-fno-strict-aliasing performance

[duchier at ps.uni-sb.de]

Konstantin Popov <kost at sics.se> writes:

> so '-fstrict-aliasing' just *happens* to produce good code on x86
> (and possibly some other platforms) right now.

I completely understand and agree with the reasons why
-fstrict-aliasing should _in_principle_ be avoided in the current
state of affairs... but "just happens to produce good code" is also
why we make a careful selection of compiler options, sensitive to
compiler, platform, and their versions.  It's a pragmatic tradeoff.

I do not care deeply about whether aliasing will be turned off on all
platforms for the next release, but it seems unnecessary and
overcautious.  That's just my opinion, of course.

Cheers,

-- 
Denys Duchier - Équipe Calligramme - LORIA, Nancy, France
AIM: duchierdenys