Re: Processor specific optimisations

"Mickaël Pointier" <mpointie@eden-studios.fr> wrote:
[...]
: So, the question finaly is: Is it possible to apply the modern
: compiler optimisation strategies to this old processor and have a
: result that an experimented 6502 assembly coder would have a hard time
: to beat ?


That question resumes the never-ending thread of whether compilers can
ever match or exceed human experts in the quality of code they
generate. By way of analogy, one can pose the same question regarding
the playing of games -- as time goes by, computers are matching or
exceeding humans at more and more games. But some games are harder to
automate than others. Last time I checked, computers were approaching
humans at chess but are still far away at go.


Consider each architecture class a different game. Why do computers
approach the performance of humans on desktop and server architectures
but not on the 6502? I have three conjectures:


  1 Perhaps the 6502 has not received the same level of effort from
      automaters. (I believe it.)


  2 Perhaps the 6502 architecture (like go) is more difficult to
      atomate. (I don't believe it.)


  3 Perhaps highly speculative desktop and server architectures are
      overly tedious for humans, because good optimiztion for such
      architectures requires lots of book-keeping and computation.
      (I believe it.)


Today, there is a lot of compiler research that is oriented toward
embedded systems, but it seems to be mostly in support of new DSP and
pixel processing architectures.


Tom Payne

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