Re: State of the art optimisations

Tom Crick <>
Thu, 18 Sep 2008 01:23:19 +0100

          From comp.compilers

Related articles
State of the art optimisations (Tom Crick) (2008-09-08)
Re: State of the art optimisations (James Harris) (2008-09-10)
Re: State of the art optimisations (cr88192) (2008-09-11)
Re: State of the art optimisations (Anton Lokhmotov) (2008-09-11)
Re: State of the art optimisations (Tom Crick) (2008-09-18)
Re: State of the art optimisations (Tom Crick) (2008-09-18)
| List of all articles for this month |

From: Tom Crick <>
Newsgroups: comp.compilers
Date: Thu, 18 Sep 2008 01:23:19 +0100
Organization: University of Bath
References: 08-09-044 08-09-055
Keywords: optimize
Posted-Date: 18 Sep 2008 18:05:28 EDT

>> Following on from the recent discussions about the "State of the Art" of
>> compiler technologies, what would you all regard as the most important
>> developments in compiler optimisations over the past ten years?
>> Do you think that is it the development of intermediate forms such as
>> SSA, which has enabled a greater range of optimisations, or are there
>> any specific optimisations that stand out as key developments?
> IMO, there haven't really been any major/drastic developments, rather lots
> of micro-level tweaking and optimizations.
> granted, SSA does add some, but the effect is by no means drastic (and IMO
> more in terms of allowing more flexibility in terms of low-level code
> generation rather than in terms of performance).
> many other changes are more in terms of the underlying architectures than
> the compiler structure.
> now, some people may disagree, but this is how it looks to me...

I agree that the improvements of late have been more incremental than
paradigm breaking. Due to the increasing complexity of the
mathematical optimisations, it raises that question in which order
should optimisations be done (e.g. "premature optimisation is the root
of all evil") or do early local optimisations prevent later
higher-level optimisations, etc.

Code generation has improved massively, but it raises the question of
how effective are modern compilers at optimising code? And how can we
measure this?


Post a followup to this message

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