Closed zztaki closed 6 months ago
I would not worry about the 32-bit issue. It would be good to add some documentation on how to use the prefetching algorithm (maybe a few lines in the cachesim.md
and a separate prefetch.md
documentation).
If you want to add the documentation after finishing everything, it is fine as well. :)
If you want to add the documentation after finishing everything, it is fine as well. :)
Yes, I am glad to add the documentations! :)
Do you mind if I create a few PRs later to add documents related to admission and prefetching? Probably within ten hours. 😄
Sounds good! I will merge this PR for now. :)
I have reconstructed the PG prefetching algorithm, and it is worth mentioning that I use direct_hash for all hashtables, which means that I used
GINT_TO_POINTER
andGPOINTER_TO_INT
to convertobj_id
orobj_size
as the key or the value. Compared tog_int64_hash
, this can reduce the overhead of allocation, dereference, and memory release.Maybe there is a problem on a 32 bit machine, where
long
andvoid*
maybe are both 4 bytes. :(