Hi,

> available, I'm not sure a pocl Windows port would be THAT much work.
Well, i have tried both Mac and Windows. Here's the windows
tale ;) ..

Clang+LLVM has windows binaries... only they're unusable with
pocl. The problem is, pocl requires a bunch of -libclang* and -libLLVM*
libraries; the windows installer doesn't ship any of these, IIRC. 

Now, seeing this, i figured i'd try doing Visual Studio build, which
was actually quite easy to do, and it produced a full clang/llvm
installation. (didn't check if it worked ;) The next thing one needs is
to build hwloc (autotools project) in Visual Studio, and if i
understand correctly, a replacement for libltdl which would work in
windows.

Feeling that would be a bit too much work, i tried to do a mingw build
of clang/llvm, hoping hwloc+ltdl would be easier in mingw. The first
warning sign was that clang/llvm has hardcoded paths for the standard
library headers (!), and those paths only work with certain mingw gcc
versions (4.7.0 being the latest..). Anyway, after a few fixes of the
build system, i got a working clang+llvm via mingw, which linked to gcc
stdlib 4.7.0. To get hwloc working, i needed a bunch of other libraries
for an outdated mingw... and i must have introduced a bit of chaos into
my mingw libraries, because somewhere down the road the "make" command
started to segfault :)

It was about that time i figured i wasted enough time on this..

> But we do not know before a Windows developer attempts to do it.
Definitely. I admit defeat :)

> Is the branch in a stable state for merging to master now?
Well... no. I'll send a separate email about that..

> Are you ready to maintain the cmake build support in pocl in case we do?
Sure, but it would be nice to have at least 1 more person that has some
cmake experience.

Regards,
-- mb

------------------------------------------------------------------------------
HPCC Systems Open Source Big Data Platform from LexisNexis Risk Solutions
Find What Matters Most in Your Big Data with HPCC Systems
Open Source. Fast. Scalable. Simple. Ideal for Dirty Data.
Leverages Graph Analysis for Fast Processing & Easy Data Exploration
http://p.sf.net/sfu/hpccsystems
_______________________________________________
pocl-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/pocl-devel

Reply via email to