This is the mail archive of the gcc-patches@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]
Other format: [Raw text]

Re: Clean up edge_def's crossing_edge field





On Aug 18, 2004, at 3:49 PM, Steven Bosscher wrote:


On Thursday 19 August 2004 00:45, Caroline Tice wrote:
Before you commit this, if it's not already too late, you should do a
profiledbootstrap, with all debugging flags
turned off (you will have to edit Makefile.in). Just to make sure your
change doesn't mess up the hot/cold partitioning please.

Actually, no, you should have fixed this long ago, when I and others asked you to.

At the time you asked me to change from using a boolean field into wrapping this
into the flags fields, I explained that I considered that to be a pure matter of taste, that
there was no pressing technical reason to make such a change, and that I wanted to focus
first on fixing potential correctness issues. Doing it your ways saves a very modest amount of
space; doing it my way saves a very modest amount of time *and makes it easier to debug*, which
was an important issue for me at the time.



I did test profiling, but not a profiledbootstrap.  And I'm not going
to either.  The patch is perfectly obvious, if this breaks something
then it was broken before.


This is so blatantly false that it's really laughable. Even if your change is technically trivial,
it is easy to make a tiny mistake (typographical, etc.) which will introduce a bug. Since bootstrapping
the normal way does NOT exercise this option at all, bootstrapping the normal way does not
guarantee that you found any such bugs in relation to this optimizations. If the profiling testing you did
was extensive then it might have exercised most possible ways that this could fail, so it might have covered everything, but I have found over the past 18 months I have been working on this optimization that one or two simple profiling tests are not nearly sufficient to prove you have found all your bugs. You did not
state what profiling testing you did perform...





Finally, I am surprised by the insulting nature of your postings. I do not believe I have done anything
to merit such abuse, and I would appreciate your refraining from it in the future.


-- Caroline Tice
ctice@apple.com
Gr.
Steven




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