Re: Cost/Benefit of compiler optimization techniques?

> I am interested in finding out the "most useful" optimizations ...


Some papers, in no special order


Effectiveness of a Machine-Level, Global Optimizer
Johnson and Miller
Sigplan '86, Compiler Construction Conference
[studies based on HP's Spectrum compiler]


Measurement of Program Improvement Algorithms
Cocke and Markstein
IFIP '80
[early measurements of the PL.8 compiler]


Simple Code Optimizations
Hanson
Software -- Practice and Experience, 1983
[argues for, and present 3 simple, effective optimizations]


A Portable Optimizing Compiler for Modula-2
Powell
Siplan '84, Compiler Construction Conference
[specifically tried for "best, simple" optimizations]


In general though, it's difficult to talk about the effectiveness and
expense of specific optimizations in isolation. Good optimizers do many
optimizations, counting in the combined effect to do a good job. The
implementation costs are especially difficult. For example, a strength
reducer alone might cost X amount. But, if you've got a dead code
eliminator, the strength reducer is simpler and can make use of some
common code (data flow analysis, ...).


For some exceptionally profitable optimizations, see


Improving Register Allocation for Subscripted Variables
Callahan, Carr, Kennedy
Sigplan '90, Programmling Language Design and Implementation


Preston
--

Return to the comp.compilers page.
Search the comp.compilers archives again.