Experience with Pipelined Multiple Instruction Streams
Article Abstract:
The Heterogeneous Element Processor (HEP) computer consists of one or more pipelined multiple instruction-multiple data (MIMD) Process Execution Modules (PEM) that share memory. The memory of the HEP system is program memory, register memory, constant memory, and data memory. Fortran is used in a modified form with the system. The machine can minimize turnaround time by partitioning work on function or on data across multiple instruction streams. Several categories of applications programs are possible. 'Dusty deck' Fortran programs are originally written for a sequential processor and require mechanical parallelization for optimization. The vector program can compare MIMD performance with single instruction-MD performance. A distinct class of programs is nonvectorizable, and MIMD parallelization is more natural. Graphs show the pipelines in the machine are fully utilized for twenty or more processes. Diagrams illustrate MIMD and SIMD architectures as well as the principle pipelines in the HEP.
Publication Name: Proceedings of the IEEE
Subject: Electronics
ISSN: 0018-9219
Year: 1984
User Contributions:
Comment about this article or add new information about this topic:
Reduced instruction set computer architecture
Article Abstract:
The general trend in CPU architecture has been toward complexity, with larger instruction sets, multiple addressing modes, and specialized registers. The relatively new RISC CPUs (characterized by one instruction per cycle, register-to-register operations, simple address modes, and simple instruction formats) are challenging this design philosophy with numerous or optimized register usage, reduced instruction sets, and pipelining. Performance benefits include: more efficient compilers; no microcode usage; more effective pipelining; and improved interrupt response. Examples of RISC systems are the experimental Berkeley RISC and the commercial MIPS R2000.
Publication Name: Proceedings of the IEEE
Subject: Electronics
ISSN: 0018-9219
Year: 1988
User Contributions:
Comment about this article or add new information about this topic:
Parallel bit-level pipelined VLSI designs for high-speed signal processing
Article Abstract:
Six bit-level pipelined VLSI designs for high-speed signal processing are explored. The issues in designing such fully pipelined architectures are: clock skew; clock distribution networks; buffering; timing simulation; area overhead from pipelining; and testing. Designs now fabricated in CMOS technology include: a multiplier; Finite Impulse Response (FIR) filter block; and a multi-channel, multiply-accumulate-add chip. Test results for functionality and speed are given. The chips presented are examples of building block FIR implementing algorithms, in which throughput and real-time operation are major concerns and latency is not a critical factor.
Publication Name: Proceedings of the IEEE
Subject: Electronics
ISSN: 0018-9219
Year: 1987
User Contributions:
Comment about this article or add new information about this topic:
- Abstracts: Implementation of digital filtering algorithms using pipelined vector processors. Technological Design Considerations for Monolithic MOS Switched-Capacitor Filtering Systems
- Abstracts: A tutorial on hidden Markov models and selected applications in speech recognition. A Tutorial on Computer-Aided Analysis of Switched-Capacitor Circuits
- Abstracts: Computations on the Massively Parallel Processor at the Goddard Space Flight Center. Three-Dimensional Wave Equation Computations on Vector Computers
- Abstracts: The Impact of Supercomputers on IC Technology Development and Design. part 2 Array processor supercomputers