this change follows CUDA convention where lower number is greater priority Change-Id: I72596a36449e818cbd8c175bf8519c51f46b1610