Short, tongue-in-cheek essay by Edsger Dijkstra on old problems in programming, many of which are embarrassingly still with us: criticizes APL, BASIC, COBOL, FORTRAN, PL/I, natural language programming.
Sustainable memory bandwidth discussion, with results on a wide variety of computer systems. From Macs and PCs to the most current and recent workstations, right up to Cray supercomputers. From the Department of Computer Science at Virginia.
Its philosophy: build compilers from parts, which might include front ends, back ends, optimizers, and the glue that holds all these pieces together. You might even generate parts automatically from compact specifications.