[mpich-discuss] Compiling with Intel icc

[Thomas Jahns]

On 09/07/2011 05:03 PM, Dave Goodell wrote:
> On Sep 7, 2011, at 8:48 AM CDT, Thomas Jahns wrote:
> 
>> I also noticed that to make the checks for missing declarations 
>> (PAC_FUNC_NEEDS_DECL) work with icc, one needs to add
>> 
>> -diag-error 147
>> 
>> to MPICH2LIB_CFLAGS.
> 
> We have previously had problems with the PAC_FUNC_NEEDS_DECL macro.  Which
> way is it failing, specifically?

icc only gives a warning where the macro expects compilation to fail. By
escalating the diagnostic for functions called in violation of a declared
interface from warning to error, the configure script correctly identifies these
functions to not need an extra interface.

> Does it cause configure to abort? Does it result in false positives
> (incorrectly determining that a local definition is required) or false
> negatives (incorrectly determining that a local definition is NOT required)?

No abort, but configure gives incorrect results which then result in the
generation of dubious declarations (i.e. substituting int for ssize_t in the
case of readlink, even though sizeof(int) != sizeof(ssize_t)), in your words:
false positives.

> I've never heard of needing "-diag-error 147" with icc, but it's not entirely
> surprising that something is needed.

Might not be required, but at least it reduces warnings during compilation
considerably.

> We put these things on the wiki sometimes, although that can only be modified
> by MPICH2 core developers.  Also the mailing list archives are indexed by
> google and can be a reasonable source of knowledge on these sorts of issues.
> 
> I've added your suggestion to our compiler quirks page:
> http://wiki.mcs.anl.gov/mpich2/index.php/Compiler_Quirks#Intel_Compiler

I see, that documents some compiler properties, but I rather meant setups people
succeeded with, i.e. compiler versions used together and corresponding configure
arguments.

If you are interested: xlf still displays its manual pages in versions 12.x and
13.x and its still just as annoying, or even more so, since the manual pages
only grow longer.

Greetings, Thomas
-- 
Thomas Jahns
DKRZ GmbH, Department: Application software

Deutsches Klimarechenzentrum
Bundesstraße 45a
D-20146 Hamburg

Phone: +49-40-460094-151
Fax: +49-40-460094-270
Email: Thomas Jahns <jahns at dkrz.de>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5884 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20110907/e409e46d/attachment-0001.bin>