Microprogrammed associative instructions

Abstract

The microprogrammed implementation of associative instructions on conventional microprogrammable computers with address-based memory-access is introduced as "vertical processing". The implementation on the processors of the EGPA-multiprocessor project is reported as well as runtime comparisons of the microprograms with equivalent HLL-programs. The contribution of the factors vertical migration and associativity to the observed overall speedup is described. Finally, a detailed analysis of the underlying microarchitecture and the handcoded microprograms is given in view of local and global compaction techniques.