Re: i386 backend

[David Edelsohn <dje at watson dot ibm dot com>]

	I personally have heard lots of vague comments about bad
performance of GCC, but rarely have been able to track it down to any
quantitative facts from the people making the comments.  What I do find is
people repeating the comments that they heard from others.

	GCC has plenty of problems and plenty of room for improvement, but
vague, unsubstantiated rumors simply are FUD.  I do not care if one person
says it or one hundred people say it (e.g., "many people from many
different places, working with many different projects, with a lot of
different code...")

	The GCC developers do work to improve code generation in response
to specific performance problem reports.  We have limited time and energy
to fight vague, qualititative assertions.

	I and other GCC developers have lots of statistics about the good
and bad performance of GCC on various benchmarks.  What surprises me is
the complaints about GCC performance from people outside the developer
community without any similar statistics.

	Please send small code examples of performance problems and we
will analyze them to uncover what optimizations need improvement.  Also,
please remember that GCC is developed by volunteers.

David