Merge branch 'gpu_through_scope_path' into 'dev'

[C++] Update description of nvtxScopeAttr_v1::path

Author: evanramos-nvidia

c/include/nvtx3/nvToolsExtPayload.h

@@ -913,20 +913,23 @@ typedef struct nvtxScopeAttr_v1
     size_t      structSize;
 
     /**
-     * Path delimited by '/' characters, relative to parentScope. Leading
+     * Path delimited by '/' characters, relative to @ref parentScope. Leading
      * slashes are ignored. Nodes in the path may use name[key] syntax to
      * indicate an array of sibling nodes, which may be combined with other
      * non-array nodes or different arrays at the same scope. Node names should
      * be UTF8 printable characters. '\' has to be used to escape '/', '[', and
      * ']' characters in node names. An empty C string "" and `NULL` are valid
      * inputs and treated equivalently.
      *
-     * A GPU can be specified using its
-     * - unique identifier (UUID) with "GPU[UUID:#]",
+     * A GPU can be specified using its:
+     * - Unique identifier (UUID) with "GPU[UUID:#]",
      * - CUDA device ID (sensitive to CUDA_VISIBLE_DEVICES) with "GPU[CUDAID:#]",
      * - NVML (nvidia-smi) device ID with "GPU[NVSMI:#]"
+     *
      * (replace `#` with the actual device ID).
      * For display purposes, a tool is recommended to show a pretty name.
+     * To clearly identify a GPU, the @ref parentScope should also match
+     * the GPU's execution context.
      */
     const char* path;