[CUDA] Document recent changes

[Hahnfeld]

Summary

• Finding installations via ptxas binary
• Relocatable device code

[tra]

Overall LGTM. We could use specific instructions for what exactly one needs to do in order to use relocatable device code in practice. Could be a separate patch.

[Hahnfeld]

In D45449#1062642, @tra wrote:

We could use specific instructions for what exactly one needs to do in order to use relocatable device code in practice. Could be a separate patch.

Maybe I should add an example to https://llvm.org/docs/CompileCudaWithLLVM.html?

docs/ReleaseNotes.rst

 
 ...
 
 OpenMP Support in Clang
 ----------------------------------
 
 - ...
 
+CUDA Support in Clang
+---------------------
+
+- Clang will now try to locate the CUDA installation next to :program:`ptxas`
+  in the `PATH` environment variable. This behavior can be turned off by passing
+  the new flag `--cuda-path-ignore-env`.
+
+- Clang now supports generating object files with relocatable device code. This
+  feature needs to be enabled with `-fcuda-rdc` and my result in performance
+  penalties compared to whole program compilation. Please note that NVIDIA's
+  :program:`nvcc` must be used for linking.
+
 Internal API Changes
 --------------------
 
 These are major API changes that have happened since the 6.0.0 release of
 Clang. If upgrading an external codebase that uses Clang as a library,
 this section should help get you past the largest hurdles of upgrading.
 
 -  ...

include/clang/Driver/Options.td

 def ptxas_path_EQ : Joined<["--"], "ptxas-path=">, Group<i_Group>,
   HelpText<"Path to ptxas (used for compiling CUDA code)">;
 def fcuda_flush_denormals_to_zero : Flag<["-"], "fcuda-flush-denormals-to-zero">,
   Flags<[CC1Option]>, HelpText<"Flush denormal floating point values to zero in CUDA device mode.">;
 def fno_cuda_flush_denormals_to_zero : Flag<["-"], "fno-cuda-flush-denormals-to-zero">;
 def fcuda_approx_transcendentals : Flag<["-"], "fcuda-approx-transcendentals">,
   Flags<[CC1Option]>, HelpText<"Use approximate transcendental functions">;
 def fno_cuda_approx_transcendentals : Flag<["-"], "fno-cuda-approx-transcendentals">;
-def fcuda_rdc : Flag<["-"], "fcuda-rdc">, Flags<[CC1Option, HelpHidden]>,
+def fcuda_rdc : Flag<["-"], "fcuda-rdc">, Flags<[CC1Option]>,
   HelpText<"Generate relocatable device code, also known as separate compilation mode.">;
 def fno_cuda_rdc : Flag<["-"], "fno-cuda-rdc">;
 def dA : Flag<["-"], "dA">, Group<d_Group>;
 def dD : Flag<["-"], "dD">, Group<d_Group>, Flags<[CC1Option]>,
   HelpText<"Print macro definitions in -E mode in addition to normal output">;
 def dI : Flag<["-"], "dI">, Group<d_Group>, Flags<[CC1Option]>,
   HelpText<"Print include directives in -E mode in addition to normal output">;
 def dM : Flag<["-"], "dM">, Group<d_Group>, Flags<[CC1Option]>,