April 2003
Intermediate to advanced
576 pages
15h 13m
English
In Chapter 5, we compared two programs, DOTLOOP and DOTCLOOP, the latter of which used the Itanium loop count register. Here we show how the same basic code fragment can be analyzed for modulo scheduling.
VLIW processor designs did not attain commercial success, in part, because software had to be recompiled for each new implementation. This section demonstrates two different Itanium instruction schedules that develop two different EPIC solutions to the obstacles of implementation dependency that exist in pure VLIW architectural designs.
In this section, we develop a pipelined instruction schedule based upon an idealized, convenient fiction: We pretend that every ...