00:18:58 <selsta> jberman[m]: is there an easy way to test the view tag speedup?
00:19:01 <selsta> performance_tests?
02:22:53 <UkoeHB> I rebased onto master and got `monero/external/randomx/src/virtual_memory.cpp:118:2: error: use of undeclared identifier 'pthread_jit_write_protect_np' pthread_jit_write_protect_np(false);`. Anyone know what that's about?
02:25:17 <hyc> are you building on a mac?
02:25:53 <hyc> you prob need RandomX pr#226
02:28:16 <UkoeHB> yes mac
02:29:30 <UkoeHB> thanks I'll wait for this fix
02:32:32 <selsta> yea it will be fixed before we tag a release
02:34:36 <hyc> gitian build uses MacOS SDK 10.11. is that new enough to have this function defined?
02:36:29 <hyc> guess I'll find out shortly. firing up a gitian build for mac 
02:42:38 <hyc> build finished, no complaints
02:44:15 <selsta> hyc: 10.11 is ancient, but I'm not sure why it doesn't fail
02:44:34 <selsta> is __APPLE__ defined when cross compiling?
02:44:46 <hyc> yes, it has to be
03:04:35 <hyc> looks like TARGET_OS_OSX doesn't exist in this SDK version, that's why it's working
03:04:51 <hyc> it's not trying to use that pthread_jit stuff at all
03:09:35 <hyc> so, gitian builds will keep worknig regardless. I guess that's a relief, even if it makes these last patches moot.
03:15:53 <selsta> hyc: yea it seems it got added in 10.12 SDK
03:15:57 <selsta> https://twitter.com/steipete/status/764692986223587328
03:17:22 <hyc> cool
03:17:28 <jtgrassie> yep TARGET_OS_OSX is not defined in 10.11
03:17:59 <hyc> so yeah, gitian builds are fine regardless
03:18:28 <hyc> even if maybe slower than necessary on newer macos machines
03:18:31 <jtgrassie> fine, if you don't want to support pthread_jit_write_protect_np
03:19:12 <jtgrassie> gittian really should use a later 10.x to avoid this sort of thing though
03:19:39 <hyc> why? sounds like it's using lowest common denominator, and playing it safe
03:19:41 <selsta> we still have users that use 10.11
03:20:29 <selsta> don't see any reason to cut support for older macOS versions just to use pthread_jit_write_protect_np
03:21:06 <hyc> particularly since it isn't making much performance difference on my measurements
03:21:18 <hyc> did you guys see any big difference in x86-64 benchmarks?
03:21:30 <jtgrassie> 10.12 will still run on 10.11
03:21:47 <selsta> what do we gain from 10.12? nothing
03:22:09 <jtgrassie> The defines, to avoid this kind of thing in the first place
03:22:28 <hyc> it has already been successfully avoided 
03:23:35 <jtgrassie> well, that's moot. Sure for *only* gitian builds, not for building directly from source.
03:23:47 <hyc> so again, unless any of you are seeing a big performance gain, sounds like a non-issue
03:23:56 <selsta> it would be easier to use something like this https://github.com/catchorg/Catch2/pull/2140/files
03:24:02 <selsta> instead of forcing everyone to upgrade their SDK
03:24:26 <selsta> but it doesn't matter anyway
03:25:01 <jtgrassie> that works too
03:25:02 <hyc> the defines did their job as-is. the _OSX symbol was missing, therefore we didn't try to use that API.
03:25:32 <jtgrassie> not if someone tried to build on 10.12+
03:25:37 <hyc> so unless you can point to a missed performance gain, I see no point in pursuing this further
03:27:14 <jtgrassie> if we're only going to support gitian builds for macos from now on, then yes.
03:46:56 <hyc> I take it neither of you saw any performance difference then
03:51:47 <jtgrassie> I don't have an ARM mac to test, sorry
03:55:08 <hyc> I was asking about x86-64
03:55:31 <selsta> I didn't see a performance gain on intel
03:55:41 <jtgrassie> Same
03:55:48 <hyc> I saw no difference on my M1 mac
09:19:29 <jberman[m]> <selsta> "jberman: is there an easy way to..." <- I tested it with chain data by modifying `out_can_be_to_acc` to call `derive_view_tag` for every output, then checking if the result is equal to 0x01, and proceeding to derive the output key if so, and short-circuiting if not (I tested with different equalities, got the same results)
09:19:36 <jberman[m]> I'll share the code on this in a separate commit. was pretty hacky
09:19:53 <jberman[m]> and the performance tests should give a benchmark idea too (once I fix that UB issue), but chain data seemed to perform better for me
09:43:54 <ErCiccione> Could somebody review https://github.com/erciccione/monero-site/pull/27? Adds the missing RPC methods up to v0.17.3 including RPC-Pay methods, to the docs on getmonero
20:39:41 <selsta> jberman[m]: performance tests worked now https://termbin.com/cgnd