19:17:19 <selsta> https://reddit.com/r/Monero/comments/gwn5lm/_/fsw3q8o/?context=1
19:18:08 <selsta> could someone reply here? can mining result in monerod taking up 26GB RAM and is the crash related to their CPU?
19:19:04 <sech1> My monerod is not mining and it takes 17 GB of RAM right now
19:19:11 <sech1> It has been running for weeks though
19:19:26 <selsta> mine uses less than 1GB RAM
19:20:07 <selsta> sech1: v0.16?
19:20:17 <sech1> 0.15.0.0
19:20:24 <sech1> I'm very conservative, lol
19:20:28 <selsta> that’s quite old :D
19:20:50 <sech1> if it works, don't touch it (c)
19:22:42 <moneromooo> I see no mention of a crash.
19:23:15 <selsta> they wrote > One time I saw 'error: verification failed' just before it crashed.
19:23:17 <selsta> though no logs
19:23:30 <sech1> It's probably early 1st-gen Ryzen hardware bug
19:23:36 <sech1> when mining RandomX
19:24:33 <selsta> did xmrig implemet a fix for it?
19:24:48 <sech1> yes
19:24:56 <sech1> more a workaround
19:25:13 <sech1> because it still crashes, but xmrig just skips a hash where it crashes
19:25:50 <selsta> ok well tell them to use xmrig for mining
19:26:17 <sech1> "I was trying to start xmrig to solomine for a bit to my own local deamon, but it wouldn't start mining. The problem appears to be that there's not enough RAM available."
20:37:01 <selsta> https://i.imgur.com/GrOw4oL.png
20:37:30 <selsta> from -pools
20:49:11 <selsta> gone already
21:56:31 <tevador> removing this one if improved performance by 25%: https://github.com/tevador/hashx/commit/b612cd2baa1a8d39dce682c6667c9ac5040b4253
21:56:33 <tevador> amazing
21:58:54 <hyc> nice. branches suck. did it get replaced by a conditional move or something?
21:59:02 <tevador> yep
21:59:32 <tevador> it was a typical unpredictable branch causing a lot of mispredictions
22:02:13 <hyc> looks like available is strictly a toggle between 1 and 0. could also get rid of the ?: operator and just use available * i.
22:07:59 <tevador> I checked the generated machine code, compiler uses cmovne, which seems to be pretty efficient and takes just 1 cycle compared to 3 cycles for multiplication
23:59:57 <hyc> fair enough