Hi, I am new to CUDA and am learning to analyze data latency with the profiler.
My system is Ubuntu, CUDA 11.0, GTX 1060 which uses the Pascal Architecture.
After running “nvvp matrix_transpose.prof”:
While in the analysis view, I go to “CUDA Application analysis” --> “Performance-Critical Kernels”, from there I see that “Perform Kernel Analysis” is Greyed out. Also, when I go to “Perform Additional analysis”, which is not greyed out, nothing happens. I was looking for the chart that shows memory latency.
Also, I can not get to a stage where I can examine Latency or Memory Bandwidth.
I realize that the new Nsight Systems CLI is what is being recommended but I did try it but found in the docs that the Pascal architecture is unsupported.