This is the mail archive of the libstdc++@sources.redhat.com mailing list for the libstdc++ project.


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

Re: slow V3 configures (was Re: ICE during bootstrap.)


[please retain the v3 list in the cc's]


On Wed, Dec 06, 2000 at 11:29:37AM +0100, Akim Demaille wrote:
> | 
> | Anyhow, I patched all the math tests to explictly do the same test that
> | AC_CACHE_VAL does before calling AC_CACHE_VAL itself.  The speedup is
> | /considerable/.  The config.cache and bits/c++config.h results are identical.
> 
> Could you detail this a little bit?  I don't understand what you
> mean.  You seem to say your wrapped AC_CACHE_VAL within AC_CACHE_VAL,
> right?

No, I wrapped AC_CACHE_VAL within the same test that CACHE_VAL will
eventually do.  If the value is cached, we skip more stuff than we would
with just CACHE_VAL.


> I never saw the code you are referring too, so I might say something
> stupid, but:

I haven't posted anything because I wanted some feedback on the question
below.  (I've gotten one "seems worth it to me" in private email.  :-)
I'll probably post+apply a cleaned-up patch shortly.


> 1. One huge AC_LANG_PUSH/POP pair is definitely the best way to go

That would be great, but it looks like we can't safely do this.  (But it's
been three months since I did tests on that part, so circumstances might
have changed.  I doubt it, though.)


> 2. If you can't, be sure to put them *inside* the AC_CACHE_VAL, don't
>    wrap it.

They always have been.  :-)


> | One downside:  it's AC_CACHE_VAL that does the printing of "(cached)".
> | My change means that a cached value still shows on the screen as
> | 
> |     checking for _expf declaration... no
> |     checking for _fabsf declaration... no
> |     checking for _floorf declaration... no
> |     checking for _fmodf declaration... no
> |     checking for _frexpf declaration... no
> |     checking for _ldexpf declaration... no
> | 
> | etc, like regular tests.  Just a /lot/ faster.  :-)  This may not be
> | acceptable; what do y'all think?  I'm hesitant to try and use the $ac_*
> | variables controlling echo(1)'s no-newline behavior, for some reason.
> 
> You're right not to depend on these, their name changed in CVS
> Autoconf.

Okay.  A user with cached values will just see these tests go by way fast.


-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.

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