Huge compilation time

[Quentin Gasper]

Hi devs,

I noticed that the NavierStokes.cpp file takes very long time to
compile. Doing "time make NavierStokes.cpp.o" prints this

[100%] Building CXX object plugins/UFEM/src/UFEM/CMakeFiles/
coolfluid_ufem.dir/NavierStokes.cpp.o

real 6m19.409s
user 6m17.484s
sys 0m2.209s

Process took about 2.1 GB of memory without swaping.

I'm experiencing this on my Linux machine here at VKI (FedoraCore 13,
gcc 4.4.5, all deps installed). I haven't checked on other OSes/
compilers.

I checked on my laptop with clang and it takes about 1 minute (I
consider it's OK since it's Proto code).

Is it normal that this file takes so much time to compile on Linux ?

[Bart Janssens]

On Thu, Nov 17, 2011 at 2:20 PM, Quentin Gasper <gque...@gmail.com> wrote:
> I noticed that the NavierStokes.cpp file takes very long time to
> compile. Doing "time make NavierStokes.cpp.o" prints this

Hi Quentin,

Unfortunately, this file is slow to compile. I have tweaked the memory
settings for GCC for that particular file, otherwise it would not
compile at all on my laptop. That said, 6m seems very long, did your
machine start to swap? I think here it takes about a minute with GCC
on linux.

Cheers,

--
Bart

[Quentin Gasper]

Hi Bart,

No the computer didn't swap.

On Nov 17, 4:49 pm, Bart Janssens <bart.janss...@lid.kviv.be> wrote:

[Bart Janssens]

On Thu, Nov 17, 2011 at 5:57 PM, Quentin Gasper <gque...@gmail.com> wrote:
> Hi Bart,
>
> No the computer didn't swap.

Is this the build in the dashboard? If so, the solution is simple:
ditch GCC 4.4 for at least GCC 4.5 :)
(there was a huge template-performance related improvement)

Cheers,

--
Bart

[Quentin Gasper]

On Nov 17, 11:10 pm, Bart Janssens <bart.janss...@lid.kviv.be> wrote:

> On Thu, Nov 17, 2011 at 5:57 PM, Quentin Gasper <gquen...@gmail.com> wrote:
> > Hi Bart,
>
> > No the computer didn't swap.
>
> Is this the build in the dashboard?

If you speak about the Experimental builds of yesterday, yes ;)

GCC 4.4 is the native compiler on VKI machines, that's why I'm using
it. I tried to compile GCC 4.6, but it failed because I'm missing the
32-bit development package for libc (I know, I can ask the system
admin to install it, but he was quite busy those last days). I will
give Clang a try.

Anyway, it's not a big problem for me since I don't plan to compile
this file everyday; but it might become a problem when people at VKI
start to work on CF3 as I don't want to force them to compile a
compiler (which takes some time in the case of GCC).

Let's leave this problem aside for the moment.

See you.

[Tamás Bányai]

Hey Guys,

Ill be brief (writing from phone in a lecture).
1.: long compilation of proto monster is indeed an issue, but more for
a student, who has to work with it.
2.: someone who does not work with ufem, just skips compiling this plugin.
3.: the real thing: please keep in mind when forcing to use latest
boost/gcc... that specializing on dependencies will in the long term
cause more pain then gain and we will end up on being able to compile
everything on the only distros we using on a daily basis.
4.: clang linker does not work properly on fedora (missing symbols
everywhere). Do i miss stg?

Th

Quoting Quentin Gasper <gque...@gmail.com>:

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

[Tamás Bányai]

Dear Devz,

Just to lift the mood and shed some optimistic light on CF3
compilation times. :)

I am compiling sage (a collection of nice all-in-one mathematics
software package), on 4 cores on my work machine and it roughly takes
4 and a half hours.

There is still room to grow. ;)

Cheers:
Th