[gentoo-user] Gcc 4.1 compile speed

[gentoo-user] Gcc 4.1 compile speed

[Mark Kirkwood]

Hi,

After upgrading to gcc to 4.1, I notice that pure c projects take 
significantly longer to compile than under 3.4 (e.g. Postgres 8.1.4 - 7 
minutes vs 5 minutes for -O2 optimization) - while c++ ones seem largely 
unaffected (i.e Mysql 5.1 - 33 minutes for -O3 in both cases).

Is this just a known issue, or is there some way to improve the compile 
times for pure c? (*apart* from the obvious dropping optimization to 
-O0...).

Thanks

Mark
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Gcc 4.1 compile speed

[Richard Fish]

On 9/6/06, Mark Kirkwood <markir@paradise.net.nz> wrote:
> After upgrading to gcc to 4.1, I notice that pure c projects take
> significantly longer to compile than under 3.4 (e.g. Postgres 8.1.4 - 7
> minutes vs 5 minutes for -O2 optimization)

To paraphrase from the amd64 list, It makes faster binaries slower.
But yeah, known issue:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21456
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23955
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18687

Bottom line is that the optimizer makes many more passes over the code
now, so produces better optimizations, but it takes longer.  We can
only hope that 4.2 begins to speed things up again.

BTW, -O3 is meant to take a long time to compile...so it is probably
more useful to do your C++ comparisons at -O2.

-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Gcc 4.1 compile speed

[Mark Kirkwood]

Richard Fish wrote:
> On 9/6/06, Mark Kirkwood <markir@paradise.net.nz> wrote:
>> After upgrading to gcc to 4.1, I notice that pure c projects take
>> significantly longer to compile than under 3.4 (e.g. Postgres 8.1.4 - 7
>> minutes vs 5 minutes for -O2 optimization)
> 
> To paraphrase from the amd64 list, It makes faster binaries slower.
> But yeah, known issue:
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21456
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23955
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18687
> 
> Bottom line is that the optimizer makes many more passes over the code
> now, so produces better optimizations, but it takes longer.  We can
> only hope that 4.2 begins to speed things up again.
> 

Thanks Richard (oops, forgot to search the buglist first, sorry).

> BTW, -O3 is meant to take a long time to compile...so it is probably
> more useful to do your C++ comparisons at -O2.
> 
>

Yeah, actually I put -O2 in CXXFLAGS, but noticed the the Mysql build is 
overriding it (if they think -O3 is better, I guess I'll let 'em insist)!

Cheers

Mark
-- 
gentoo-user@gentoo.org mailing list