--- Log opened Sun Jan 31 00:00:32 2021
02:11 -!- sev_sam [~s3v_s4m@2601:642:c301:c180:55b6:9f7e:f33e:a0e] has quit [Remote host closed the connection]
09:57 -!- jonatack [jon@gateway/vpn/airvpn/jonatack] has quit [Ping timeout: 240 seconds]
10:00 -!- jonatack [jon@gateway/vpn/airvpn/jonatack] has joined #secp256k1
10:25 -!- belcher_ [~belcher@unaffiliated/belcher] has joined #secp256k1
10:27 -!- belcher [~belcher@unaffiliated/belcher] has quit [Ping timeout: 264 seconds]
10:31 -!- belcher_ [~belcher@unaffiliated/belcher] has quit [Ping timeout: 256 seconds]
10:31 -!- belcher_ [~belcher@unaffiliated/belcher] has joined #secp256k1
13:01 -!- belcher_ is now known as belcher
14:02 -!- virtu [~virtu@gateway/tor-sasl/virtu] has quit [Remote host closed the connection]
14:03 -!- virtu [~virtu@gateway/tor-sasl/virtu] has joined #secp256k1
15:41 -!- sev_sam [~s3v_s4m@2601:642:c301:c180:255f:6342:84df:91b0] has joined #secp256k1
16:04 < elichai2> https://twitter.com/hashbreaker/status/1355933909788225536
16:05 < elichai2> I wonder how the perf of this is compared to the current PR
16:15 < sipa> a factor 2x to 1.5x
16:15 < sipa> compared to the constant time version (with hddivsteps)
16:16 < sipa> our variable time version is closer, but not faster than his (constant time) code
16:18 < sipa> elichai2: https://gist.github.com/57c16caec7fcafca6115813330e76312
16:19 < sipa> that's djb's test/benchmark code, with our modinv module patched in
16:42 < elichai2> Damn he writes fast code (well, this is asm, but I didn't expect 2x)
16:49 < sipa> it very heavily relies on AVX2
16:50 < sipa> being able to operate on 256 bits at once tends to be significantly faster than 64 bits at a time ;)
16:51 < sipa> elichai2: also, it interleaves things in a fairly crazy way that a compiler would never figure out (it's doing the matrix computation of the next round while still applying the matrix of the previous round to the state)
16:52 < sipa> downside is that if you don't run it on an architecture that's very similar to what it's optimized for (with different latency/throughput numbers per opcode), you quickly lose a lot
18:37 -!- meshcollider [meshcollid@gateway/shell/ircnow/x-gxrhsfvhkvhshaty] has quit [Ping timeout: 265 seconds]
18:53 -!- meshcollider [meshcollid@gateway/shell/ircnow/x-euyphuspntddllnf] has joined #secp256k1
20:28 -!- sev_sam [~s3v_s4m@2601:642:c301:c180:255f:6342:84df:91b0] has quit [Read error: Connection reset by peer]
20:28 -!- sev_sam [~s3v_s4m@2601:642:c301:c180:255f:6342:84df:91b0] has joined #secp256k1
20:37 -!- sev_sam [~s3v_s4m@2601:642:c301:c180:255f:6342:84df:91b0] has quit [Read error: Connection reset by peer]
20:37 -!- sev_sam [~s3v_s4m@2601:642:c301:c180:255f:6342:84df:91b0] has joined #secp256k1
21:53 -!- lukedashjr [~luke-jr@unaffiliated/luke-jr] has joined #secp256k1
21:53 -!- luke-jr [~luke-jr@unaffiliated/luke-jr] has quit [Ping timeout: 264 seconds]
21:57 -!- lukedashjr is now known as luke-jr
22:44 -!- fjahr [sid374480@gateway/web/irccloud.com/x-rcvxpwznmxtiozvi] has quit [Ping timeout: 260 seconds]
22:45 -!- fjahr [sid374480@gateway/web/irccloud.com/x-dqspzugbuztizvad] has joined #secp256k1
--- Log closed Mon Feb 01 00:00:32 2021