--- Log opened Fri Mar 08 00:00:10 2019
00:27 -!- harding [~quassel@li1258-130.members.linode.com] has quit [Ping timeout: 250 seconds]
00:29 -!- harding [~quassel@li1258-130.members.linode.com] has joined #secp256k1
00:51 -!- cornfeedhobo [~cornfeedh@unaffiliated/cornfeed] has quit [Quit: ZNC - https://znc.in]
00:51 -!- cornfeedhobo [~cornfeedh@unaffiliated/cornfeed] has joined #secp256k1
08:09 -!- elichai2 [uid212594@gateway/web/irccloud.com/x-immaryhfbnctletn] has joined #secp256k1
13:25 < gmaxwell> sipa: do you have a running odroid c2 right now?   I'd like to replicate the benchmarks on https://github.com/bitcoin-core/secp256k1/pull/596 on it.. and want to know if I should go dig up one of mine.
13:26 < sipa> gmaxwell: i have one, not plugged in right no
13:26 < gmaxwell> okay I'll find mine.
13:51 -!- afk11 [~afk11@unaffiliated/afk11] has quit [Ping timeout: 252 seconds]
15:05 -!- belcher [~belcher@unaffiliated/belcher] has quit [Ping timeout: 250 seconds]
15:18 -!- belcher [~belcher@unaffiliated/belcher] has joined #secp256k1
15:23 -!- elichai2 [uid212594@gateway/web/irccloud.com/x-immaryhfbnctletn] has quit [Quit: Connection closed for inactivity]
16:22 < gmaxwell> sipa: I believe my benchmarks show we should not be reducing the endmorophism table size.
16:26 -!- belcher [~belcher@unaffiliated/belcher] has quit [Ping timeout: 268 seconds]
16:30 < gmaxwell> If anything endomorphisms benefit from larger tables is greater.
16:33 < gmaxwell> e.g. on non-endo tr2 is pretty much flatlined at 13, but with endo 17 (really 16) looks like a better choice.  In general if two numbers are really close we probably want to pick the smaller: it'll be more robust against other parallel workloads putting pressure on the cache, and will perform more consistently on smaller devices.
16:39 -!- belcher [~belcher@unaffiliated/belcher] has joined #secp256k1
18:46 < gmaxwell> would anyone hate too much if we took openssl out of the benchmarks? it's really slow. :P
20:29 < gmaxwell> exhaustive_tests have clearly never been used with arm asm, the build fails.
20:31 < gmaxwell> due to missing $(COMMON_LIB) ... I'll fix
22:11 < gmaxwell> nickler: "As an aside I can get a 10% speedup with ecdsa verify using window 10."
22:11 < gmaxwell> On what hardware?
22:11 < gmaxwell> I'm seeing nothing like that (testing arm cortex-a9 right now and also seeing small improvements from increasing it)
22:14 < gmaxwell> And similar on power9, crystal well, zen+, sandy bridge, and xeon haswell.
22:37 < midnightmagic> I wouldn't hate it!
22:37 < midnightmagic> :-D
23:21 -!- sipa [~pw@gateway/tor-sasl/sipa1024] has quit [Remote host closed the connection]
23:26 -!- sipa [~pw@gateway/tor-sasl/sipa1024] has joined #secp256k1
--- Log closed Sat Mar 09 00:00:11 2019