Hi! I cleaned up your code for you!

lib/cuda/driver/function.rb

@@ -187,7 +187,7 @@ def launch_grid(xdim, ydim = 1)
     # will execute on the default stream 0.
     # @overload launch_grid_async(xdim, stream)
     # @overload launch_grid_async(xdim, ydim, stream)
-    # @param [Integer] xdim The x dimensional size 
+    # @param [Integer] xdim The x dimensional size
     def launch_grid_async(xdim, ydim = 1, stream)
         s = Pvt::parse_stream(stream)
         status = API::cuLaunchGridAsync(self.to_api, xdim, ydim, s)

lib/cuda/runtime/memory.rb

@@ -34,7 +34,7 @@ class CudaDeviceMemory
 
     # Allocate memory on the device.
     # @param [Integer] nbytes The number of bytes of memory to allocate.
-    # @return [*SGC::Memory::MemoryPointer] A memory pointer to the allocated device memory. 
+    # @return [*SGC::Memory::MemoryPointer] A memory pointer to the allocated device memory.
     #
     # @note The returned memory pointer is enabled to call _free_ method on itself.
     def self.malloc(nbytes)

lib/cuda/runtime/stream.rb

@@ -86,9 +86,9 @@ def wait_event(event, flags = 0)
 
 
     # Let all future operations submitted to any CUDA stream wait until _event_ complete before beginning execution.
-    # @overload wait_event(event)                      
-    # @overload wait_event(event, flags)               
-    # @param (see CudaStream#wait_event) 
+    # @overload wait_event(event)
+    # @overload wait_event(event, flags)
+    # @param (see CudaStream#wait_event)
     def self.wait_event(event, flags = 0)
         status = API::cudaStreamWaitEvent(nil, event.to_api, flags)
         Pvt::handle_error(status, "Failed to make any CUDA stream's future operations to wait event: flags = #{flags}.")