🚀 go-pugleaf

On 3 Nov 90 04:16:26 GMT, grover@brahmand.Eng.Sun.COM (Vinod Grover) said:

grover> In article <PCG.90Nov2231021@athene.cs.aber.ac.uk>
grover> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes:

pcg> The burden of proof is on proponents of aggressive optimization,
pcg> becasue they have to overcome the obvious observation that anything
pcg> that enlarges or complicates a program will tend to decrease its
pcg> reliability.

grover> I fail to see how an optimiser can complicate a program if it
grover> performs its transfromation at an intermediate level. The source
grover> program written by the programmer is *not modified* in any way.

Sorry if It was not clear from what I wrote that the program I was
referring to was the *compiler*.

grover> The translation to a complex intermediate form, if correct, does
grover> not affect the reliability of the source program.

I intended to say that the code that effects the transformation adds to
the size and complexity of the compiler and therefore affects the
compiler's reliability. Each individual transformation may look simple
and safe, but once you consider that you need code also to do the
analysis that makes possible and drives the transformation process and
you also need to consider he interactions between tranformations, which
are difficult to model in languages that have not been designed with
transsformations in mind... You have a large amount of hairy code in the
compiler's bowels, and this is baaaad.
--
Piercarlo "Peter" Grandi           | ARPA: pcg%uk.ac.aber.cs@nsfnet-relay.ac.uk
Dept of CS, UCW Aberystwyth        | UUCP: ...!mcsun!ukc!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk