I am new to cuda programming. Was trying to profile my convolution implementation in cuda. Every time I run nvprof, it gives a different order of GPU activities and API calls sorted in the order of time taken. Is there a way to fix the order to how they are called in the code?
I am using the following command:
nvprof --csv --log-file
Also, if anyone knows of any useful and common commands to use while profiling, I would really appreciate your comments.