UObject base layout did this in the early days. Doesn’t scale well to inheritance, though, and automatic sequential prefetching by the CPU seems to help. Really the problem with Unreal GC isn’t the amortized overhead - like 0.1%, it’s the single-threaded pause time.
-
-
This Tweet is unavailable.
-
-
-
- 5 more replies
New conversation -
-
To the first question: yes. The reference is here: https://ieeexplore.ieee.org/document/6494675/ …. There's also tricks for computing the Fourier transform on graph faster.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Unless your graph is time variant, you can express it as a state space filter: https://ccrma.stanford.edu/~jos/fp/State_Space_Filters.html … Old school control sys folks use them a lot. Once you get your graph in a state space representation you can gauge its frequency response
-
These are specifically time-varying. Discrete time, but all computers use discrete time so nbd.
- 2 more replies
New conversation -
-
-
I've wondered about this. I suspect non-deterministic GC could be a fruitful research program; it's really not a problem if a certain % of the space is wrongly marked as live at any given point, as long as the chance converges to 0 eventually
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
The FT is the SVD of a Toeplitz matrix. Maybe if your incidence matrices had a specific structure there would be a transform for it (wavelets?). But I don't think there can be a general form for "all graphs" since the set of all arbitrary matrices includes the uncompressible ones
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Spectral graph theory I guess, but I don't think that has much grip on digraphs yet.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.