SAMBA has been designed at IRISA, in the API team. The API project is interested in special-purpose VLSI architectures, especially parallel architectures.This research is covered by designing and building architectures for certain application domains, such as biological sequence comparison, image compression, data compression, coding and cryptography,
The design of SAMBA was decided in september 1993. The main decision was to choose the technology of the systolic array. The choice was between a programmable or a fully dedicated architecture. It was decided to build a systolic array with fully dedicated processors, as the result would be about 10 times faster than a programmable version, and there would be no need to design a programming environment - a task which usually requires a lot of effort. Moreover, the algorithms to be implemented were very stables, and a parameterized architecture would make it possible to execute all the commonly used algorithms.
The design effort was estimated to 11 person/months: 1 month for functional specification and simulation, one month for architectural specification and simulation, 2 months for architectural synthesis, 2 month for chip testing, 0.5 month for board design, 4 month for interface design, and 0.5 month for final integration.
The SAMBA prototype works since the end of 1995.