Re: optimization and compilation speed

In the various talks that Bill Wulf used to give about the PQCC
(Production Quality Compiler Compiler) technology, he would talk about
the optimizer running in "negative time", due to the reduction in the
size of the IL, which was both (a) verbose, and (b) communicated
between compiler phases (of which there were many in PQCC) in ASCII.
This translation to/from internal binary to ASCII was itself an
expensive process, so there was a direct correlation between IL file
size and compile time. The papers date back to 1979-1980.


-AndyJ


Saumya K. Debray <debray@CS.Arizona.EDU> wrote in message
news:99-09-044@comp.compilers...
> I have this impression that various people have observed that some
> amount of (peephole) optimization can lead to an overall reduction in
> compilation time, compared to no optimization at all, despite the
> additional time spent in doing the optimization, because the reduced
> number of instructions reaching later phases means less time is spent
> in those phases. I thought this observation was due to Wulf in his
> book about the Bliss-11 compiler, but can't seem to find it there.

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