jlebar added inline comments.
================
Comment at: lib/Driver/ToolChains.cpp:4902
+ DeviceOffloadingKind == Action::OFK_Cuda) &&
+ "The offloading kind is not OpenMP or CUDA.");
----------------
Not sure this assertion message helps us much beyond what's already in the code.
================
Comment at: lib/Driver/ToolChains.cpp:4914
+ options::OPT_fno_cuda_approx_transcendentals,
+ false))
+ CC1Args.push_back("-fcuda-approx-transcendentals");
----------------
Are these changes related to this patch?
I have no problem cleaning up whitespace errors like these, but would prefer
for them to be split out separately if possible.
================
Comment at: lib/Driver/ToolChains.cpp:4961
+ // If this is an OpenMP device we only need to append the gpu name.
+ if (DeviceOffloadKind == Action::OFK_OpenMP) {
----------------
s/device/compilation/?
================
Comment at: lib/Driver/ToolChains.cpp:4961
+ // If this is an OpenMP device we only need to append the gpu name.
+ if (DeviceOffloadKind == Action::OFK_OpenMP) {
----------------
jlebar wrote:
> s/device/compilation/?
An "otherwise" would probably be helpful in this comment.
================
Comment at: lib/Driver/ToolChains.cpp:4966
+
+ // TO DO: get the right arch from the offloading arguments when not
+ // using the default compute capability of sm_20.
----------------
TODO is idiomatically one word
================
Comment at: lib/Driver/Tools.cpp:12136
// Obtain architecture from the action.
- CudaArch gpu_arch = StringToCudaArch(JA.getOffloadingArch());
assert(gpu_arch != CudaArch::UNKNOWN &&
----------------
Why does JA.getOffloadingArch have the wrong value? Isn't the purpose of
JA.getOffloadingArch to have this particular value? If it's broken, can we fix
it instead of doing this hack?
Repository:
rL LLVM
https://reviews.llvm.org/D29647
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
