This is the mail archive of the
mailing list for the GCC project.
Re: Faster compilation speed
- From: kaih at khms dot westfalen dot de (Kai Henningsen)
- To: gcc at gcc dot gnu dot org
- Date: 13 Aug 2002 21:00:00 +0200
- Subject: Re: Faster compilation speed
- Comment: Unsolicited commercial mail will incur an US$100 handling fee per received mail.
- Organization: Organisation? Me?! Are you kidding?
- References: <20020813173609.3DC80F2D49@nile.gnat.com>
firstname.lastname@example.org (Robert Dewar) wrote on 13.08.02 in <20020813173609.3DC80F2D49@nile.gnat.com>:
> But remember that work you put in on speeding up the compiler is work
> that you do not put in on improving the compiler. As time goes on, quality
> of generated code continues to be critical, compiler speed is less critical.
Not necessarily. It depends on what you do to get the speed up.
If, for example, you get the speed up by noticing that gcc does some
useless work, and you eliminate that, that is most definitely an
improvement. Or if you find that another algorithm gives at least as well
a result in much shorter time.
On the other hand, if you find you can improve the speed by writing lots
of spaghetti code, that is probabl not an improvement.
> Very little in practice. You do not rebuild a million line system every
> two minutes after all, and in practice once the build time for a large
> system is down in the ten minute range, the gains in making it faster
> diminish rapidly. This is not a guess, as I say, this is an observation
Well, my personal observation is that a change much like this ten minutes
to one has recently made me *much* more productive. For one thing, if the
right change will cause a complete rebuild to the project (because a
common header file changes), I no longer look for alternate changes just
to avoid that. Nor do I have as much trouble remembering what exactly that
last change was that I'm supposed to test now, as I have already spent
time thinking about (and possibly implementing) three more changes.
Sure, ten hours to one hour gives you back much more time. But ten minutes
to one is still enough for a significant change in development style.
Ten seconds to one does not look like it would be all that important - but
not having lived with it, I can't say for sure.