gonioscopy 发表于 2025-3-26 21:37:53

http://reply.papertrans.cn/89/8818/881776/881776_31.png

PANEL 发表于 2025-3-27 02:22:52

The Cray-1,were employed with the CDC 7600. STACKLIB, a collection of subroutines which processed arrays of data, was developed to permit the CDC 7600 to operate at maximum speed. Utilizing tightly coded loops to process vectors of data suggested the development of a hardware implementation.

后退 发表于 2025-3-27 09:16:59

The IBM 360-91,he IBM 360-91 and 95 central processors employed a highly sophisticated look-ahead, pipeline/overlap system for improving performance in a manner transparent to the programmer. Many of the features of these systems, such as forwarding, combining, and the floating-point multiply strategy, were taken directly from the IBM Stretch.

myopia 发表于 2025-3-27 11:20:05

http://reply.papertrans.cn/89/8818/881776/881776_34.png

eczema 发表于 2025-3-27 15:11:43

http://reply.papertrans.cn/89/8818/881776/881776_35.png

费解 发表于 2025-3-27 20:20:15

The CDC 6600,lizing a novel approach to high density packaging and cooling, the system contained 400,000 silicon transistors. The result was a highly reliable system which became the standard of performance against which other scientific computers would be measured. As Thornton points out, the CDC 6600 b

半身雕像 发表于 2025-3-27 23:43:11

The IBM 360-91,models of the series share a common instruction set architecture and are compatible. The technology used to implement different models varies widely, as does the architectural approach. Low end, small machines have narrow data paths (8-bits wide), sequential functional units, and slow components. Hi

饮料 发表于 2025-3-28 02:08:35

http://reply.papertrans.cn/89/8818/881776/881776_38.png

生命层 发表于 2025-3-28 07:59:53

http://reply.papertrans.cn/89/8818/881776/881776_39.png

Limousine 发表于 2025-3-28 11:12:17

The CYBER 205,olved from experience with the STAR-100. Scalar computations on the STAR-100 were inefficient due to the long latency of the vector pipelines. A series of scalar floating-point additions was ten times slower than the same number of additions performed in vector mode. On the CYBER-205, register opera
页: 1 2 3 [4] 5
查看完整版本: Titlebook: Supercomputer Architecture; Paul B. Schneck Book 1987 Kluwer Academic Publishers 1987 algorithms.computer.computer architecture.processor.