problem compiling cbenchmarks

[Erven Rohou]

Hello,

I haven't used the cbenchmarks for a while, and I started with a fresh checkout.
Some benchmarks do not compile because of missing macros. For example
telecom_gsm needs a -DSASR (or a fix in the source code). I have
similar problems with consumer_lame, consumer_tiff. None of the
provided Makefiles use those flags.
Do you have special Makefiles, or am I missing something?

Thanks,

--
Erven.

[Abdul Wahid Memon]

Hello

I too had the problem with telecom_gsm and consumer_lame but consumer_tiff is working well with the makefiles provided. consumer_mad has the compilation problem too. May be they need a source code fix.

Regards

Abdul

--

You received this message because you are subscribed to the Google Groups "ctuning-discussions" group.
To post to this group, send email to ctuning-d...@googlegroups.com.
To unsubscribe from this group, send email to ctuning-discuss...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/ctuning-discussions?hl=en.

[Yuri Kashnikoff]

Yes, the problem is in Makefiles. I will fix them tomorrow.

Thanks!

[Grigori Fursin]

Thanks, Yuri...

That's my fault - I have been updating all Makefiles
for the new CCC framework, automated the converting but
didn't process them correctly. I just realized that now...
So, we need to look back at the previous correct version
and add all the -D or other specific flags ...

Important - add them as a new environment variable but
don't touch the old ones since I modify them through CCC
framework.

There was another issue with -pg profiling - we need to
add one more environment variable for linking. Yuri,
if you will have time and can add that as well, that
would be great - we can discuss that later face-to-face ...

Cheers,
Grigori

> discussions...@googlegroups.com.

[Erven Rohou]

On Wed, Dec 9, 2009 at 09:25, Grigori Fursin <grigori...@inria.fr> wrote:
> Thanks, Yuri...
>
> That's my fault - I have been updating all Makefiles
> for the new CCC framework, automated the converting but
> didn't process them correctly. I just realized that now...
> So, we need to look back at the previous correct version
> and add all the -D or other specific flags ...

Ok, it makes sense now.

>
> Important - add them as a new environment variable but
> don't touch the old ones since I modify them through CCC
> framework.

I understand that you modify some flags through the CCC framework.
However, some flags are needed to compile the benchmarks,
independently of CCC (for example, telecom_gsm does not compile
without the -DSASR flag).
I would put the required ones in the Makefiles, and the variable ones
appended through environment variables.
That way, people can use the benchmarks in "standalone" mode. And you
can use them in the CCC framework.

Comments?


--
Erven.

> There was another issue with -pg profiling - we need to
> add one more environment variable for linking. Yuri,
> if you will have time and can add that as well, that
> would be great - we can discuss that later face-to-face ...
>
> Cheers,
> Grigori
>
>
>> -----Original Message-----
>> From: ctuning-disc...@googlegroups.com [mailto:ctuning-
>> discussi...@googlegroups.com] On Behalf Of Yuri Kashnikoff
>> Sent: Tuesday, December 08, 2009 7:45 PM
>> To: ctuning-d...@googlegroups.com
>> Subject: Re: problem compiling cbenchmarks
>>

>> discussions...@googlegroups.com.

[Grigori Fursin]

Hi Erven,

> > Important - add them as a new environment variable but
> > don't touch the old ones since I modify them through CCC
> > framework.
>
> I understand that you modify some flags through the CCC framework.
> However, some flags are needed to compile the benchmarks,
> independently of CCC (for example, telecom_gsm does not compile
> without the -DSASR flag).
> I would put the required ones in the Makefiles, and the variable ones
> appended through environment variables.
> That way, people can use the benchmarks in "standalone" mode. And you
> can use them in the CCC framework.

I totally agree and that was my intention to keep standalone and CCC
related flags together but I mixed it up ... Basically I change
flags through CCC_OPT variable so we can add another variable with
standalone flags that will not be touched by CCC framework ...

Cheers,
Grigori

[Joern Rennecke]

Can I add new attributes to the xml records of executed passes that
the adapt plugin creates, and will these new attributes be then present
when I read in a re-ordered pass list?

[Yuri Kashnikoff]

> Can I add new attributes to the xml records of executed passes that
> the adapt plugin creates, and will these new attributes be then present
> when I read in a re-ordered pass list?

It should be so. But this depends on how you will read them.