Compare cpu vs gpu execution time with google benchmark

Hello, I want to compare execution time of different implementations of linear algebra algorithms, some of them use cpu libraries (e.g. Intel MKL) and some of them use gpu libraries (e.g. cuBLAS). I want also to compare how they scale according to the size of the data (vector or matrix size).
Is google benchmark a good solution ? I’v seen that they suggest a way to benchmark CUDA code : benchmark/docs/user_guide.md at main · google/benchmark · GitHub.

Does running benchmarks with this library on GPU vs CPU would give meaningfull and accurate results ?