--- Day changed Wed Nov 25 2015
01:54 -!- Madars [~null@unaffiliated/madars] has quit [Ping timeout: 272 seconds]
01:56 -!- Madars [~null@unaffiliated/madars] has joined #secp256k1
02:31 -!- andytoshi [~andytoshi@unaffiliated/andytoshi] has quit [Ping timeout: 255 seconds]
02:41 -!- andytoshi [~andytoshi@wpsoftware.net] has joined #secp256k1
03:31 -!- andytoshi [~andytoshi@wpsoftware.net] has quit [Ping timeout: 246 seconds]
03:52 -!- andytoshi [~andytoshi@wpsoftware.net] has joined #secp256k1
04:12 < gmaxwell> nickler: Did you end up giving klee a try?
06:08 < nickler> gmaxwell: not yet, I'll have a look after the Scaling Bitcoin conference
11:37 -!- jtimon [~quassel@74.29.134.37.dynamic.jazztel.es] has quit [Ping timeout: 265 seconds]
12:45 < gmaxwell> sipa: hm. jacobi mode is less of a speedup than I expected from the bench internal numbers.
12:52 -!- belcher [~user@unaffiliated/belcher] has joined #secp256k1
13:13 < sipa> gmaxwell: but but but... no inverse!!
13:16 < sipa> gmaxwell: but indeed, i found the same; only a factor 2 approximately (when using both inverse and jacobi from gmp)
13:22 < sipa> it does make schnorr consistently faster than ecdsa though
13:22 < sipa> even only marginally
14:28 -!- midnightmagic [~midnightm@unaffiliated/midnightmagic] has quit [Ping timeout: 252 seconds]
14:34 -!- midnightmagic [~midnightm@unaffiliated/midnightmagic] has joined #secp256k1
14:40 -!- belcher [~user@unaffiliated/belcher] has quit [Read error: Connection reset by peer]
14:40 -!- belcher [~user@unaffiliated/belcher] has joined #secp256k1
15:59 -!- jtimon [~quassel@74.29.134.37.dynamic.jazztel.es] has joined #secp256k1
16:43 < gmaxwell> ::sigh:: bench_verify now crashes on rhel because its openssl lacks secp256k1 support.
16:49 < sipa> heh, the autotools code was supposed to check for that?
16:50 < gmaxwell> I'd not noticed it on tests, so perhaps something fishy is up, will diagnose when I'm done with these benchmark runs.
17:04 < gmaxwell> I hate benchmarking.
17:05 < gmaxwell> ECDSA is faster for me.
17:05 < sipa> Inconceivable!
17:05 < gmaxwell> I know, and I increased the test rep count. to 1000.
17:06 < gmaxwell> schnorr_verify: min 63.6us / avg 63.6us / max 70.4us
17:06 < gmaxwell> ecdsa_verify: min 62.7us / avg 62.7us / max 72.1us
17:06 < gmaxwell> schnorr_verify: min 63.0us / avg 63.0us / max 64.6us
17:06 < sipa> with or without endomorphism?
17:06 < sipa> (i tested without)
17:06 < gmaxwell> with.
17:07 < gmaxwell> I'm testing with to try the worst case (all method indifferent times minimized)
17:24 < sipa> gmaxwell: can you try without endo?
17:25 < gmaxwell> yes.
17:28 < sipa> not that it matters much... it's close in any case, but it's a bit sad that's it not unconditionally faster
17:30 < gmaxwell> well that must be noise or a suggestion that our test is stupid (distored by cache effects.
17:30 < gmaxwell> )
17:32 < sipa> the gmp jacobi code looks readable and short: https://gmplib.org/repo/gmp-6.1/file/tip/mpz/jacobi.c
17:38 < andytoshi> sipa: it calls into mpn_jacobi_base :)
17:38 < andytoshi> mpz code is just thin wrappers around the mpn stuff, which is nasty
17:38 < sipa> ugh, ok
17:40 < sipa> mpn_jacobi_base also looks readable :)
17:41 < andytoshi> one sec..
17:41 < sipa> ah, but mpn_jacobi_n doesn't really
17:43 < andytoshi> yeah, mpn_jacobi_n is the nasty one
17:43 < andytoshi> but i'm curious now, i don't see where jacobi_base calls it..
17:43 < sipa> at the end
17:43 -!- jtimon [~quassel@74.29.134.37.dynamic.jazztel.es] has quit [Ping timeout: 252 seconds]
17:43 < andytoshi> ohh, i see, damn
17:44 < andytoshi> i'd have been so happy had you discovered that HGCD stuff was not even used..
18:35 -!- belcher [~user@unaffiliated/belcher] has quit [Quit: Leaving]
23:28 -!- maaku [~quassel@botbot.xen.prgmr.com] has quit [Ping timeout: 265 seconds]
23:31 -!- luke-jr_ [~luke-jr@unaffiliated/luke-jr] has joined #secp256k1
23:31 -!- wump [~quassel@pdpc/supporter/professional/wumpus] has joined #secp256k1
23:31 -!- sipa_ [~pw@2a02:348:86:3011::1] has joined #secp256k1
23:36 -!- Netsplit *.net <-> *.split quits: btcdrak
23:36 -!- Netsplit *.net <-> *.split quits: Madars, Apocalyptic
23:36 -!- Netsplit *.net <-> *.split quits: jonasschnelli, Luke-Jr
23:38 -!- Netsplit over, joins: Apocalyptic
23:40 -!- Netsplit *.net <-> *.split quits: wumpus
23:41 -!- Netsplit *.net <-> *.split quits: waxwing, fkhan, sipa
23:42 -!- fkhan [weechat@unaffiliated/loteriety] has joined #secp256k1
23:55 -!- btcdrak_ [uid115429@gateway/web/irccloud.com/session] has joined #secp256k1
23:56 -!- Netsplit *.net <-> *.split quits: evoskuil
23:56 -!- jonasschnelli [~jonasschn@unaffiliated/jonasschnelli] has joined #secp256k1
23:56 -!- Netsplit over, joins: evoskuil
23:57 -!- Netsplit over, joins: waxwing
23:59 -!- luke-jr_ is now known as Luke-Jr