dr|z3d
Not really, RN, could be that whatever client you're using over i2cp you closed
RN
hmmm. ok. thanks. :)
zlatinb
zzz: re GMP on M1 gmplib.org/repo/gmp/rev/5f32dbc41afc
zlatinb
looks like a fix
zlatinb
also str4d said he might look into Rust implementations of ModPow
zzz
zlatinb, that's good news
zzz
they only release like every two years, last one was a year and a half ago
zzz
you may as well check in your build changes for it. I already changed 6.2.0 to 6.2.1
zlatinb
hmm the latest tag is from 2015 looks like
zzz
Download: gmp-6.2.1.tar.lz | Release notes NEW 2020-11-14
zzz
The current stable release is 6.2.1, released 2020-11-14.
zlatinb
changes to the gmp build or mac jpackage?
zzz
gmplib.org release
zzz
what tag are you talking about?
zlatinb
oh the last one I see is 6.1 0
zlatinb
or did they move to github?
zzz
dunno, I'm just looking at their home page
zlatinb
did you want me to check in the changes I made to the jbigi build script to get it to build on m1 or to the jpackage scripts?
zzz
the jbigi build script
zlatinb
let me see what I did...
zzz
lol that's my point, before you lose it
zlatinb
MR !60
zlatinb
hmm 4 years of silence before 6.2.1 gmplib.org/list-archives/gmp-announce
zzz
so choices are we could patch, or just use 6.2.1 as is, or don't bother with jbigi
zlatinb
yeah
zlatinb
oh and we would need the cpuid lib ported too if we choose a) or b)
zzz
no, cpuid is only for x86
zlatinb
panic off :)
zzz
but might need changes in the loader code in NativeBitInteger?
zlatinb
looks like it should fall to armv8
zlatinb
until my right hand is functioning again it'd be very hard for me to try things... eche|on eche|off maybe you can help?
zzz
the test is to figure out what name NBI wants, then name it that and put it in installer/lib/jbigi, then build updaterWithJbigi, then see if it extracts it out of jbigi.jar into $I2P on restart
zlatinb
NBI wants SystemVersion.isARM() which wants os.arch startsWith "arm"/"aarch64"
zlatinb
but afaik the web console showed "(arm64)"
zlatinb
(armv8)*
zzz
should be libjbigi-osx-armv8_64.jnilib based on how the linux one is named
zzz
you can see isARM() and everything else with: java -jar i2p.jar systemversion
zlatinb
oh os.arch is aarch64 so that's good
zzz
as long as the ARM: and MAC: lines are right we should be good
zzz
and 64 bit:
zlatinb
both are correct
zlatinb
slow is false though?
zlatinb
64 bit is true
zzz
slow will be false until we have jbigi, although as I mentioned before we could special-case that if we decide not to bother
zzz
SystemVersion is great, I'm so glad we centralized all of that in one place
zlatinb
can we just not care for jbigi if jbigi.enable=false
zzz
sure, that's one way to do it
zlatinb
so modpow was 40% faster, similar to the rpi4
zlatinb
idk if that means it's worth it
zlatinb
modinv was 11 times faster
zzz
modinv is crazy fast anyway, we shouldn't make decisions based on that alone
zlatinb
when would the checkin deadline be for 1.9.0? ETA on my hand is end of July
zzz
late august
zzz
wow you really messed it up bad?
zlatinb
finger is broken entire hand is in a cast, have an operation on Monday
zzz
yikes. want to tell us how it happened, or too embarrassing? :)
zlatinb
a drunken neighbor was looking for a fight and I helped him find one :)
zzz
oof
zzz
that really sucks, sorry to hear
zzz
so you're out for the #ls2 meeting monday then
zlatinb
I hope to be back from surgery by then
zlatinb
should def. be back for the monthly one
zzz
orignal can't make it and it's 4th of july anyway, I think I'll cancel
zzz
or maybe we can do android post-mortem instead
zzz
you want to join us for that? won't require typing
zlatinb
I'm not at all up to speed
zzz
only prereq. is good software development practices
zzz
up to you
zlatinb
if i'm home by then i'll attend
zlatinb
zzz: str4d did put together a nice benchmark and the results on the M1 are very surprising - check twitter for screenshots
zzz
got a tl;dr ?
zlatinb
pure java is slower than on Ryzen 9 5950x, everything else between 15-30% faster
zlatinb
let me run them on my iMac
zlatinb
the one that just wras GMP is fastest