C or C++ programming mutual assistance group(s) ?

Lennart Sorensen lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org
Sat Aug 21 01:06:49 UTC 2004


On Fri, Aug 20, 2004 at 07:35:26PM +0000, James McIntosh wrote:
> My brother is on the team at IBM Research Laboratories to optimize code
> from compilers.
> 
> If I understand him correctly, C++ has more generalized cases to consider,
> and cannot be optimized well unless more information is retained from
> initial to final phases of compilation, so that ambiguities can be resolved.
> 
> An approach to optimizing C destroys information after it is no longer
> necessary.
> 
> Although it is not necessary for optimization of C, it is necessary for
> proper optimization of C++. At the phase where the information is needed,
> it has already been destroyed. 
> 
> It is politically difficult to tell people to retain information their work
> has no need for, just because the work at a later phase of compilation
> requires it.
> 
> When one organization innovates by retaining the information so that it can
> be used later in the compilation, then I imagine others will follow, and we
> will see an industry-wide improvement in execution speed of C++ programmes.

The documentation on gcc 3.4's new parser tree might interest you.  I
read a small document on it eariler this year and it sound like that is
excactly the kind of thing they are doing with it.  Or at least planning
to do now that the infastructure exists for it.

http://gcc.gnu.org/projects/tree-ssa/

If was merged into what is now gcc 3.4 in May of this year.

Lennart Sorensen
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml





More information about the Legacy mailing list