[gmx-developers] Zero cell size

[Roland Schulz]

On Tue, Jul 22, 2008 at 10:52 AM, David van der Spoel <spoel at xray.bmc.uu.se>
wrote:

> Axel Kohlmeyer wrote:
>
>> On Tue, 22 Jul 2008, Roland Schulz wrote:
>>
>> hi roland,
>>
>> RS> Sorry for making the thread confusing by replying to myself.
>> RS> RS> The error: "The maximum allowed number of cells is: X 0 Y 0 Z 5"
>> must have
>> RS> been caused by a bad build. I couldn't reproduce it after rebuilding
>> again.
>> RS> RS> The error: "The X-size (0.000000) times" seems to be caused by PGI
>> 7.1.6,
>> RS> because PGI 7.2 and Gcc produce correct results.
>>
>> why do you even try to compile with PGI? i only would only do it after
>> getting paid for the debugging. in my experience the PGI C/C++ compilers
>> always have been bug ridden (much more than their fortran 77/90 compilers
>> which have their own share of problems, particularly pgf90).
>> i've been testing them on and off for 10 years now, and produce with
>> few exceptions significantly slower executables than gcc or intel.
>>
>> having PGI compilers as default (and no viable alternative to pgf90)
>> has caused our group a great deal of grief on cray xt3 machines and
>> cost me a lot of time finding and implementing workarounds...
>>
>>  You said the word: Cray. As with IBM their salesforce are smarter than
> their engineers and these boxes are popping up like mushrooms at least in
> europe. And they make it very difficult to compile stuff with something else
> then PGI compilers...
>
Yes here it is also Cray. Building works also very good with GNU and
Pathscale but the Profiling Tools work often better with PGI because PGI is
their default.

Roland
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20080722/81e522ff/attachment.html>