No label defined (Q3813282)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 4082955
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 4082955 |
Statements
1988
0 references
parallel reduction
0 references
graph reduction
0 references
cell token flow model
0 references
0.7889764308929443
0 references
[For the entire collection see Zbl 0639.00038.] The theoretically attractive reduction model gives rise to challanging problems when trying to find an appropriate machine architecture. Furthermore, while the implementation of sequential reduction is somewhat easy to handle (stack control mechanism), parallel reduction machine implementation depends on solving the problem of reduction control in parallel execution (in the first place, triggering/delaying a specified reduction). This paper offers a model (G-reduction) and an implementation of parallel graph reduction machine, based on the data flow computing scheme. A parallel computation mechanism of lambda-calculus is introduced, using a graph reduction model. The data flow computing scheme can carry on usual (object) data, as well as the data flow graph itself (meta-data), by means of object operators and several meta-operators. The 8 meta- operations support high-level facilities such as partial function application (leading to a higher-order function mechanism) and delay-eval computation mechanism (the basis for parallel reduction control strategy); nevertheless, they increase the implementation cost. The machine architecture is designed on a multi-thread control flow concept. Efficiency is gaind by means of elimination of redundant data flow control. In the implementation, a cell token flow model is used instead of reducing the data flow graph directly. The processor is designed as a circular pipeline system; highly parallel MIMD computation is achieved due also to the packet communication network of processors.
0 references