Compressão de codigo de programa usando fatoração de operandos

AUTOR(ES)
DATA DE PUBLICAÇÃO

1999

RESUMO

The increasing use of embedded systems is a clear trend in the telecommunication, multimedia and consumer electronics industry. Because these systems are designed for high-volume market, a cost reduction can have a large impact in the final price of the product. Driven by the need to reduce cost, these systems are implemented by putting together a core processor, an ASIC (Application Specific Integrated Circuits) and a program/data memory into a single chip. These systems are known as SOC - System-On-a-Chip. As embedded systems are becoming more complex, the size of embedded programs is growing considerably large. The results are systems in which program memories account for the largest share of total die area, more than the area of the others modules. Thus, minimizing program size has become an important part of the design effort of these systems. We propose a program code compression technique called operand factorization. The key idea of operand factorization is the separation of program expression trees into sequences of operators and operands (registers and immediates). Using operand factorization we show that tree and operand sequences have exponential frequency distributions. A set of experiments is performed to determine the best encoding technique that explores this feature. The experimental results, using the studied techniques, show an average compression ratio of 43% for SPEC CINT95 programs, running on a MIPS R2000 processor and 60% for a set embedded programs running on the TMS320C25... Note: The complete abstract is available with the full electronic digital thesis or dissertations.

ASSUNTO(S)

circuitos integrados - integração em escala muito ampla sistemas embutidos de computador

ACESSO AO ARTIGO

Documentos Relacionados