This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: basic-block and profile-based optimizing (was Re: New attribute "infrequent"?)


On Fri, Aug 31, 2001 at 06:25:09PM +0200, Andreas Jaeger wrote:

> You cannot profile accuratly the C library.  It can be profiled - but
> I don't expect that it's realistic:
> 
> - What kind of application will you use?  This needs to be a
>   self-written application that calls all calls with typical
>   arguments.
> - How will you handle exceptional stuff?  Should those be called by
>   the special application?  For example should strerror be called - or
>   not?
> - There're plenty of different types of applications and it's
>   difficult to model even some of them.

I agree getting a useful profile for the general case is as you point out not easy at all,
and carries the risk of un-optimising common cases under particular workloads.

your points are valid for a library intended for general use. However, it is possible
to optimise a specific instance of the library on a particular machine.

This can be useful if machines are intended for a particular workload, and solves
the problem of needing a "typical" library user - you use the /real/ ones.

This requires a continuous profiler such as DCPI or (shameless plug) my x86 prototype
oprofile.sf.net

regards
john

-- 
"I'm not sure which upsets me more: that people are so unwilling to accept
 responsibility for their own actions, or that they are so eager to regulate
 everyone else's."
	- Kee Hinckley


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]