[cfe-dev] [Openmp-dev] CMAKE_INSTALL_PREFIX vs -fopenmp and clang search library search path
cbergstrom at pathscale.com
Fri May 15 21:44:53 PDT 2015
maybe we're talking about the same thing
Each STL has a different name.. libc++ isn't named libstl.so
So calling it libiomp.so or libwhocares.so ensures that it doesn't
conflict with something down the road. Again it doesn't really matter
since the packager can now easily change the name. We may want to
allow that same cmake variable to be used inside llvm/clang so that if
everything is built together it all plays nice. This isn't something
user facing. The compiler will/should automatically add the right
As an example
clang -mp foo.c # User never see's the name of the lib unless they do
ldd later or something. Assuming the default is llvm/openmp.
I hope the noise on this topic dies down soon. It's really all bikeshed
On Sat, May 16, 2015 at 11:35 AM, Chris Lattner <clattner at apple.com> wrote:
>> On May 15, 2015, at 3:47 PM, C Bergström <cbergstrom at pathscale.com> wrote:
>> Adding a prefix may be desireable to avoid conflicts or confusion.
> Sure, OTOH, we have precedent with this. There were many c++ standard libraries before libc++. I'm ok with LLVM's implementation being the defacto implementation, aren't you?
>> On Sat, May 16, 2015 at 5:35 AM, Chris Lattner <clattner at apple.com> wrote:
>>> On May 13, 2015, at 9:29 AM, Peyton, Jonathan L
>>> <jonathan.l.peyton at intel.com> wrote:
>>> We understand that iomp is undesirable. What we don’t know is what name is
>>> What name is desirable instead of iomp? Do we want to link via…
>>> -lnotintelomp ( just kidding J )
>>> … Some other name? I get the feeling someone outside of Intel should make
>>> this choice.
>>> Why not -lomp?
>>> Openmp-dev mailing list
>>> Openmp-dev at dcs-maillist2.engr.illinois.edu
More information about the cfe-dev