| |
Sections
|
 DOCUMENT (160 pages). This file explains all my work.
REPORT (7 pages). It is a miserable bureaucrat copy+paste of the first one.
MOTIVATIONS: In june 2005 Barcelona emerged at the TOP500 ranking with MareNostrum. The third position was the best introduction for the new BSC. This year MareNostrum is eleventh. At european level it has descended from the first to the third. This is the inexorable descendent way for all computers, and that is known. The will to make short this situation brings some movements in computer centers, as planning the life for the actual processors. CBEA get in scene: a new architecture multi-core from IBM with a dessign that breaks some traditional schemes. The study of this new architecture may to resolve if the more performance justifies the economical cost. Will CBEA be a good architecture for MareNostrum? This work points to this: after a deep study of CBEA and its programming techniques, I code the Smith-Waterman algorithm to prove his superior performance. Main resources.  This file contains both sequencial and Altivec source code of FASTA.
This tarball contains the Altivec source of ssearch34 compilable with GCC.
IBM docs. Introduction to the Cell multiprocessor
CBE Architecture
CBE Handbook
CBE Linux Reference Implementation Application Binary Interface Specification
CBE Registers
SPU Application Binary Interface Specification
SPU Assembly Language Specification
SPU C-C++ Language Extensions
SPU-ISA
Altivec Programming Interface Manual
Related papers. A Model of evolutionary change in proteins (1978)
Smith-Waterman (1981)
Amino acid substitution matrices from protein blocks (1992)
Protein sequence comparison and Protein evolution (2001)
 FASTA algorithm
|
Copyleft 2o12 ~ www.akamc2.cat ~
miquel
akamc2