update #1 after review

.clang-format

@@ -3,7 +3,7 @@ BasedOnStyle: Google
 ---
 Language:               Cpp
 Cpp11BracedListStyle:   true
-Standard:               Cpp11
+Standard:               c++11
 DerivePointerAlignment: false
 PointerAlignment:       Right
 ---

k2/csrc/cuda/arch.h

@@ -9,7 +9,15 @@
 namespace k2 {
 
 /**
- * K2_PTX_ARCH get the target PTX version through the active compiler (or zero for the host).
+ * @def K2_PTX_ARCH
+ *
+ * @brief
+ *  K2_PTX_ARCH get the target PTX version through the active compiler.
+ *
+ * @details
+ *  - for host, it's 0.
+ *  - for device, it's `__CUDA_ARCH__`, which indicates the
+ *                compute compatibility, should >= 200.
  */
 #ifndef K2_PTX_ARCH
   #ifndef __CUDA_ARCH__
@@ -19,6 +27,6 @@ namespace k2 {
   #endif
 #endif
 
-}  // end namespace k2
+}  // namespace k2
 
 #endif  // K2_CSRC_CUDA_ARCH_H_

k2/csrc/cuda/context.cu

@@ -9,7 +9,7 @@
 #include <cstdlib>
 
 #include "k2/csrc/cuda/context.h"
-#include "k2/csrc/cuda/debug.cuh"
+#include "k2/csrc/cuda/debug.h"
 #include "k2/csrc/cuda/error.h"
 
 static constexpr size_t kAlignment = 64;