This talk will outline work conducted in my group (in Montreal) during the last few years on the design of portable and efficient parallel functional languages. It offers a personal view of the tradeoffs forced upon language designers by the use of massively-parallel architectures. Our view of the area is based on the opposition between special-purpose and generalpurpose programming. This duality is mirrored by the semantic notions of macroscopic (or dataparallel) and microscopic (or communicationexplicit) parallel languages. Many research problems in parallel programming can be seen as a search for theoretical and software foundations of hybrid systems which will balance the relative advantages of general-purpose and special-purpose programming: portability and efficiency. In accordance with this view, we are engaged in two complementary activities. One is centered on a microscopic functional language called DPML. Another has been on a macroscopic language called MOA and now a semanti...