You are considering adding a hardware accelerator to your current desktop machine to speedup the vector matrix operations found in machine learning code. Datasheets for the hardware accelerator claim a $10$x boost $($performance increase$)$ for code that runs on the accelerator. The percentage of code that can run on the accelerator is defined as the percentage of time spent running the original code on the CPU $($i$.$e$.$ without the accelerator$)$