Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!columbia!rutgers!caip!segall
From: segall@caip.RUTGERS.EDU (Ed Segall)
Newsgroups: net.lang
Subject: Re: Optimized Jumps in High-level Languages
Message-ID: <3844@caip.RUTGERS.EDU>
Date: Sun, 2-Nov-86 18:54:49 EST
Article-I.D.: caip.3844
Posted: Sun Nov  2 18:54:49 1986
Date-Received: Tue, 4-Nov-86 01:29:15 EST
References: <1225@ncr-sd.UUCP>
Distribution: net.lang,net.lang.c
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 21

This sounds like trace scheduling, where the compiler uses information
about the likelihood of the direction of branches to optimize use
of pipelines, parallelism (e.g. in a microcoded machine), and perhaps
other performance enhancers. John Fisher, who was at Yale until a 
couple years ago, wrote a trace scheduling compiler for VLIW machines
(Very Long Instruction Word - i.e. parallel machines controlled with 
lots of bits, such as horizontally microprogrammed machines). He wrote
a book on it called something like "Bulldog - a Trace Scheduling
Compiler for VLIW Machines," which is published by MIT Press. There
are also Yale technical reports on the work. I've also seen a
reference to trace scheduling in VLSI design. I don't remember the
book, but I'm almost certain that it is a Springer-Verlag monograph,
and that it has an introduction Edsger Dijkstra (sp?). Or maybe the
intro was by Hoare. Anyway, it was somebody famous for being very
insistant about formality. I think it was Dijkstra, and I think the
work was done in a Northern European country, which would fit. How's
that for free association? 

Have fun,

Ed Segall