Re: Compiler back-ends [Q]

>Ben.Sloman@reading.ac.uk (Ben Sloman) writes:
>> `the back-ends of most production compilers are generated automatically'
>> True or false? On what grounds (examples/references appreciated)?
>> ...
>> [I'll eat my hat if it's true. -John]


In 95-10-114 cliffc@ami.sps.mot.com (Cliff Click) writes:
>[`lcc', `gcc', TI compilers, Motorola compilers.]


Not to forget the Greenhills compilers, which use a proprietary machine
description and have been ported to a variety of systems including x86
and KSR1.


FWIW (though this hardly counts as ``most''), I've heard a rumor that
SunSoft has, for several years now, tuned their schedulers using an
automated system rather like Henry Baker's ``Precise Scheduling Without
a Machine Model''. According to rumor, the scheduler executes a suite
of benchmarks, trying each benchmark a variety of scheduling options
and tuning to those options that provide the best measured performance.
They iterate through this several times across a variety of machines,
eventually selecting the option that provides the best overall speedups
without substantially hurting code quality on other Sun platforms. At
least that's the rumor I heard.


(Fortunately, John's hat is made of bread and the strings of Pasta.)


;-D oN ( The back-ends justify the means ) Pardo
--

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