space S [1]. With the idea of parallel computations the number N of processors became an additional important complexity measure. Over the years, since about 1980, compute-intensive parallel algorithms have been implemented as full-custom or semicustom very-large-scale integration (VLSI) chips. Since then the chip area A became one of the important measures for such implementations. Combinations like AT and AT2 were widely used [2] to evaluate the quality of VLSI algorithms. Lower bounds for the time complexity of parallel algorithms could be proven by means of these combinations [3]. In the last 10 years another complexity measure became dominant: power consumption P. As heat dissipation is one of the major problems of modern high-performance computer systems, the power is often the limiting parameter for computational performance [4].
