Probably should look at one of the profile guided binary layout optimizers (e.g.facebook's bolt) and see how much that helps. But I think we also need to look at why our code density/locality is so shitty that we get such a huge number of faults.https://twitter.com/AndresFreundTec/status/1214305610172289024 …
-
-
I think it's more that there's traversal through quite a few layers when processing a query (network handling, protocol handling, query execution initialization, query execution, protocol handling, network), and they're all one or more C translation units.
-
There's basically no locality in how that's laid out for a query that quickly goes through all of those. The iTLB for my CPU is 128 4k entries large, which theoretically - assuming no associativity conflicts, and no HT conflicts - could cover 512kb of instructions.
- 1 more reply
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.