Re: WANTED: Rating of common optimization routines.

dlmoore@ix.netcom.com writes:
> Which optimizations produce how much will depend upon the program and
> the machine.


And also what you had to start with. For instance, saying "how much
does register allocation improve things" is kind of nonsense, because
you have to do *some* kind of register allocation just to get a
runnable program. Presumably, the requester meant something like
"global register allocation", but even then it depends on how good
your local register allocator does.


Beyond that, optimizations often become interrelated in many
compilers. For instance, some optimizers will perform strength
reduction in a way that relies on copy propagation and dead code
elimination to "clean up" afterwards. That makes it hard to tell how
much the optimizations contribute individually.


There are simply too many algorithms for each optimization, each one
at least somewhat dependent on the intermediate form used, the
machine, the language, the subject program, and other parts of the
compiler to be able to measure the effect of one optimization in
isolation.


--
Bill Leonard
Harris Computer Systems Corporation
2101 W. Cypress Creek Road
Fort Lauderdale, FL 33309
Bill.Leonard@mail.hcsc.com
--

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