Re: latest trends in compiler optimization research?

Ray Dillinger <bear@sonic.net> wrote:
>
>I don't know what the latest research work or trends are, but
>I can say what some of the biggest problems that compiler
>writers need to solve in modern architectures are.
>
>Effective use of multi-core CPU's,
>Effective use of multi-CPU architectures,
>Effective use of shared memory and cache in multi-CPU architectures,
>Effective use of per-CPU memory or per-core memory in multi-CPU
> architectures.


I don't think these are problems that can be tackled just by compiler
writers: they are more in the realm of language designers or system
architects (depending on whether you put the features in the language
or the library). Promising approaches include:


* message-passing, such as in Erlang
* software transactional memory
* data-parallel array ops


all of which require the right kind of application design to work.
Of course the compiler writer has to make them work *well*, but that's
out of scope if the compiler implements C.


Tony.
--
f.a.n.finch <dot@dotat.at> http://dotat.at/

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