In other words, there are data-structures (better than queues) to implement local names, since we know, at compile time, how many receptors there will be on a given local name. In particular, replicated names in Pict are always local.
this leads me to ask a question: we have all these fancy translations from higher-level languages into pi and then into assembler (pict etc)
f g
HLL ---------> pi -----------> assemblerand via conventional conventional compilers directly into assembler:
h
HLL -----------> assemblerwhat happens if we compare the output of g o f with h? do we essentially get the same code? or does translating via pi introduce inefficiencies? if the latter, are these inefficiences of a systematic nature?
martin