--- Log opened Mon May 31 00:00:24 2021 01:51 -!- thedragon [~thedragon@user/thedragon] has quit [Quit: Leaving] 02:34 -!- thedragon [~thedragon@user/thedragon] has joined #hplusroadmap 03:36 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:28fa:6bee:ad7f:96c6] has joined #hplusroadmap 07:25 -!- thedragon_ [~thedragon@user/thedragon] has joined #hplusroadmap 07:27 -!- thedragon [~thedragon@user/thedragon] has quit [Read error: Connection reset by peer] 10:57 < kanzure> .title https://github.com/stnolting/neorv32 10:57 < saxo> GitHub - stnolting/neorv32: A size-optimized, customizable full-scale 32-bit RISC-V soft-core CPU and SoC written in platform-independent VHDL. 10:57 < kanzure> "on-chip debugger (OCD) via JTGA - compatible to the Minimal RISC-V Debug Specification Version 0.13.2 and compatible with the OpenOCD and gdb" 13:31 < nmz787> fenn: Agatha: I wouldn't say STM is a chonkier AFM... they work by significantly different means 13:35 < kanzure> he meant the other STM 13:39 < fenn> yeah my sleep deprived brain just made a bizarre connection that wasn't there at all 14:19 < nmz787> kanzure: yeah, I just care less about standard template librarys 14:19 < nmz787> oh, that's STL 14:19 < nmz787> hrmm 14:19 < nmz787> I guess I have no idea what STM is in connection with Rust, the microchip company? 14:26 < kanzure> nmz787: https://en.wikipedia.org/wiki/Software_transactional_memory 15:00 < lsneff> yes, software transactional memory, not scanning tunneling microscopes 15:01 < lsneff> I've been reading through the literature 15:01 < lsneff> it seems that no one has figured out an all-around good way of doing it 15:11 -!- L29Ah [~L29Ah@user/l29ah] has quit [Remote host closed the connection] 15:14 -!- srk [~sorki@user/srk] has quit [Ping timeout: 252 seconds] 15:20 < kanzure> lsneff: whatcha need it for? 15:23 < kanzure> "In vivo partial reprogramming of myofibers promotes muscle regeneration by remodeling the stem cell niche" https://www.nature.com/articles/s41467-021-23353-z 15:30 -!- srk [~sorki@user/srk] has joined #hplusroadmap 15:45 -!- acertain [sid470584@stonehaven.irccloud.com] has quit [Ping timeout: 252 seconds] 15:45 -!- FelixWeis [sid154231@id-154231.stonehaven.irccloud.com] has quit [Read error: Connection reset by peer] 15:51 -!- FelixWeis [sid154231@id-154231.stonehaven.irccloud.com] has joined #hplusroadmap 15:51 -!- acertain [sid470584@id-470584.stonehaven.irccloud.com] has joined #hplusroadmap 16:09 -!- L29Ah [~L29Ah@user/l29ah] has joined #hplusroadmap 16:18 < maaku> lsneff: I recall hardware STM in intel chips led to some pretty severe vulnabilities and was disabled in firmware. I don't recall if they later re-enabled it. 16:19 < maaku> It seems like a pretty low-level observation from Sweeney. STM probably is the future for concurrent data structures in real-time interactive simulations 16:19 < maaku> but it's not like it drastically changes the architecture or anything. 16:26 < kanzure> concurrency is dark magic 16:27 < kanzure> in particular the atomic transactional type of concurrency 17:05 < Agatha> Personally, I like lock-free datastructures, because they have approximately zero runtime cost over your usual critical section; even a futex is twice the cost on the fast path. It's honestly not that hard either; I feel like people go about teaching concurrency the wrong way, with toy examples instead of practice. 17:07 < Agatha> Whenever I hear transactional memory, I always (unreasonably) think of near-memory transactional computing. I mention it because it has relevance to WBE and the like; solving problems like simulating the human brain requires a lot of memory, but not a lot of compute; you can get that done faster and with a lot less energy if you build circuitry into your DRAM and do stuff right there where it's needed. 17:47 -!- yashgaroth [~ffffffff@2601:5c4:c780:6aa0:28fa:6bee:ad7f:96c6] has quit [Quit: Leaving] 17:47 < nmz787> "atomic transactional concurrency" sounds like it could have come out of the Freitas gang 17:54 < lsneff> Yeah, it's not a particularly paradigm shifting observation, aside from the fact that you really need to take advantage of concurrency if you have a million players, and other approaches get complicated. 17:55 < lsneff> Really, you need zero overhead when there's no conflict, which only tm does (theoretically) 17:55 < lsneff> Sweeney said that Epic is looking into it, so I'm wondering if they've figured out a better than state of the art approach for STM 18:12 -!- Agatha [~agatha@user/agatha] has quit [Ping timeout: 264 seconds] 23:24 -!- johest [~johest@user/johest] has joined #hplusroadmap --- Log closed Tue Jun 01 00:00:25 2021