Re: kickass optimizing compilers?

I have a similar story only what we did was take an assembly program
and re-code it into C retaining the algorithmic implementation.


I explained to our customer that we would prefer a well developed
piece of code that was as tight as possible. Our customer asked us
when we would need the code and could they review it and tighten it up
a little before they gave it to us. They assigned three people for 5
weeks to work on a slightly over 2K byte piece of embedded system
code. Our alpha compiler with the C equivalent generated an object 120
bytes less and used 5 bytes of RAM less. (It also ran faster) Our
customer was convinced.


Detained study later by both our customer and us showed most of the
saving was the attention to detail, full use of the instruction set by
the compiler and optimizations that the compiler used that would be
very high risk for the asm programmer.


Walter Banks


Robert Knighten wrote:


> There is also a matter of context. Years ago I worked on an Ada
> compiler that was required to generate code that was within 10% of the
> performance of the best hand coded assembler that our customer could
> produce.

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