[llvm-commits] [llvm-gcc-4.2] r55650 - /llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp

[Duncan Sands]

> >> Don't convert atomics that aren't supported by the target.
> >> This gets us undefined functions at linktime instead of
> >> crashes in llvm, which is what gcc does.
> >
> > how about teaching llvm to output the required library calls
> > if a target does not support an atomic intrinsic (I'm assuming
> > such library calls exist...)?
> 
> That could be done; what you would get is the same library call
> that already exists in the gcc IL.  Why change it and change it back?

Because llvm-gcc is not the only front-end to llvm, and also
that way you don't need to put information about which targets
support what into front-ends (which risks being incomplete and
can get out of sync with llvm).

Ciao,

Duncan.