--- Log opened Sun Dec 01 00:00:24 2019
01:53 -!- jonatack [~jon@2a01:e35:8aba:8220:6627:dad:d967:649d] has quit [Quit: jonatack]
06:30 < elichai2> fanquake: Yeah, I discussed that with cfields_ a while ago :)
06:30 < elichai2> the problem is clang's llvm and rustc's llvm *have* to match
06:30 < elichai2> otherwise all hell breaks loose
06:30 < fanquake> Yea. Looks like you basically have to use Clang 8
06:31 < fanquake> elichai2 I've been playing around with LTO a bit. Did you have any particular issues using thinLTO ?
06:31 < elichai2> I actually had to manually compile the right llvm to get this to work. but maybe it was because I tested it on a nightly that had higher llvm version than latest clang
06:31 < elichai2> fanquake: with rust only?
06:31 < fanquake> nah compiling Bitcoin Core
06:31 < elichai2> oh. no it worked pretty good
06:31 < fanquake> I'm pretty sure you used thinLTO in your PR from a while ago
06:31 < elichai2> yeah
06:32 < elichai2> I personally am a fan of LTO, even  if it doesn't always deliver what we hope for
06:33 < fanquake> Cool. I'd like to get james to benchmark some IBD with at least a thin vs full LTO
06:33 < elichai2> fanquake: I always compile bitcoin-core with thinLTO and never had a problem
06:33 < elichai2> fanquake: that'd be awesome. I tried a while back but it's really hard to benchmark IBD on a laptop
06:33 < elichai2> too many interferences
06:33 < fanquake> Yea
06:34 < fanquake> When compiling Core are there any other particular compiler flags you've been using regularly?
06:34 < fanquake> I've been thinking about https://github.com/bitcoin/bitcoin/issues/17344 a bit, and what we might want to start turning on
06:35 < fanquake> Also, what compiler do you normally use?
06:36 < elichai2> some of my flags are highly controversial though lol
06:37 < fanquake> -ffast-math
06:37 < elichai2> changing secp to `+ac_configure_args="${ac_configure_args} --disable-shared --with-pic --with-bignum=gmp --enable-module-recovery --disable-jni --enable-endomorphism --disable-tests --disable-benchmarks --with-scalar=64bit --with-ecmult-window=24 --with-asm=x86_64 --with-field=64bit"`
06:37 < elichai2> fanquake: lol not that much hehe, and probably doesn't matter in core
06:38 < fanquake> What's the reasoning behind the secp modifications? I can understand disabling things for build speed, but why "--with-ecmult-window=24" for example. If that differs from defaults.
06:38 < elichai2> but for core itself i'm using `CC=clang CXX=clang++ LDFLAGS="-flto=thin" CXXFLAGS="-O3 -flto=thin" CFLAGS="-O3 -flto=thin" ./configure --with-incompatible-bdb --disable-debug --disable-tests --enable-hardening --enable-upnp --disable-bench --enable-werror --with-utils  --with-qrencode --enable-upnp-default --with-gui=qt5`
06:39 < elichai2> hmm for secp gmp + endo is a huge speed bump
06:39 < elichai2> and the ecmult window that's a tradeoff between memory usage and speed (default is 15)
06:40 < fanquake> Ok
06:41 < elichai2> gmp is controversial for consensus reasons. endomorphism is controversial for patent reasons
06:42 < elichai2> both of them together makes sig verification take 40% less time
06:42 < fanquake> Fair enough. I'll have to look into the secp build system / tuning a bit more.
06:42 < elichai2> (on my machine)
08:35 -!- davterra [~dulyNoded@195.242.213.120] has quit [Ping timeout: 268 seconds]
08:36 -!- davterra [~dulyNoded@195.242.213.120] has joined #bitcoin-builds
09:11 -!- Netsplit *.net <-> *.split quits: kanzure, ryanofsky, wumpus, elichai2
09:12 -!- Netsplit over, joins: kanzure, ryanofsky, elichai2, wumpus
09:12 -!- elichai2 [uid212594@gateway/web/irccloud.com/x-siqapwxxfgnswisr] has quit [Max SendQ exceeded]
09:15 -!- elichai2 [sid212594@gateway/web/irccloud.com/x-btcupeghbgvkjake] has joined #bitcoin-builds
13:20 < jonasschnelli> MarcoFalke: my nighly builds are still broken since the switch from the amd to the linux64 base image...
13:20 < jonasschnelli> I get "Could not download some packages, please run gbuild --upgrade"... any idea how to fix this?
20:24 -!- alko [~alko@cpe-85-10-28-138.static.amis.net] has joined #bitcoin-builds
20:24 -!- alko89 [~alko@cpe-85-10-28-138.static.amis.net] has quit [Ping timeout: 250 seconds]
23:25 < wumpus> jonasschnelli: is it fixed if you change the string back? if so, we should probably merge that PR for now
23:25 < wumpus> if not, there might be something wrong with the base image
23:26 < wumpus> I think that particular error happens if one of the initial apt-gets in the gitian VM fails
--- Log closed Mon Dec 02 00:00:25 2019