Does Jetson TK1 support cudaHostRegister ?

I have tried to compile code containing cudaHostRegister() with three possible flags(default, portable and mapped), however all of then show “operation not supported” error string during execution.

I read from this post http://stackoverflow.com/questions/23600403/cudamallocmanaged-returns-operation-not-supported that mapped host memory is not supported in 32-bit machines. However, I cannot find any other docs talking about the availability of pinned-host-memory on different GPU architectures.

Since asyncMemcpy is quite important to the speed of my code, I wonder if there is anyway to transfer memory asynchronously on Jetson TK1?

Thanks!