[PATCH] Default to "nvptx[64]-unknown-cuda" triple for device-side CUDA compilation.

[Eric Christopher]

Since -fcuda-is-device is just a -cc1 option can we require that the user just pass the triple in? Seems like a small shortcut that's unneeded?

-eric


http://reviews.llvm.org/D8393

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