10-05-2016, 12:04 AM
(This post was last modified: 10-05-2016, 12:06 AM by royce.
Edit Reason: more
)
Theoretically, could you generate those kernels on an otherwise identical setup, and then copy them to the kernel cache area?
Also, if it's the kernel generation, I would expect the delay before any given benchmark would only happen once, and then that kernel would be cached.
Do you have a view into what resources the system is consuming during the long pause? (CPU? GPU? I/O?)
Also, if it's the kernel generation, I would expect the delay before any given benchmark would only happen once, and then that kernel would be cached.
Do you have a view into what resources the system is consuming during the long pause? (CPU? GPU? I/O?)
~