05:23:16 <elucidator> https://www.qualcomm.com/news/releases/2023/10/qualcomm-to-bring-risc-v-based-wearable-platform-to--wear-os-by- google wants rV android as well. these are good news. 
07:00:37 <hyc> my ryzen 5900 laptop got 415H/s softAes, 706H/s hwAES for 1 thread. so that gives an indication how much these risc-v chips are missing without crypto instructions
07:29:07 <sech1> hyc is it for the original RandomX or v2?
07:29:17 <sech1> because v2 has doubled AES
07:45:34 <sech1> heh, github now has Apple M1 runners: https://github.blog/2023-10-02-introducing-the-new-apple-silicon-powered-m1-macos-larger-runner-for-github-actions/
07:46:07 <sech1> But only for Github Team & Enterprise plans (you have to pay $3.67/month at least + $0.16 per minute used by the M1 runner)
07:47:44 <sech1> I wonder how they got around Apple's licensing restrictions (the minimum time to "rent" an M1 is 24 hours)
20:42:00 <sech1> Can anyone with Apple M1 or M2 run randomx-tests from https://github.com/SChernykh/RandomX/tree/aarch64-clear-cache ? It should theoretically fix incorrect hashes with the latest macOS SDK, although it doesn't pass tests on Github runner (but maybe Github runner is just weird).
20:42:05 <sech1> hyc selsta ^
20:42:42 <selsta> will test
20:44:30 <selsta> [84] Hash test 2a (compiler)                  ... Assertion failed: (equalsHex(hash, "639183aae1bf4c9a35884cb46b09cad9175f04efd7684e7262a0ac1c2f0b4e3f")), function operator(), file tests.cpp, line 974.
20:44:53 <selsta> sech1: I can give you access to the M1 VM
20:46:10 <sech1> I think I have login/password somewhere. Are you talking about VM with IP xxx.xxx.xxx.x73 ?
20:46:44 <selsta> yes
20:48:17 <sech1> yeap, logged in
20:48:27 <sech1> I have no idea what's broken, but I'll try to fix it
20:48:42 <sech1> At least I can iterate faster than with Github CI
20:49:11 <selsta> ty, would be really nice if this can get fixed
20:52:01 <sech1> I have the code which cleans the cache properly, now it's all about understanding why even this code doesn't work :D
20:52:23 <sech1> I thought __builtin__clear_cache was somehow broken in the latest SDK, but it's something deeper
23:36:12 <hyc> sech1: my numbers were from V1
23:37:12 <hyc> the actual cache clearing code is so simple, can't imagine that part being broken
23:37:34 <hyc> maybe we need to look at the source code of something that successfully uses jit, like Safari or whatnot