|Code generation for multiway branches email@example.com (2004-05-24)|
|From:||firstname.lastname@example.org (Steven Bosscher)|
|Date:||24 May 2004 00:31:52 -0400|
|Keywords:||code, optimize, comment|
|Posted-Date:||24 May 2004 00:31:52 EDT|
Does anyone know of a paper about code generation for multiway
branches using profile information?
I have found "Compiler code ceneration for Multiway branch statements
as a static search problem" by D.A. Spuler, but the algoriths
discussed in this paper and the ones it refers to all assume that all
cases are equally likely. That's exactly what I don't want :-)
Has anyone seen a paper that takes branch probabilities into account?
[Well, there was the Fortran FREQUENCY statement which they dropped
when they realized that programmers have no idea what the probabilities
Return to the
Search the comp.compilers archives again.