Abstract
A set of dedicated instructions for associative processing, including compares and searches, parallel arithmetic and system type operations has been implemented on the microprogrammable processors of the Erlangen General Purpose Array (EGPA) multiprocessor system using an unconventional data representation. Detailed measurements and the analysis of the runtimes for the microprogrammed operations as compared to optimized higher level language code show the different contributions of vertical migration and associativity to the total speedup obtained. An interpretation of the results is given.
Originalsprache | Englisch |
---|---|
Seiten (von - bis) | 257-261 |
Seitenumfang | 5 |
Fachzeitschrift | Microprocessing and Microprogramming |
Jahrgang | 8 |
Ausgabenummer | 3-5 |
DOIs | |
Publikationsstatus | Veröffentlicht - 1981 |
Extern publiziert | Ja |