[PATCH] LLVM OpenMP CMake Overhaul

[Chandler Carruth]

REPOSITORY
  rL LLVM

================
Comment at: runtime/cmake/LibompHandleFlags.cmake:44
@@ +43,3 @@
+    if(${IA32})
+        libomp_append(flags_local -m32 LIBOMP_HAVE_M32_FLAG)
+        libomp_append(flags_local /arch:SSE2 LIBOMP_HAVE_ARCH_SSE2_FLAG)
----------------
jhowarth wrote:
> chandlerc wrote:
> > Why do you want to blindly pass -m32 when it is supported? Seems odd that this is the only predicate..
> This may be for the building the fat libomp.dyib on x86_64 darwin (containing both i386 and x86_64 code) which requires -m32 to be  explicitly passed to the native x86_64 compiler.
Sure, I know that this is what -m32 does, but it doesn't make sense to *blindly* pass it. I feel like we should be very specifically testing for targeting i386 with a x86_64 toolchain first.

http://reviews.llvm.org/D10656

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/