Jared Tobin

> Expect this to be ready to go next week. Well, almost -- I found some of that aforementioned low-hanging fruit. The change to the +feud jets about doubles the...

> Ping. Progress? Still chugging. There is an issue with memory usage; while the time differential is not so so bad, the caching compiler uses *way* too much memory as...

(There was a similar issue with the hoon-level cache leaking into the *vere*-level memo cache, which is really pathological. The solution here was to manually memoize +burp, rather than using...

Yeah, that's precisely it -- `|=(axe=@ud ..)` is the dodgy thunk in question. Will update after some probing!

@tacryt-socryp Cool, so punching the cache out of the subject prior to creating the thunk indeed seems to kill the memory leak. The disadvantage here is that, of course, the...

Some additional simple benchmarks: ``` +mint:ut (hoon.hoon, layer five) took s/42.069.090 (jetted) took s/90.689.227 (caching) max cache size: total: MB/71.655.480 (hoon.hoon, layer four) took s/7.168.708 (jetted) took s/16.708.800 (caching) max...

Ok, I've opened this up. One note re: the remark on +fond in the OP: > N.b., some work has also been done on +fond to make it more amenable...

> Tbh merge it 😂 The compiler is not in the hot path for commonly performed tasks, and it is unclean to have long-running PRs! Plus the LRU stuff is...

Ok, well what might be ideal for the time being would be to: 1) Flesh out the PSQ / LRU cache API's so that they're useful more generally, i.e. not...

> Release the PSQ / LRU stuff as a library or two so that it can just be imported trivially in whatever application. When we eventually need to move it...