Chris Wilson writes:
> We can reduce the frequency of calling madvise by only using it for
> buffers that are inactive (busy buffers will be pinned by they use by the
> GPU, so should not significantly add to mempressure, but will still be
> reaped by our timed cache.)
>
> v2: Remember about APPL
We can reduce the frequency of calling madvise by only using it for
buffers that are inactive (busy buffers will be pinned by they use by the
GPU, so should not significantly add to mempressure, but will still be
reaped by our timed cache.)
v2: Remember about APPLE_purgeable_memory that also calls
