Re: alignment of data-types

"Mark McIntyre" <markmcintyre@spamcop.net> wrote in message
news:u6i04vgrps4ar212a1s8knie78jm193e63@4ax.com...
> On Tue, 4 Feb 2003 14:22:00 +0100, in comp.lang.c , "R.A. Scheltema"
> <r.a.scheltema@dacolian.nl> wrote:
>
> >I'm currently working on a project which needs a lot of optimazation.
I've
> >been reading some stuff on aligning data-types, which would mainly be
> >important for avoiding cach-splits and some other preformance
optimizations.
>
> The first rule of optimisation is - don't do it yet.
> The second rule is - optimise the ALGORITHM first not the code.
>
> The sort of stuff you're talking about is very low level and should be
> done automatically by any good optimising compiler.
>
> You have little or no chance of being better at it than the guys who
> wrote it, at least not till you're a guru programmer, by which stage
> you won't need to ask us for advice....


Should, but not always true.


It is common for x86 compilers to allocate double variables on 4 byte
boundaries. This was optimal on the 386 and 486, but not on pentium
and later processors. The difference can be large.


I now added comp.compilers because I didn't think of a better place to
discuss the properties of a large number of compilers on different
systems. It might be nice to know which allocate on 4 and which on 8
byte boundaries.


-- glen

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