[PATCH] D47849: [OpenMP][Clang][NVPTX] Enable math functions called in an OpenMP NVPTX target device region to be resolved as device-native function calls

[Gheorghe-Teodor Bercea via Phabricator via cfe-commits]

gtbercea added a comment.

Just to address any generality concerns:

This patch fixes the problem of calling libdevice math functions for all platform combinations. It ensures that the OpenMP NVPTX target region will NOT call any host math functions (which ever host that may be) IF equivalent device functions are available.

I think there was a confusion regarding header file inclusion. This patch does not address any issues that might arise from the user including header files (be it math.h or some other header). Any failure related to header file inclusion (such as the reported x restriction issue on x86) is unrelated to what this patch aims to do. Before the functionality in this patch can kick in, any user-included headers must successfully pass all checks in place for the NVPTX toolchain. A fix in the direction of the one proposed in one of the comments above is probably required. The fix would also needs its own separate patch.


Repository:
  rC Clang

https://reviews.llvm.org/D47849