Christopher Berner
Christopher Berner
Ah yes, I think I understand now. Thanks for explaining it in detail! I'll try it out
Just implemented this and got a nice ~15% speedup: https://github.com/cberner/raptorq/pull/87 . Thanks for the help! Do you know of any tricks for optimizing the graph selection for r=2 in the...
Ah yep, I already keep an index of values by column to speed up those adjustments. Ya, I already tried doing graph maintenance for each operation, but that turned out...
@sleepybishop one other question, do you know if implementing this part of the RFC: `Furthermore, the row operations required for the HDPC rows may be performed for all such rows...
Cool, thanks. Ya, I'd like to implement that decoding optimization at some point. Right now I'm trying to finish up optimizing the encode path
I'm not familiar with Markowitz pivoting, do you have any pointers to how it works? And is the tradeoff that the plan compilation is faster, but the resulting plan will...
Cool, thanks! I may give that a try. I'm still thinking through some potential data structure changes to speed up the graph calculation. I feel like there might be a...
Yep, I already leverage that by selecting an edge from the longest cycle during the `r=2` selection step
I optimized the graph substep using a union-find data structure: https://github.com/cberner/raptorq/pull/95 It's now a much smaller fraction of the time
Ya, about 10% on large symbol counts. I just released it in 1.6.2. It's spread out over a fair number of things now, but the sparse FMAs are the largest...