17:06:23
jpk68:
How did 7 blocks just get mined in the past 3 minutes
17:06:44
jpk68:
Must be a crazy fluke (??)
17:10:50
hyc:
huh github says I'm ratelimited. I haven't done anything yet.
17:20:50
sech1:
Maybe github doesn't like your IP address
17:21:16
sech1:
Meanwhile, I tested RISC-V vector code on Orange Pi RV2: https://github.com/xmrig/xmrig/pull/3746#issuecomment-3708264695
17:21:38
sech1:
+2.5% only, but it's heavily limited by RAM (only 512 KB L2 per a cluster of 4 cores)
17:21:51
sech1:
I need to test it single-threaded, and with 512 KB scratchpad
17:43:38
sech1:
Did a test of a single thread with 256 KB scratchpad. 34.61 H/s scalar code, 43.43 H/s vector code. +25% speedup
17:44:22
sech1:
scalar: https://p2pool.io/u/11498f17985cb346/Screenshot%20from%202026-01-04%2018-43-22.png
17:44:41
sech1:
vector: https://p2pool.io/u/f5073be9c575b440/Screenshot%20from%202026-01-04%2018-38-44.png
17:46:13
sech1:
With 512 KB scratchpad it was 30.61 -> 38.8 H/s (+26.7%)
17:46:39
sech1:
I guess because this CPU doesn't have hardware AES, and vectorized soft AES is much faster, so it matters more for bigger scratchpads
17:52:00
sech1:
and 25.34 H/s for the regular RandomX on a single thread (2 MB scratchpad)
17:52:26
sech1:
that's vector code already ^
18:22:03
hyc:
nice