Re: Compiler Optimisation?

[Marin David Condic <condicma@pwfl.com>]

Iain Bate wrote:
> 
> I am looking for some specific information on compiler optimisation
> and I was wondering if anyone could help. Basically, I would like to
> get some idea how much faster code executes due to optimisation. I am
> looking for the information at two levels:
> 
> 1. How much faster code executes due to the overall optimisation
> process?
> 2. How much faster code executes due to specific optimisation stages,
> e.g. the speed-up due to the peephole stage?
> 
If you get an answer to this, it is almost certainly going to be
compiler dependent and probably misleading. Consider that if a compiler
generates really crappy code in non-optimized mode, the optimization
could show a huge delta in performance. Whereas another compiler may
generate very tight code in non-optimized mode and hence get a much
smaller delta in optimization. So almost any quote you get ("Our
optimizer results in 50% improvement in performance! - by taking out all
the no-ops we stuck in there in the code generation...") is bound to be
misleading.

Your best bet is to get a representative sample of code for your
particular applications and run some timing studies comparing one
compiler to another at wound-full-out optimization. Almost anything else
is not going to provide you with much useful information.

MDC

-- 
Marin D. Condic
Real Time & Embedded Systems, Propulsion Systems Analysis
United Technologies, Pratt & Whitney, Large Military Engines
M/S 731-95, P.O.B. 109600, West Palm Beach, FL, 33410-9600
Ph: 561.796.8997         Fx: 561.796.4669

"The reasonable man adapts himself to the world; the unreasonable one
persists in trying to adapt the world to himself. Therefore all progress
depends on the unreasonable man."

        --  G.B. Shaw