Hi,

 > It seems I've been fighting with static qt the whole night. I still
> haven't managed to compile it statically. The official script for static
> qt compilation on windows does not work, of course. And manual compiling
> with mingw32-make fails for no sensible reason... I sometimes feel like
> they make things complicated on purpose.

wait - are we allowed to link Qt statically? I thought I read something 
in the net that the open source version of Qt only allows dynamic links?


>     Namik just pushed a version which allows to also build it
>     dynamically. Will this work sufficiently well?
>
>
> Unfortunately, it didn't work out the way I thought it would. I still
> don't know what's the proper way to dynamically link libarchive with
> CMake. In QMake it works just fine by specifying -llibarchive...

Well, if the static link works, why should we bother with a dynamic 
link? Note that QMake fails on Linux because there is no precompiled 
shared library in the repository, so the -llibarchive is only a shortcut 
and does not tell the full story.


>       We will have to make sure Windows finds these dlls, which is
>     cumbersome... Any ideas how to ensure this? Copying the .dll files
>     to the .exe?
>
>
> Copying dll files is the simplest solution. And it shouldn't be too
> complicated, if only dynamic linking worked...

As far as I can tell, we only need to dynamically link with Qt, don't we?


>     1.) Boost: Namik, can you please remove all use of Boost (sources
>     plus build system files)? This is just required at very few
>     locations, mostly sparse stuff, so it shouldn't be a big deal. I'll
>     commit some matrix generation code soon, which will allow us to get
>     rid of all the IO and the testdata files, which is the primary use
>     of Boost.
>
>
> I tried removing Boost usage some time ago, but failed since sparse
> benchmark heavily depends on Boost. I couldn't manage to make it work by
> replacing Boost ublas with ViennaCL linalg functions.

Ok, I'll take care of this.


> We still need the IO for benchmarking with MatrixMarket files, right?

No, this can also be achieved with the sparse matrix wrapper in ViennaCL.


>     3.) libarchive+zlib: It seems like in the current build setup the
>     zlib is not enabled in libarchive, as it tries to run 'gzip' as a
>     fallback. Namik, can you please have a look at that?
>
>
> Libarchive will use zlib if ZLIB_INCLUDE_DIR and ZLIB_LIBRARY variables
> are defined. ZLIB_INCLUDE_DIR would be the source code in
> projectSourceFolder/external/zlib and ZLIB_LIBRARY would be the
> generated zlib library in the projectBuildFolder/external/zlib. I
> couldn't test this as I haven't managed to compile with CMake yet.

Which problems did you run into? I got everything to compile on a fairly 
virgin Windows 7 yesterday, so my setup shouldn't be that different from 
yours. The one thing that I found cumbersome, though, is that a 
reconfiguration in CMake does not work, so you always have to wipe the 
whole build folder whenever you change something in CMakeLists.txt

My guide to a CMake build for the developer-repo will be finished soon, 
I'll drop you a short note when it's ready.

Best regards,
Karli


------------------------------------------------------------------------------
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
_______________________________________________
ViennaCL-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/viennacl-devel

Reply via email to