WEBVTT Kind: captions Language: en 00:00:00.000 --> 00:00:04.350 align:start position:0% foreign 00:00:04.350 --> 00:00:04.360 align:start position:0% 00:00:04.360 --> 00:00:16.689 align:start position:0% [Music] 00:00:16.689 --> 00:00:16.699 align:start position:0% 00:00:16.699 --> 00:00:19.370 align:start position:0% library<00:00:17.699> is<00:00:18.060> something<00:00:18.180> that<00:00:18.660> generically 00:00:19.370 --> 00:00:19.380 align:start position:0% library is something that generically 00:00:19.380 --> 00:00:21.710 align:start position:0% library is something that generically just<00:00:19.560> takes<00:00:19.920> any<00:00:20.220> message<00:00:20.520> any<00:00:20.939> a<00:00:21.240> narratory 00:00:21.710 --> 00:00:21.720 align:start position:0% just takes any message any a narratory 00:00:21.720 --> 00:00:25.070 align:start position:0% just takes any message any a narratory length<00:00:22.080> message<00:00:22.320> as<00:00:23.100> in<00:00:23.400> a<00:00:23.520> sequence<00:00:23.699> of<00:00:23.880> bits 00:00:25.070 --> 00:00:25.080 align:start position:0% length message as in a sequence of bits 00:00:25.080 --> 00:00:27.890 align:start position:0% length message as in a sequence of bits of<00:00:25.439> any<00:00:25.619> length<00:00:25.980> and<00:00:26.460> It<00:00:26.580> just<00:00:26.760> fits<00:00:27.240> out<00:00:27.480> out 00:00:27.890 --> 00:00:27.900 align:start position:0% of any length and It just fits out out 00:00:27.900 --> 00:00:30.529 align:start position:0% of any length and It just fits out out of<00:00:28.019> that<00:00:28.260> 32<00:00:28.920> bytes<00:00:29.340> the<00:00:29.820> digest<00:00:30.240> of<00:00:30.420> this 00:00:30.529 --> 00:00:30.539 align:start position:0% of that 32 bytes the digest of this 00:00:30.539 --> 00:00:33.110 align:start position:0% of that 32 bytes the digest of this thing<00:00:30.720> is<00:00:30.900> 256<00:00:31.679> bits<00:00:32.099> the<00:00:32.460> procedure<00:00:32.820> to<00:00:33.000> go 00:00:33.110 --> 00:00:33.120 align:start position:0% thing is 256 bits the procedure to go 00:00:33.120 --> 00:00:35.630 align:start position:0% thing is 256 bits the procedure to go through<00:00:33.300> those<00:00:33.540> consists<00:00:34.260> of<00:00:34.380> three<00:00:34.680> parts<00:00:34.980> so 00:00:35.630 --> 00:00:35.640 align:start position:0% through those consists of three parts so 00:00:35.640 --> 00:00:37.549 align:start position:0% through those consists of three parts so I'll<00:00:35.820> just<00:00:36.000> go<00:00:36.180> over<00:00:36.300> the<00:00:36.480> three<00:00:36.660> parts<00:00:36.840> the 00:00:37.549 --> 00:00:37.559 align:start position:0% I'll just go over the three parts the 00:00:37.559 --> 00:00:40.310 align:start position:0% I'll just go over the three parts the first<00:00:37.680> part<00:00:37.980> is<00:00:38.399> breaking<00:00:39.300> your<00:00:39.480> message<00:00:39.719> into 00:00:40.310 --> 00:00:40.320 align:start position:0% first part is breaking your message into 00:00:40.320 --> 00:00:43.430 align:start position:0% first part is breaking your message into multiples<00:00:40.980> of<00:00:41.360> 64<00:00:42.360> bytes<00:00:42.780> so<00:00:42.960> those<00:00:43.140> are<00:00:43.320> the 00:00:43.430 --> 00:00:43.440 align:start position:0% multiples of 64 bytes so those are the 00:00:43.440 --> 00:00:46.729 align:start position:0% multiples of 64 bytes so those are the chunks<00:00:43.800> 512<00:00:44.760> bits<00:00:45.180> the<00:00:45.780> second<00:00:45.899> part<00:00:46.200> consists 00:00:46.729 --> 00:00:46.739 align:start position:0% chunks 512 bits the second part consists 00:00:46.739 --> 00:00:49.850 align:start position:0% chunks 512 bits the second part consists on<00:00:46.980> scheduling<00:00:47.940> words<00:00:48.239> Computing<00:00:49.020> a<00:00:49.559> bunch<00:00:49.680> of 00:00:49.850 --> 00:00:49.860 align:start position:0% on scheduling words Computing a bunch of 00:00:49.860 --> 00:00:51.770 align:start position:0% on scheduling words Computing a bunch of double<00:00:50.100> Words<00:00:50.399> which<00:00:50.820> are<00:00:50.940> four<00:00:51.120> bytes<00:00:51.539> each 00:00:51.770 --> 00:00:51.780 align:start position:0% double Words which are four bytes each 00:00:51.780 --> 00:00:53.650 align:start position:0% double Words which are four bytes each and<00:00:52.320> the<00:00:52.440> third<00:00:52.559> part<00:00:52.800> consists<00:00:53.399> on 00:00:53.650 --> 00:00:53.660 align:start position:0% and the third part consists on 00:00:53.660 --> 00:00:56.029 align:start position:0% and the third part consists on performing<00:00:54.660> a<00:00:54.840> bunch<00:00:54.960> of<00:00:55.079> rounds<00:00:55.440> which<00:00:55.920> is 00:00:56.029 --> 00:00:56.039 align:start position:0% performing a bunch of rounds which is 00:00:56.039 --> 00:00:57.470 align:start position:0% performing a bunch of rounds which is just<00:00:56.219> the<00:00:56.340> function<00:00:56.640> that<00:00:56.820> actually<00:00:57.000> does<00:00:57.239> the 00:00:57.470 --> 00:00:57.480 align:start position:0% just the function that actually does the 00:00:57.480 --> 00:00:59.450 align:start position:0% just the function that actually does the hashing<00:00:57.840> itself<00:00:58.140> the<00:00:58.500> function<00:00:58.920> that<00:00:59.100> is<00:00:59.280> not 00:00:59.450 --> 00:00:59.460 align:start position:0% hashing itself the function that is not 00:00:59.460 --> 00:01:01.610 align:start position:0% hashing itself the function that is not invertible<00:01:00.000> here's<00:01:00.539> just<00:01:00.719> a<00:01:00.899> brief 00:01:01.610 --> 00:01:01.620 align:start position:0% invertible here's just a brief 00:01:01.620 --> 00:01:04.070 align:start position:0% invertible here's just a brief description<00:01:02.460> of<00:01:02.760> this<00:01:03.300> this<00:01:03.420> slide<00:01:03.780> I<00:01:03.960> just 00:01:04.070 --> 00:01:04.080 align:start position:0% description of this this slide I just 00:01:04.080 --> 00:01:06.590 align:start position:0% description of this this slide I just stole<00:01:04.500> from<00:01:04.680> somewhere<00:01:04.979> in<00:01:05.159> the<00:01:05.280> internet<00:01:05.600> but 00:01:06.590 --> 00:01:06.600 align:start position:0% stole from somewhere in the internet but 00:01:06.600 --> 00:01:09.230 align:start position:0% stole from somewhere in the internet but I<00:01:07.320> want<00:01:07.380> to<00:01:07.560> go<00:01:07.740> in<00:01:07.979> some<00:01:08.280> detail<00:01:08.640> in<00:01:08.939> each<00:01:09.060> one 00:01:09.230 --> 00:01:09.240 align:start position:0% I want to go in some detail in each one 00:01:09.240 --> 00:01:11.690 align:start position:0% I want to go in some detail in each one of<00:01:09.360> the<00:01:09.479> three<00:01:09.659> parts<00:01:09.840> so<00:01:10.740> let's<00:01:10.920> start<00:01:11.220> with 00:01:11.690 --> 00:01:11.700 align:start position:0% of the three parts so let's start with 00:01:11.700 --> 00:01:14.030 align:start position:0% of the three parts so let's start with the<00:01:11.820> easiest<00:01:12.119> one<00:01:12.420> which<00:01:12.900> is<00:01:13.080> scheduling 00:01:14.030 --> 00:01:14.040 align:start position:0% the easiest one which is scheduling 00:01:14.040 --> 00:01:14.990 align:start position:0% the easiest one which is scheduling words 00:01:14.990 --> 00:01:15.000 align:start position:0% words 00:01:15.000 --> 00:01:16.969 align:start position:0% words so<00:01:15.659> scheduling<00:01:16.080> words<00:01:16.260> as<00:01:16.500> you<00:01:16.619> see<00:01:16.799> here 00:01:16.969 --> 00:01:16.979 align:start position:0% so scheduling words as you see here 00:01:16.979 --> 00:01:19.490 align:start position:0% so scheduling words as you see here you're<00:01:17.700> given<00:01:18.119> a<00:01:18.240> message<00:01:18.479> let's<00:01:19.140> suppose 00:01:19.490 --> 00:01:19.500 align:start position:0% you're given a message let's suppose 00:01:19.500 --> 00:01:21.830 align:start position:0% you're given a message let's suppose that<00:01:19.619> we've<00:01:19.860> already<00:01:20.040> broken<00:01:20.400> the<00:01:20.759> message<00:01:20.939> in 00:01:21.830 --> 00:01:21.840 align:start position:0% that we've already broken the message in 00:01:21.840 --> 00:01:25.010 align:start position:0% that we've already broken the message in pieces<00:01:22.500> each<00:01:22.860> one<00:01:22.979> of<00:01:23.100> them<00:01:23.220> is<00:01:23.400> 64<00:01:23.759> bytes 00:01:25.010 --> 00:01:25.020 align:start position:0% pieces each one of them is 64 bytes 00:01:25.020 --> 00:01:27.490 align:start position:0% pieces each one of them is 64 bytes so<00:01:25.380> the<00:01:25.500> first<00:01:25.619> thing<00:01:25.740> you<00:01:25.979> do<00:01:26.220> is<00:01:26.759> you<00:01:27.000> compute 00:01:27.490 --> 00:01:27.500 align:start position:0% so the first thing you do is you compute 00:01:27.500 --> 00:01:31.370 align:start position:0% so the first thing you do is you compute 48<00:01:28.500> double<00:01:28.979> words<00:01:29.340> four<00:01:30.060> bytes<00:01:30.479> each<00:01:30.780> the 00:01:31.370 --> 00:01:31.380 align:start position:0% 48 double words four bytes each the 00:01:31.380 --> 00:01:33.950 align:start position:0% 48 double words four bytes each the first<00:01:31.560> you<00:01:31.860> need<00:01:32.100> 64<00:01:32.640> in<00:01:32.880> total<00:01:33.119> the<00:01:33.479> first<00:01:33.659> 16 00:01:33.950 --> 00:01:33.960 align:start position:0% first you need 64 in total the first 16 00:01:33.960 --> 00:01:36.770 align:start position:0% first you need 64 in total the first 16 are<00:01:34.619> your<00:01:34.860> your<00:01:35.159> message<00:01:35.460> are<00:01:35.880> the<00:01:36.060> 64<00:01:36.420> bytes 00:01:36.770 --> 00:01:36.780 align:start position:0% are your your message are the 64 bytes 00:01:36.780 --> 00:01:38.510 align:start position:0% are your your message are the 64 bytes that<00:01:36.900> you're<00:01:37.020> giving<00:01:37.200> these<00:01:37.860> are<00:01:37.979> 16<00:01:38.159> words 00:01:38.510 --> 00:01:38.520 align:start position:0% that you're giving these are 16 words 00:01:38.520 --> 00:01:40.730 align:start position:0% that you're giving these are 16 words with<00:01:39.180> those<00:01:39.360> 16<00:01:39.600> words<00:01:39.960> you<00:01:40.200> can<00:01:40.320> compute<00:01:40.619> the 00:01:40.730 --> 00:01:40.740 align:start position:0% with those 16 words you can compute the 00:01:40.740 --> 00:01:43.609 align:start position:0% with those 16 words you can compute the next<00:01:40.860> 16<00:01:41.100> words<00:01:41.479> and<00:01:42.479> with<00:01:42.900> those<00:01:43.079> 16<00:01:43.259> words 00:01:43.609 --> 00:01:43.619 align:start position:0% next 16 words and with those 16 words 00:01:43.619 --> 00:01:45.170 align:start position:0% next 16 words and with those 16 words that<00:01:43.860> you<00:01:43.979> just<00:01:44.100> compare<00:01:44.400> to<00:01:44.579> come<00:01:44.700> to<00:01:45.060> the 00:01:45.170 --> 00:01:45.180 align:start position:0% that you just compare to come to the 00:01:45.180 --> 00:01:46.730 align:start position:0% that you just compare to come to the next<00:01:45.299> ones<00:01:45.659> and<00:01:45.900> the<00:01:46.079> next<00:01:46.140> one<00:01:46.320> and<00:01:46.619> you 00:01:46.730 --> 00:01:46.740 align:start position:0% next ones and the next one and you 00:01:46.740 --> 00:01:49.249 align:start position:0% next ones and the next one and you compare<00:01:47.040> the<00:01:47.220> 64<00:01:47.579> that<00:01:47.820> you<00:01:47.939> needed<00:01:48.600> and<00:01:49.140> the 00:01:49.249 --> 00:01:49.259 align:start position:0% compare the 64 that you needed and the 00:01:49.259 --> 00:01:50.149 align:start position:0% compare the 64 that you needed and the important<00:01:49.500> thing<00:01:49.680> that<00:01:49.860> you<00:01:49.979> need<00:01:50.040> to 00:01:50.149 --> 00:01:50.159 align:start position:0% important thing that you need to 00:01:50.159 --> 00:01:52.370 align:start position:0% important thing that you need to remember<00:01:50.340> from<00:01:50.700> this<00:01:50.880> slide<00:01:51.119> is<00:01:51.659> that<00:01:51.960> to 00:01:52.370 --> 00:01:52.380 align:start position:0% remember from this slide is that to 00:01:52.380 --> 00:01:54.830 align:start position:0% remember from this slide is that to compute<00:01:52.740> those<00:01:53.040> scheduled<00:01:53.820> words<00:01:54.060> the<00:01:54.720> only 00:01:54.830 --> 00:01:54.840 align:start position:0% compute those scheduled words the only 00:01:54.840 --> 00:01:56.690 align:start position:0% compute those scheduled words the only thing<00:01:55.079> that<00:01:55.320> you<00:01:55.439> need<00:01:55.560> is<00:01:56.220> the<00:01:56.579> previous 00:01:56.690 --> 00:01:56.700 align:start position:0% thing that you need is the previous 00:01:56.700 --> 00:01:58.609 align:start position:0% thing that you need is the previous words<00:01:57.119> nothing<00:01:57.780> else 00:01:58.609 --> 00:01:58.619 align:start position:0% words nothing else 00:01:58.619 --> 00:02:00.710 align:start position:0% words nothing else that<00:01:59.159> means<00:01:59.460> that 00:02:00.710 --> 00:02:00.720 align:start position:0% that means that 00:02:00.720 --> 00:02:03.649 align:start position:0% that means that when<00:02:01.259> you're<00:02:01.380> scheduling<00:02:01.860> words<00:02:02.479> well<00:02:03.479> I 00:02:03.649 --> 00:02:03.659 align:start position:0% when you're scheduling words well I 00:02:03.659 --> 00:02:05.450 align:start position:0% when you're scheduling words well I think<00:02:03.780> the<00:02:03.899> consensus<00:02:04.320> people<00:02:04.920> are<00:02:05.159> starting 00:02:05.450 --> 00:02:05.460 align:start position:0% think the consensus people are starting 00:02:05.460 --> 00:02:06.590 align:start position:0% think the consensus people are starting to<00:02:05.700> come<00:02:05.880> in 00:02:06.590 --> 00:02:06.600 align:start position:0% to come in 00:02:06.600 --> 00:02:09.229 align:start position:0% to come in anyways<00:02:07.320> I've<00:02:07.860> already<00:02:08.099> started<00:02:08.520> guys 00:02:09.229 --> 00:02:09.239 align:start position:0% anyways I've already started guys 00:02:09.239 --> 00:02:11.570 align:start position:0% anyways I've already started guys uh<00:02:10.080> so<00:02:10.380> when<00:02:10.500> you're<00:02:10.619> scheduling<00:02:11.039> words<00:02:11.220> the 00:02:11.570 --> 00:02:11.580 align:start position:0% uh so when you're scheduling words the 00:02:11.580 --> 00:02:13.729 align:start position:0% uh so when you're scheduling words the important<00:02:11.819> thing<00:02:12.000> is<00:02:12.239> that 00:02:13.729 --> 00:02:13.739 align:start position:0% important thing is that 00:02:13.739 --> 00:02:15.890 align:start position:0% important thing is that you<00:02:14.400> don't<00:02:14.580> need<00:02:14.819> to<00:02:15.000> know<00:02:15.239> the<00:02:15.720> previous 00:02:15.890 --> 00:02:15.900 align:start position:0% you don't need to know the previous 00:02:15.900 --> 00:02:18.770 align:start position:0% you don't need to know the previous state<00:02:16.260> of<00:02:16.560> the<00:02:16.739> Hashi<00:02:17.220> you<00:02:18.060> only<00:02:18.180> need<00:02:18.480> to<00:02:18.599> know 00:02:18.770 --> 00:02:18.780 align:start position:0% state of the Hashi you only need to know 00:02:18.780 --> 00:02:20.869 align:start position:0% state of the Hashi you only need to know what<00:02:19.379> were<00:02:19.620> the<00:02:19.860> previous<00:02:20.040> scheduled<00:02:20.700> words 00:02:20.869 --> 00:02:20.879 align:start position:0% what were the previous scheduled words 00:02:20.879 --> 00:02:24.110 align:start position:0% what were the previous scheduled words in<00:02:21.780> particular<00:02:22.080> it<00:02:22.980> only<00:02:23.099> depends<00:02:23.640> on<00:02:23.940> the 00:02:24.110 --> 00:02:24.120 align:start position:0% in particular it only depends on the 00:02:24.120 --> 00:02:25.910 align:start position:0% in particular it only depends on the chunk<00:02:24.300> that<00:02:24.720> you're<00:02:24.840> hashing<00:02:25.260> now<00:02:25.440> and<00:02:25.739> it 00:02:25.910 --> 00:02:25.920 align:start position:0% chunk that you're hashing now and it 00:02:25.920 --> 00:02:27.350 align:start position:0% chunk that you're hashing now and it doesn't<00:02:26.099> depend<00:02:26.400> on<00:02:26.520> the<00:02:26.700> other<00:02:26.760> chunks<00:02:27.239> that 00:02:27.350 --> 00:02:27.360 align:start position:0% doesn't depend on the other chunks that 00:02:27.360 --> 00:02:29.030 align:start position:0% doesn't depend on the other chunks that you're<00:02:27.480> going<00:02:27.660> to<00:02:27.780> hatch<00:02:28.020> so<00:02:28.500> if<00:02:28.680> your<00:02:28.800> message 00:02:29.030 --> 00:02:29.040 align:start position:0% you're going to hatch so if your message 00:02:29.040 --> 00:02:32.150 align:start position:0% you're going to hatch so if your message consists<00:02:29.819> of<00:02:30.000> 10<00:02:30.780> 000<00:02:30.840> chunks<00:02:31.379> you<00:02:32.040> can 00:02:32.150 --> 00:02:32.160 align:start position:0% consists of 10 000 chunks you can 00:02:32.160 --> 00:02:34.490 align:start position:0% consists of 10 000 chunks you can compute<00:02:32.760> the<00:02:33.120> scheduled<00:02:33.540> words<00:02:33.720> for<00:02:34.020> the<00:02:34.319> ten 00:02:34.490 --> 00:02:34.500 align:start position:0% compute the scheduled words for the ten 00:02:34.500 --> 00:02:37.010 align:start position:0% compute the scheduled words for the ten times<00:02:34.739> and<00:02:34.920> challenge<00:02:35.160> without<00:02:36.060> caring<00:02:36.660> about 00:02:37.010 --> 00:02:37.020 align:start position:0% times and challenge without caring about 00:02:37.020 --> 00:02:39.170 align:start position:0% times and challenge without caring about how<00:02:37.440> you<00:02:37.560> attach<00:02:37.980> each<00:02:38.340> one<00:02:38.520> of<00:02:38.640> them<00:02:38.760> and 00:02:39.170 --> 00:02:39.180 align:start position:0% how you attach each one of them and 00:02:39.180 --> 00:02:41.270 align:start position:0% how you attach each one of them and without<00:02:39.300> caring<00:02:39.720> about<00:02:39.900> the<00:02:40.080> rounds<00:02:40.379> part<00:02:40.560> so 00:02:41.270 --> 00:02:41.280 align:start position:0% without caring about the rounds part so 00:02:41.280 --> 00:02:43.070 align:start position:0% without caring about the rounds part so scheduling<00:02:41.700> words<00:02:41.940> only<00:02:42.300> requires<00:02:42.840> the 00:02:43.070 --> 00:02:43.080 align:start position:0% scheduling words only requires the 00:02:43.080 --> 00:02:45.710 align:start position:0% scheduling words only requires the previously<00:02:43.800> scheduled<00:02:44.519> words<00:02:44.760> so<00:02:45.480> the 00:02:45.710 --> 00:02:45.720 align:start position:0% previously scheduled words so the 00:02:45.720 --> 00:02:47.570 align:start position:0% previously scheduled words so the diagram<00:02:46.140> there<00:02:46.440> is<00:02:46.680> taken<00:02:47.040> out<00:02:47.160> of<00:02:47.340> an<00:02:47.459> Angel 00:02:47.570 --> 00:02:47.580 align:start position:0% diagram there is taken out of an Angel 00:02:47.580 --> 00:02:50.509 align:start position:0% diagram there is taken out of an Angel paper<00:02:48.060> describing<00:02:48.900> what<00:02:49.379> were<00:02:49.620> two<00:02:50.280> new 00:02:50.509 --> 00:02:50.519 align:start position:0% paper describing what were two new 00:02:50.519 --> 00:02:53.030 align:start position:0% paper describing what were two new instructions<00:02:51.060> that<00:02:51.360> do<00:02:51.599> this<00:02:51.900> scheduling<00:02:52.739> of 00:02:53.030 --> 00:02:53.040 align:start position:0% instructions that do this scheduling of 00:02:53.040 --> 00:02:54.949 align:start position:0% instructions that do this scheduling of four<00:02:53.280> words<00:02:53.519> at<00:02:53.819> a<00:02:53.940> time<00:02:54.060> with<00:02:54.540> only<00:02:54.720> two 00:02:54.949 --> 00:02:54.959 align:start position:0% four words at a time with only two 00:02:54.959 --> 00:02:56.930 align:start position:0% four words at a time with only two instructions<00:02:55.440> this<00:02:55.980> can<00:02:56.160> be<00:02:56.280> done<00:02:56.519> now<00:02:56.760> on 00:02:56.930 --> 00:02:56.940 align:start position:0% instructions this can be done now on 00:02:56.940 --> 00:02:59.390 align:start position:0% instructions this can be done now on Modern<00:02:57.180> CPUs<00:02:57.780> and<00:02:58.379> this<00:02:58.560> is<00:02:58.620> just<00:02:58.800> a<00:02:58.980> sketch 00:02:59.390 --> 00:02:59.400 align:start position:0% Modern CPUs and this is just a sketch 00:02:59.400 --> 00:03:01.790 align:start position:0% Modern CPUs and this is just a sketch that<00:02:59.819> how<00:03:00.239> diagram<00:03:00.660> diagrammatically<00:03:01.620> you 00:03:01.790 --> 00:03:01.800 align:start position:0% that how diagram diagrammatically you 00:03:01.800 --> 00:03:04.369 align:start position:0% that how diagram diagrammatically you compute<00:03:02.340> four<00:03:02.819> words<00:03:03.000> at<00:03:03.300> a<00:03:03.480> time<00:03:03.599> but<00:03:04.200> it's 00:03:04.369 --> 00:03:04.379 align:start position:0% compute four words at a time but it's 00:03:04.379 --> 00:03:06.170 align:start position:0% compute four words at a time but it's irrelevant<00:03:04.980> the<00:03:05.220> method<00:03:05.519> that<00:03:05.760> you<00:03:05.819> use<00:03:06.000> to 00:03:06.170 --> 00:03:06.180 align:start position:0% irrelevant the method that you use to 00:03:06.180 --> 00:03:07.910 align:start position:0% irrelevant the method that you use to compute<00:03:06.480> those<00:03:06.599> scheduled<00:03:07.019> words<00:03:07.200> what<00:03:07.800> I 00:03:07.910 --> 00:03:07.920 align:start position:0% compute those scheduled words what I 00:03:07.920 --> 00:03:10.009 align:start position:0% compute those scheduled words what I want<00:03:08.040> you<00:03:08.160> to<00:03:08.280> remember<00:03:08.400> is<00:03:08.819> that<00:03:09.120> to<00:03:09.660> compute 00:03:10.009 --> 00:03:10.019 align:start position:0% want you to remember is that to compute 00:03:10.019 --> 00:03:11.690 align:start position:0% want you to remember is that to compute them<00:03:10.200> you<00:03:10.500> need<00:03:10.620> to<00:03:10.739> know<00:03:10.920> the<00:03:11.280> previous<00:03:11.400> words 00:03:11.690 --> 00:03:11.700 align:start position:0% them you need to know the previous words 00:03:11.700 --> 00:03:14.390 align:start position:0% them you need to know the previous words and<00:03:11.879> nothing<00:03:12.120> else<00:03:12.360> rounds<00:03:13.680> we<00:03:14.159> don't<00:03:14.280> care 00:03:14.390 --> 00:03:14.400 align:start position:0% and nothing else rounds we don't care 00:03:14.400 --> 00:03:16.490 align:start position:0% and nothing else rounds we don't care about<00:03:14.580> what<00:03:14.879> rounds<00:03:15.300> is<00:03:15.480> we<00:03:15.900> only<00:03:16.080> care<00:03:16.319> that 00:03:16.490 --> 00:03:16.500 align:start position:0% about what rounds is we only care that 00:03:16.500 --> 00:03:18.290 align:start position:0% about what rounds is we only care that it's<00:03:16.739> a<00:03:16.920> function<00:03:17.159> that<00:03:17.580> is<00:03:17.700> not<00:03:17.819> invertible 00:03:18.290 --> 00:03:18.300 align:start position:0% it's a function that is not invertible 00:03:18.300 --> 00:03:20.030 align:start position:0% it's a function that is not invertible that<00:03:18.720> is<00:03:18.840> hard<00:03:19.019> that<00:03:19.440> is<00:03:19.620> computationally 00:03:20.030 --> 00:03:20.040 align:start position:0% that is hard that is computationally 00:03:20.040 --> 00:03:23.030 align:start position:0% that is hard that is computationally hard<00:03:20.400> to<00:03:20.580> invert<00:03:20.940> but<00:03:21.620> the<00:03:22.620> important<00:03:22.920> thing 00:03:23.030 --> 00:03:23.040 align:start position:0% hard to invert but the important thing 00:03:23.040 --> 00:03:25.550 align:start position:0% hard to invert but the important thing that<00:03:23.220> we<00:03:23.340> need<00:03:23.459> to<00:03:23.580> remember<00:03:23.760> is<00:03:24.120> this<00:03:24.560> we 00:03:25.550 --> 00:03:25.560 align:start position:0% that we need to remember is this we 00:03:25.560 --> 00:03:27.710 align:start position:0% that we need to remember is this we already<00:03:26.099> we<00:03:26.519> were<00:03:26.819> given<00:03:27.120> the<00:03:27.239> message<00:03:27.360> we 00:03:27.710 --> 00:03:27.720 align:start position:0% already we were given the message we 00:03:27.720 --> 00:03:29.690 align:start position:0% already we were given the message we broke<00:03:27.959> it<00:03:28.080> into<00:03:28.200> pieces<00:03:28.500> of<00:03:28.680> 64<00:03:29.040> bytes<00:03:29.459> that<00:03:29.580> I 00:03:29.690 --> 00:03:29.700 align:start position:0% broke it into pieces of 64 bytes that I 00:03:29.700 --> 00:03:32.030 align:start position:0% broke it into pieces of 64 bytes that I haven't<00:03:29.760> told<00:03:29.940> you<00:03:30.060> how<00:03:30.239> we<00:03:30.959> computed<00:03:31.500> those 00:03:32.030 --> 00:03:32.040 align:start position:0% haven't told you how we computed those 00:03:32.040 --> 00:03:34.190 align:start position:0% haven't told you how we computed those scheduled<00:03:32.519> words<00:03:32.640> that<00:03:33.060> we<00:03:33.239> need<00:03:33.360> and<00:03:34.080> then 00:03:34.190 --> 00:03:34.200 align:start position:0% scheduled words that we need and then 00:03:34.200 --> 00:03:36.649 align:start position:0% scheduled words that we need and then what<00:03:34.440> we<00:03:34.620> do<00:03:34.800> is<00:03:34.920> we<00:03:35.159> pass 00:03:36.649 --> 00:03:36.659 align:start position:0% what we do is we pass 00:03:36.659 --> 00:03:38.630 align:start position:0% what we do is we pass an<00:03:37.140> incoming<00:03:37.620> Digest 00:03:38.630 --> 00:03:38.640 align:start position:0% an incoming Digest 00:03:38.640 --> 00:03:40.490 align:start position:0% an incoming Digest if<00:03:39.180> it's<00:03:39.300> the<00:03:39.540> very<00:03:39.720> first<00:03:39.840> chunk<00:03:40.080> that<00:03:40.379> you're 00:03:40.490 --> 00:03:40.500 align:start position:0% if it's the very first chunk that you're 00:03:40.500 --> 00:03:42.649 align:start position:0% if it's the very first chunk that you're slash<00:03:40.920> that<00:03:41.159> you're<00:03:41.280> hashing<00:03:41.760> we<00:03:42.239> pass<00:03:42.480> a 00:03:42.649 --> 00:03:42.659 align:start position:0% slash that you're hashing we pass a 00:03:42.659 --> 00:03:45.229 align:start position:0% slash that you're hashing we pass a constant<00:03:42.959> digest<00:03:43.379> that<00:03:43.560> the<00:03:43.739> method<00:03:43.980> has<00:03:44.239> if 00:03:45.229 --> 00:03:45.239 align:start position:0% constant digest that the method has if 00:03:45.239 --> 00:03:47.089 align:start position:0% constant digest that the method has if it's<00:03:45.480> the<00:03:45.780> third<00:03:45.900> chunk<00:03:46.260> you're<00:03:46.560> gonna<00:03:46.739> pass 00:03:47.089 --> 00:03:47.099 align:start position:0% it's the third chunk you're gonna pass 00:03:47.099 --> 00:03:50.750 align:start position:0% it's the third chunk you're gonna pass the<00:03:48.180> past<00:03:48.959> that<00:03:49.379> the<00:03:49.560> second<00:03:49.799> chunk<00:03:50.159> produced 00:03:50.750 --> 00:03:50.760 align:start position:0% the past that the second chunk produced 00:03:50.760 --> 00:03:52.610 align:start position:0% the past that the second chunk produced the<00:03:51.420> point<00:03:51.540> is<00:03:51.720> that<00:03:51.900> you<00:03:52.080> have<00:03:52.260> a<00:03:52.440> status 00:03:52.610 --> 00:03:52.620 align:start position:0% the point is that you have a status 00:03:52.620 --> 00:03:55.490 align:start position:0% the point is that you have a status which<00:03:53.099> is<00:03:53.220> your<00:03:53.400> current<00:03:53.580> hash<00:03:53.959> and<00:03:54.959> you<00:03:55.379> pass 00:03:55.490 --> 00:03:55.500 align:start position:0% which is your current hash and you pass 00:03:55.500 --> 00:03:57.410 align:start position:0% which is your current hash and you pass it<00:03:55.680> through<00:03:55.860> this<00:03:56.040> function<00:03:56.340> that<00:03:56.640> takes<00:03:56.879> this 00:03:57.410 --> 00:03:57.420 align:start position:0% it through this function that takes this 00:03:57.420 --> 00:04:00.229 align:start position:0% it through this function that takes this status<00:03:57.599> the<00:03:58.200> hash<00:03:58.379> the<00:03:58.799> 32<00:03:59.099> bytes<00:03:59.459> it<00:04:00.060> takes 00:04:00.229 --> 00:04:00.239 align:start position:0% status the hash the 32 bytes it takes 00:04:00.239 --> 00:04:02.390 align:start position:0% status the hash the 32 bytes it takes one<00:04:00.840> of<00:04:01.019> the<00:04:01.140> scheduled<00:04:01.620> words<00:04:01.799> that<00:04:02.280> you 00:04:02.390 --> 00:04:02.400 align:start position:0% one of the scheduled words that you 00:04:02.400 --> 00:04:05.149 align:start position:0% one of the scheduled words that you computed<00:04:02.819> and<00:04:03.599> it<00:04:03.780> takes<00:04:04.080> another<00:04:04.560> constant 00:04:05.149 --> 00:04:05.159 align:start position:0% computed and it takes another constant 00:04:05.159 --> 00:04:07.369 align:start position:0% computed and it takes another constant word<00:04:05.400> that<00:04:05.760> the<00:04:06.000> protocol<00:04:06.360> has 00:04:07.369 --> 00:04:07.379 align:start position:0% word that the protocol has 00:04:07.379 --> 00:04:09.110 align:start position:0% word that the protocol has so<00:04:07.739> it<00:04:07.799> takes<00:04:07.920> this<00:04:08.159> data<00:04:08.519> and<00:04:08.700> it<00:04:08.760> produces 00:04:09.110 --> 00:04:09.120 align:start position:0% so it takes this data and it produces 00:04:09.120 --> 00:04:11.509 align:start position:0% so it takes this data and it produces for<00:04:09.299> you<00:04:09.420> a<00:04:09.540> new<00:04:09.720> hash<00:04:10.019> so<00:04:10.920> what<00:04:11.159> do<00:04:11.340> you<00:04:11.459> need 00:04:11.509 --> 00:04:11.519 align:start position:0% for you a new hash so what do you need 00:04:11.519 --> 00:04:13.250 align:start position:0% for you a new hash so what do you need to<00:04:11.640> remember<00:04:11.760> from<00:04:12.120> this<00:04:12.299> is<00:04:12.599> that<00:04:12.780> you<00:04:13.140> need 00:04:13.250 --> 00:04:13.260 align:start position:0% to remember from this is that you need 00:04:13.260 --> 00:04:15.350 align:start position:0% to remember from this is that you need to<00:04:13.379> have<00:04:13.560> computed<00:04:14.159> at<00:04:14.400> least<00:04:14.580> that<00:04:14.939> scheduled 00:04:15.350 --> 00:04:15.360 align:start position:0% to have computed at least that scheduled 00:04:15.360 --> 00:04:16.310 align:start position:0% to have computed at least that scheduled word 00:04:16.310 --> 00:04:16.320 align:start position:0% word 00:04:16.320 --> 00:04:18.949 align:start position:0% word before<00:04:16.919> passing<00:04:17.400> through<00:04:17.519> the<00:04:17.699> rounds<00:04:18.060> and 00:04:18.949 --> 00:04:18.959 align:start position:0% before passing through the rounds and 00:04:18.959 --> 00:04:21.830 align:start position:0% before passing through the rounds and you<00:04:19.380> cannot<00:04:19.680> do<00:04:19.919> this<00:04:20.100> in<00:04:20.280> parallel<00:04:20.639> to<00:04:21.419> pass 00:04:21.830 --> 00:04:21.840 align:start position:0% you cannot do this in parallel to pass 00:04:21.840 --> 00:04:23.330 align:start position:0% you cannot do this in parallel to pass through<00:04:22.139> the<00:04:22.320> round<00:04:22.500> you<00:04:22.979> need<00:04:23.100> to<00:04:23.160> have 00:04:23.330 --> 00:04:23.340 align:start position:0% through the round you need to have 00:04:23.340 --> 00:04:25.310 align:start position:0% through the round you need to have computer<00:04:23.520> already<00:04:23.940> in<00:04:24.180> hash<00:04:24.419> before<00:04:24.660> so<00:04:25.199> this 00:04:25.310 --> 00:04:25.320 align:start position:0% computer already in hash before so this 00:04:25.320 --> 00:04:26.510 align:start position:0% computer already in hash before so this is<00:04:25.440> something<00:04:25.560> that<00:04:25.800> you<00:04:25.919> cannot<00:04:26.160> do<00:04:26.340> in 00:04:26.510 --> 00:04:26.520 align:start position:0% is something that you cannot do in 00:04:26.520 --> 00:04:27.469 align:start position:0% is something that you cannot do in paragraph 00:04:27.469 --> 00:04:27.479 align:start position:0% paragraph 00:04:27.479 --> 00:04:30.469 align:start position:0% paragraph okay<00:04:28.020> so<00:04:28.500> the<00:04:28.620> padding<00:04:28.979> block<00:04:29.160> so<00:04:29.699> this<00:04:29.940> is<00:04:30.060> the 00:04:30.469 --> 00:04:30.479 align:start position:0% okay so the padding block so this is the 00:04:30.479 --> 00:04:33.770 align:start position:0% okay so the padding block so this is the first<00:04:31.020> part<00:04:31.259> of<00:04:31.440> the<00:04:31.500> three<00:04:31.919> uh<00:04:32.820> process<00:04:33.419> which 00:04:33.770 --> 00:04:33.780 align:start position:0% first part of the three uh process which 00:04:33.780 --> 00:04:35.870 align:start position:0% first part of the three uh process which is<00:04:33.960> breaking<00:04:34.680> the<00:04:34.860> message<00:04:35.040> into<00:04:35.460> multiples 00:04:35.870 --> 00:04:35.880 align:start position:0% is breaking the message into multiples 00:04:35.880 --> 00:04:38.450 align:start position:0% is breaking the message into multiples of<00:04:35.940> 64<00:04:36.360> bytes<00:04:36.840> how<00:04:37.259> do<00:04:37.380> you<00:04:37.440> do<00:04:37.620> this<00:04:37.800> well<00:04:38.220> you 00:04:38.450 --> 00:04:38.460 align:start position:0% of 64 bytes how do you do this well you 00:04:38.460 --> 00:04:39.170 align:start position:0% of 64 bytes how do you do this well you just 00:04:39.170 --> 00:04:39.180 align:start position:0% just 00:04:39.180 --> 00:04:41.689 align:start position:0% just break<00:04:39.780> your<00:04:40.320> message<00:04:40.440> into<00:04:40.740> multiple<00:04:41.160> of<00:04:41.340> 64 00:04:41.689 --> 00:04:41.699 align:start position:0% break your message into multiple of 64 00:04:41.699 --> 00:04:44.810 align:start position:0% break your message into multiple of 64 bytes<00:04:42.120> you<00:04:43.020> add<00:04:43.380> one<00:04:44.040> bit 00:04:44.810 --> 00:04:44.820 align:start position:0% bytes you add one bit 00:04:44.820 --> 00:04:46.610 align:start position:0% bytes you add one bit at<00:04:45.240> the<00:04:45.360> very<00:04:45.419> end<00:04:45.660> of<00:04:45.780> the<00:04:45.900> message<00:04:46.020> just<00:04:46.320> to 00:04:46.610 --> 00:04:46.620 align:start position:0% at the very end of the message just to 00:04:46.620 --> 00:04:49.490 align:start position:0% at the very end of the message just to Signal<00:04:46.979> this<00:04:47.759> message<00:04:48.000> has<00:04:48.419> ended<00:04:48.840> so<00:04:49.320> here 00:04:49.490 --> 00:04:49.500 align:start position:0% Signal this message has ended so here 00:04:49.500 --> 00:04:52.730 align:start position:0% Signal this message has ended so here the<00:04:49.800> message<00:04:49.979> is<00:04:50.280> less<00:04:50.520> than<00:04:50.699> 64<00:04:51.180> bytes<00:04:51.600> is<00:04:52.139> 24 00:04:52.730 --> 00:04:52.740 align:start position:0% the message is less than 64 bytes is 24 00:04:52.740 --> 00:04:55.550 align:start position:0% the message is less than 64 bytes is 24 bits<00:04:53.340> these<00:04:54.060> three<00:04:54.180> bytes<00:04:54.600> there<00:04:54.780> a<00:04:55.020> b<00:04:55.199> and<00:04:55.320> c 00:04:55.550 --> 00:04:55.560 align:start position:0% bits these three bytes there a b and c 00:04:55.560 --> 00:04:59.450 align:start position:0% bits these three bytes there a b and c this<00:04:56.400> one<00:04:56.639> there<00:04:57.000> is<00:04:57.660> showing<00:04:58.020> that<00:04:58.320> we<00:04:58.979> added 00:04:59.450 --> 00:04:59.460 align:start position:0% this one there is showing that we added 00:04:59.460 --> 00:05:01.189 align:start position:0% this one there is showing that we added that<00:04:59.759> extra<00:05:00.000> bit<00:05:00.300> to<00:05:00.479> show<00:05:00.600> the<00:05:00.720> message<00:05:00.900> has 00:05:01.189 --> 00:05:01.199 align:start position:0% that extra bit to show the message has 00:05:01.199 --> 00:05:04.969 align:start position:0% that extra bit to show the message has finished<00:05:01.440> and<00:05:02.400> then<00:05:02.520> you<00:05:02.759> pad<00:05:03.000> with<00:05:03.300> zero<00:05:03.720> bits 00:05:04.969 --> 00:05:04.979 align:start position:0% finished and then you pad with zero bits 00:05:04.979 --> 00:05:08.930 align:start position:0% finished and then you pad with zero bits up<00:05:05.759> to<00:05:06.000> a<00:05:06.240> multiple<00:05:06.600> of<00:05:06.720> 64<00:05:07.199> bytes<00:05:07.620> minus<00:05:08.040> eight 00:05:08.930 --> 00:05:08.940 align:start position:0% up to a multiple of 64 bytes minus eight 00:05:08.940 --> 00:05:10.310 align:start position:0% up to a multiple of 64 bytes minus eight because<00:05:09.300> you're<00:05:09.540> going<00:05:09.720> to<00:05:09.780> use<00:05:09.900> the<00:05:10.199> last 00:05:10.310 --> 00:05:10.320 align:start position:0% because you're going to use the last 00:05:10.320 --> 00:05:13.430 align:start position:0% because you're going to use the last eight<00:05:10.620> bytes<00:05:11.040> or<00:05:11.340> 64<00:05:11.759> bits<00:05:12.180> to<00:05:12.660> encode<00:05:13.080> the 00:05:13.430 --> 00:05:13.440 align:start position:0% eight bytes or 64 bits to encode the 00:05:13.440 --> 00:05:15.590 align:start position:0% eight bytes or 64 bits to encode the length<00:05:13.680> of<00:05:13.860> the<00:05:14.100> whole<00:05:14.220> message 00:05:15.590 --> 00:05:15.600 align:start position:0% length of the whole message 00:05:15.600 --> 00:05:17.749 align:start position:0% length of the whole message as<00:05:16.080> is<00:05:16.259> there<00:05:16.380> in<00:05:16.680> binary<00:05:16.979> that's<00:05:17.280> the<00:05:17.580> number 00:05:17.749 --> 00:05:17.759 align:start position:0% as is there in binary that's the number 00:05:17.759 --> 00:05:19.550 align:start position:0% as is there in binary that's the number 24<00:05:18.000> which<00:05:18.419> is<00:05:18.600> the<00:05:18.720> actual<00:05:18.960> length<00:05:19.259> of<00:05:19.380> this 00:05:19.550 --> 00:05:19.560 align:start position:0% 24 which is the actual length of this 00:05:19.560 --> 00:05:20.330 align:start position:0% 24 which is the actual length of this message 00:05:20.330 --> 00:05:20.340 align:start position:0% message 00:05:20.340 --> 00:05:23.150 align:start position:0% message okay<00:05:20.520> so<00:05:21.000> this<00:05:21.120> is<00:05:21.240> the<00:05:21.360> procedure<00:05:21.720> to<00:05:21.840> pad<00:05:22.160> uh 00:05:23.150 --> 00:05:23.160 align:start position:0% okay so this is the procedure to pad uh 00:05:23.160 --> 00:05:25.430 align:start position:0% okay so this is the procedure to pad uh your<00:05:23.580> message<00:05:24.060> which<00:05:24.419> was<00:05:24.600> arbitrary<00:05:25.080> length 00:05:25.430 --> 00:05:25.440 align:start position:0% your message which was arbitrary length 00:05:25.440 --> 00:05:28.249 align:start position:0% your message which was arbitrary length into<00:05:25.979> a<00:05:26.340> multiple<00:05:26.699> of<00:05:26.820> 64<00:05:27.180> bytes 00:05:28.249 --> 00:05:28.259 align:start position:0% into a multiple of 64 bytes 00:05:28.259 --> 00:05:30.650 align:start position:0% into a multiple of 64 bytes vectorization<00:05:28.979> which<00:05:29.340> is<00:05:29.460> the<00:05:29.639> main<00:05:30.000> topic<00:05:30.360> of 00:05:30.650 --> 00:05:30.660 align:start position:0% vectorization which is the main topic of 00:05:30.660 --> 00:05:32.570 align:start position:0% vectorization which is the main topic of this<00:05:30.780> talk<00:05:31.020> is<00:05:31.259> vectorization 00:05:32.570 --> 00:05:32.580 align:start position:0% this talk is vectorization 00:05:32.580 --> 00:05:34.430 align:start position:0% this talk is vectorization so<00:05:32.940> the<00:05:33.180> first<00:05:33.300> thing<00:05:33.419> to<00:05:33.600> notice<00:05:33.900> is<00:05:34.080> that 00:05:34.430 --> 00:05:34.440 align:start position:0% so the first thing to notice is that 00:05:34.440 --> 00:05:36.710 align:start position:0% so the first thing to notice is that you're<00:05:35.039> not<00:05:35.280> going<00:05:35.580> to<00:05:35.759> beat<00:05:36.060> any<00:05:36.479> of<00:05:36.660> the 00:05:36.710 --> 00:05:36.720 align:start position:0% you're not going to beat any of the 00:05:36.720 --> 00:05:39.230 align:start position:0% you're not going to beat any of the hazards<00:05:37.259> out<00:05:37.380> there<00:05:37.560> every<00:05:38.520> implementation 00:05:39.230 --> 00:05:39.240 align:start position:0% hazards out there every implementation 00:05:39.240 --> 00:05:40.850 align:start position:0% hazards out there every implementation that<00:05:39.539> I've<00:05:39.660> reviewed<00:05:40.020> before<00:05:40.259> according<00:05:40.560> to 00:05:40.850 --> 00:05:40.860 align:start position:0% that I've reviewed before according to 00:05:40.860 --> 00:05:42.830 align:start position:0% that I've reviewed before according to this<00:05:41.039> and<00:05:41.639> I'm<00:05:41.820> not<00:05:41.940> an<00:05:42.060> expert<00:05:42.419> at<00:05:42.539> role<00:05:42.720> on 00:05:42.830 --> 00:05:42.840 align:start position:0% this and I'm not an expert at role on 00:05:42.840 --> 00:05:44.570 align:start position:0% this and I'm not an expert at role on this<00:05:42.960> I<00:05:43.199> I<00:05:43.440> come<00:05:43.919> from<00:05:44.039> a<00:05:44.220> completely 00:05:44.570 --> 00:05:44.580 align:start position:0% this I I come from a completely 00:05:44.580 --> 00:05:45.770 align:start position:0% this I I come from a completely different<00:05:44.699> subjects<00:05:45.300> but<00:05:45.600> every 00:05:45.770 --> 00:05:45.780 align:start position:0% different subjects but every 00:05:45.780 --> 00:05:47.450 align:start position:0% different subjects but every implementation<00:05:46.259> I've<00:05:46.500> seen<00:05:46.680> is<00:05:46.979> equivalent 00:05:47.450 --> 00:05:47.460 align:start position:0% implementation I've seen is equivalent 00:05:47.460 --> 00:05:50.390 align:start position:0% implementation I've seen is equivalent to<00:05:48.000> Intel's<00:05:48.660> original<00:05:48.840> paper<00:05:49.560> white<00:05:49.919> paper<00:05:50.160> on 00:05:50.390 --> 00:05:50.400 align:start position:0% to Intel's original paper white paper on 00:05:50.400 --> 00:05:52.790 align:start position:0% to Intel's original paper white paper on this<00:05:50.940> and<00:05:51.600> it's<00:05:51.780> equivalent<00:05:52.080> to<00:05:52.199> what<00:05:52.380> opens 00:05:52.790 --> 00:05:52.800 align:start position:0% this and it's equivalent to what opens 00:05:52.800 --> 00:05:55.370 align:start position:0% this and it's equivalent to what opens the<00:05:52.919> cell<00:05:53.100> for<00:05:53.280> example<00:05:53.460> does<00:05:54.180> all<00:05:55.080> of<00:05:55.259> them 00:05:55.370 --> 00:05:55.380 align:start position:0% the cell for example does all of them 00:05:55.380 --> 00:05:57.890 align:start position:0% the cell for example does all of them implement<00:05:55.860> the<00:05:56.100> following<00:05:56.460> you<00:05:57.240> can<00:05:57.360> compute 00:05:57.890 --> 00:05:57.900 align:start position:0% implement the following you can compute 00:05:57.900 --> 00:05:59.629 align:start position:0% implement the following you can compute your<00:05:58.380> scheduled<00:05:58.740> words<00:05:58.919> as<00:05:59.160> I<00:05:59.340> told<00:05:59.520> you 00:05:59.629 --> 00:05:59.639 align:start position:0% your scheduled words as I told you 00:05:59.639 --> 00:06:01.610 align:start position:0% your scheduled words as I told you without<00:06:00.419> knowing<00:06:00.840> what<00:06:00.960> the<00:06:01.139> previously 00:06:01.610 --> 00:06:01.620 align:start position:0% without knowing what the previously 00:06:01.620 --> 00:06:04.490 align:start position:0% without knowing what the previously hashed<00:06:02.460> uh<00:06:03.419> what<00:06:03.720> the<00:06:03.900> previous<00:06:04.020> what<00:06:04.380> the 00:06:04.490 --> 00:06:04.500 align:start position:0% hashed uh what the previous what the 00:06:04.500 --> 00:06:06.469 align:start position:0% hashed uh what the previous what the current<00:06:04.740> status<00:06:05.039> of<00:06:05.280> the<00:06:05.400> hashing<00:06:05.759> is 00:06:06.469 --> 00:06:06.479 align:start position:0% current status of the hashing is 00:06:06.479 --> 00:06:08.810 align:start position:0% current status of the hashing is so<00:06:06.900> all<00:06:07.080> of<00:06:07.259> them<00:06:07.440> use<00:06:07.860> Vector<00:06:08.340> instructions 00:06:08.810 --> 00:06:08.820 align:start position:0% so all of them use Vector instructions 00:06:08.820 --> 00:06:12.050 align:start position:0% so all of them use Vector instructions to<00:06:09.780> compute<00:06:10.139> several<00:06:10.560> words<00:06:10.800> at<00:06:11.039> a<00:06:11.160> time<00:06:11.280> if 00:06:12.050 --> 00:06:12.060 align:start position:0% to compute several words at a time if 00:06:12.060 --> 00:06:16.850 align:start position:0% to compute several words at a time if your<00:06:12.180> CPU<00:06:12.600> supports<00:06:13.340> uh<00:06:14.600> 128<00:06:15.600> bits<00:06:16.139> registers 00:06:16.850 --> 00:06:16.860 align:start position:0% your CPU supports uh 128 bits registers 00:06:16.860 --> 00:06:19.430 align:start position:0% your CPU supports uh 128 bits registers like<00:06:17.100> these<00:06:17.340> ones<00:06:17.639> and<00:06:18.180> almost<00:06:18.360> every<00:06:18.720> CPU<00:06:19.199> now 00:06:19.430 --> 00:06:19.440 align:start position:0% like these ones and almost every CPU now 00:06:19.440 --> 00:06:22.010 align:start position:0% like these ones and almost every CPU now does<00:06:20.160> then<00:06:20.639> you're<00:06:20.820> gonna<00:06:21.000> compute<00:06:21.660> four 00:06:22.010 --> 00:06:22.020 align:start position:0% does then you're gonna compute four 00:06:22.020 --> 00:06:23.749 align:start position:0% does then you're gonna compute four words<00:06:22.259> at<00:06:22.500> a<00:06:22.620> time<00:06:22.740> four<00:06:22.919> double<00:06:23.160> words<00:06:23.400> at<00:06:23.639> a 00:06:23.749 --> 00:06:23.759 align:start position:0% words at a time four double words at a 00:06:23.759 --> 00:06:26.390 align:start position:0% words at a time four double words at a time<00:06:24.300> you<00:06:24.900> start<00:06:25.080> with<00:06:25.319> your<00:06:25.560> 16<00:06:25.800> double<00:06:26.160> Words 00:06:26.390 --> 00:06:26.400 align:start position:0% time you start with your 16 double Words 00:06:26.400 --> 00:06:28.790 align:start position:0% time you start with your 16 double Words which<00:06:26.699> is<00:06:26.819> your<00:06:27.060> message<00:06:27.240> and<00:06:27.840> you<00:06:28.080> can<00:06:28.199> put 00:06:28.790 --> 00:06:28.800 align:start position:0% which is your message and you can put 00:06:28.800 --> 00:06:31.129 align:start position:0% which is your message and you can put those<00:06:29.280> 16<00:06:29.639> double<00:06:30.000> words<00:06:30.240> in<00:06:30.660> only<00:06:30.900> four 00:06:31.129 --> 00:06:31.139 align:start position:0% those 16 double words in only four 00:06:31.139 --> 00:06:34.670 align:start position:0% those 16 double words in only four registers<00:06:31.680> if<00:06:32.340> your<00:06:32.520> computer<00:06:32.699> supports<00:06:33.780> uh 00:06:34.670 --> 00:06:34.680 align:start position:0% registers if your computer supports uh 00:06:34.680 --> 00:06:36.230 align:start position:0% registers if your computer supports uh avx2 00:06:36.230 --> 00:06:36.240 align:start position:0% avx2 00:06:36.240 --> 00:06:38.330 align:start position:0% avx2 that<00:06:36.900> you<00:06:37.380> can<00:06:37.560> do<00:06:37.680> our<00:06:37.860> genetic<00:06:38.160> and 00:06:38.330 --> 00:06:38.340 align:start position:0% that you can do our genetic and 00:06:38.340 --> 00:06:40.670 align:start position:0% that you can do our genetic and registers<00:06:38.699> that<00:06:38.880> are<00:06:39.060> 256<00:06:39.780> bits<00:06:40.259> then<00:06:40.560> you're 00:06:40.670 --> 00:06:40.680 align:start position:0% registers that are 256 bits then you're 00:06:40.680 --> 00:06:42.290 align:start position:0% registers that are 256 bits then you're gonna<00:06:40.860> compute<00:06:41.220> eight<00:06:41.520> you're<00:06:42.000> gonna<00:06:42.120> use 00:06:42.290 --> 00:06:42.300 align:start position:0% gonna compute eight you're gonna use 00:06:42.300 --> 00:06:43.610 align:start position:0% gonna compute eight you're gonna use only<00:06:42.479> two<00:06:42.720> registers<00:06:43.199> and<00:06:43.319> you're<00:06:43.440> gonna 00:06:43.610 --> 00:06:43.620 align:start position:0% only two registers and you're gonna 00:06:43.620 --> 00:06:45.590 align:start position:0% only two registers and you're gonna compute<00:06:43.979> eight<00:06:44.160> words<00:06:44.400> at<00:06:44.639> a<00:06:44.819> time<00:06:44.940> if<00:06:45.419> your 00:06:45.590 --> 00:06:45.600 align:start position:0% compute eight words at a time if your 00:06:45.600 --> 00:06:48.409 align:start position:0% compute eight words at a time if your computer<00:06:45.780> supports<00:06:46.620> AVX<00:06:47.039> 512<00:06:47.880> you're<00:06:48.240> gonna 00:06:48.409 --> 00:06:48.419 align:start position:0% computer supports AVX 512 you're gonna 00:06:48.419 --> 00:06:51.290 align:start position:0% computer supports AVX 512 you're gonna compute<00:06:48.900> all<00:06:49.680> of<00:06:49.800> them<00:06:49.919> the<00:06:50.220> 16<00:06:50.520> of<00:06:50.819> them<00:06:50.940> at<00:06:51.180> a 00:06:51.290 --> 00:06:51.300 align:start position:0% compute all of them the 16 of them at a 00:06:51.300 --> 00:06:53.529 align:start position:0% compute all of them the 16 of them at a time 00:06:53.529 --> 00:06:53.539 align:start position:0% 00:06:53.539 --> 00:06:56.029 align:start position:0% 1024<00:06:54.539> does<00:06:54.720> not<00:06:54.900> exist<00:06:55.259> yet<00:06:55.440> but<00:06:55.740> it's<00:06:55.860> already 00:06:56.029 --> 00:06:56.039 align:start position:0% 1024 does not exist yet but it's already 00:06:56.039 --> 00:06:58.490 align:start position:0% 1024 does not exist yet but it's already in<00:06:56.280> the<00:06:56.400> books<00:06:56.940> there's<00:06:57.600> two<00:06:57.840> options<00:06:58.139> into 00:06:58.490 --> 00:06:58.500 align:start position:0% in the books there's two options into 00:06:58.500 --> 00:07:00.590 align:start position:0% in the books there's two options into Computing<00:06:59.039> This<00:06:59.280> Modern<00:06:59.759> implementations 00:07:00.590 --> 00:07:00.600 align:start position:0% Computing This Modern implementations 00:07:00.600 --> 00:07:04.070 align:start position:0% Computing This Modern implementations either<00:07:01.080> do<00:07:01.620> this<00:07:02.220> this<00:07:02.580> vectorization<00:07:03.180> or<00:07:03.900> if 00:07:04.070 --> 00:07:04.080 align:start position:0% either do this this vectorization or if 00:07:04.080 --> 00:07:06.950 align:start position:0% either do this this vectorization or if your<00:07:04.259> CPU<00:07:05.120> implements<00:07:06.120> cryptographic 00:07:06.950 --> 00:07:06.960 align:start position:0% your CPU implements cryptographic 00:07:06.960 --> 00:07:09.650 align:start position:0% your CPU implements cryptographic extensions<00:07:07.440> like<00:07:08.280> what's<00:07:09.240> there<00:07:09.419> in<00:07:09.600> a 00:07:09.650 --> 00:07:09.660 align:start position:0% extensions like what's there in a 00:07:09.660 --> 00:07:12.890 align:start position:0% extensions like what's there in a picture<00:07:09.780> then<00:07:10.380> they<00:07:10.620> will<00:07:10.740> use<00:07:11.039> 128<00:07:11.940> bits 00:07:12.890 --> 00:07:12.900 align:start position:0% picture then they will use 128 bits 00:07:12.900 --> 00:07:14.689 align:start position:0% picture then they will use 128 bits registers<00:07:13.620> regardless<00:07:14.160> if<00:07:14.400> your<00:07:14.520> computer 00:07:14.689 --> 00:07:14.699 align:start position:0% registers regardless if your computer 00:07:14.699 --> 00:07:16.670 align:start position:0% registers regardless if your computer has<00:07:15.060> larger<00:07:15.419> registers<00:07:15.960> because 00:07:16.670 --> 00:07:16.680 align:start position:0% has larger registers because 00:07:16.680 --> 00:07:19.010 align:start position:0% has larger registers because cryptographic<00:07:17.639> extensions<00:07:18.120> can<00:07:18.600> compute 00:07:19.010 --> 00:07:19.020 align:start position:0% cryptographic extensions can compute 00:07:19.020 --> 00:07:20.930 align:start position:0% cryptographic extensions can compute forwards<00:07:19.560> at<00:07:19.680> a<00:07:19.800> time<00:07:19.919> much<00:07:20.160> faster<00:07:20.520> than 00:07:20.930 --> 00:07:20.940 align:start position:0% forwards at a time much faster than 00:07:20.940 --> 00:07:23.870 align:start position:0% forwards at a time much faster than vectorized<00:07:21.979> computations<00:07:22.979> but<00:07:23.520> the<00:07:23.639> point<00:07:23.759> I 00:07:23.870 --> 00:07:23.880 align:start position:0% vectorized computations but the point I 00:07:23.880 --> 00:07:25.189 align:start position:0% vectorized computations but the point I want<00:07:24.000> to<00:07:24.120> make<00:07:24.180> is<00:07:24.360> that<00:07:24.539> this<00:07:24.660> vectorization 00:07:25.189 --> 00:07:25.199 align:start position:0% want to make is that this vectorization 00:07:25.199 --> 00:07:27.409 align:start position:0% want to make is that this vectorization is<00:07:25.440> there<00:07:25.680> in<00:07:26.220> every<00:07:26.400> current<00:07:26.699> implementation 00:07:27.409 --> 00:07:27.419 align:start position:0% is there in every current implementation 00:07:27.419 --> 00:07:30.770 align:start position:0% is there in every current implementation this<00:07:27.780> is<00:07:27.900> since<00:07:28.080> Intel's<00:07:28.880> uh<00:07:29.880> since<00:07:30.419> that's 00:07:30.770 --> 00:07:30.780 align:start position:0% this is since Intel's uh since that's 00:07:30.780 --> 00:07:32.689 align:start position:0% this is since Intel's uh since that's white<00:07:31.020> paper<00:07:31.259> also<00:07:32.039> Computing<00:07:32.400> the<00:07:32.580> words 00:07:32.689 --> 00:07:32.699 align:start position:0% white paper also Computing the words 00:07:32.699 --> 00:07:35.510 align:start position:0% white paper also Computing the words like<00:07:32.940> this<00:07:33.120> is<00:07:33.539> useful<00:07:33.900> because<00:07:34.560> since<00:07:35.280> you 00:07:35.510 --> 00:07:35.520 align:start position:0% like this is useful because since you 00:07:35.520 --> 00:07:37.010 align:start position:0% like this is useful because since you can<00:07:35.639> compute<00:07:35.940> the<00:07:36.060> words<00:07:36.240> and<00:07:36.539> at<00:07:36.720> the<00:07:36.840> same 00:07:37.010 --> 00:07:37.020 align:start position:0% can compute the words and at the same 00:07:37.020 --> 00:07:37.790 align:start position:0% can compute the words and at the same time 00:07:37.790 --> 00:07:37.800 align:start position:0% time 00:07:37.800 --> 00:07:40.129 align:start position:0% time uh<00:07:38.639> pass<00:07:38.940> through<00:07:39.180> some<00:07:39.360> rounds<00:07:39.720> so<00:07:39.840> let's<00:07:40.020> say 00:07:40.129 --> 00:07:40.139 align:start position:0% uh pass through some rounds so let's say 00:07:40.139 --> 00:07:42.050 align:start position:0% uh pass through some rounds so let's say that<00:07:40.319> you<00:07:40.500> computed<00:07:40.919> the<00:07:41.099> fifth<00:07:41.400> word<00:07:41.580> then 00:07:42.050 --> 00:07:42.060 align:start position:0% that you computed the fifth word then 00:07:42.060 --> 00:07:44.990 align:start position:0% that you computed the fifth word then you<00:07:42.300> can<00:07:42.479> pass<00:07:42.840> up<00:07:43.080> to<00:07:43.259> the<00:07:43.380> fifth<00:07:43.680> round<00:07:44.000> then 00:07:44.990 --> 00:07:45.000 align:start position:0% you can pass up to the fifth round then 00:07:45.000 --> 00:07:47.270 align:start position:0% you can pass up to the fifth round then you<00:07:45.419> can<00:07:45.539> mix<00:07:45.840> scalar<00:07:46.620> operations<00:07:46.919> with 00:07:47.270 --> 00:07:47.280 align:start position:0% you can mix scalar operations with 00:07:47.280 --> 00:07:48.770 align:start position:0% you can mix scalar operations with Vector<00:07:47.580> operations<00:07:47.940> and<00:07:48.180> the<00:07:48.300> CPU<00:07:48.599> would 00:07:48.770 --> 00:07:48.780 align:start position:0% Vector operations and the CPU would 00:07:48.780 --> 00:07:51.650 align:start position:0% Vector operations and the CPU would perform<00:07:48.960> this<00:07:49.259> in<00:07:49.440> parallel<00:07:49.860> this<00:07:50.400> views<00:07:50.940> can 00:07:51.650 --> 00:07:51.660 align:start position:0% perform this in parallel this views can 00:07:51.660 --> 00:07:54.409 align:start position:0% perform this in parallel this views can take<00:07:51.900> has<00:07:52.680> several<00:07:52.979> ports<00:07:53.460> and<00:07:53.940> can<00:07:54.180> take 00:07:54.409 --> 00:07:54.419 align:start position:0% take has several ports and can take 00:07:54.419 --> 00:07:56.330 align:start position:0% take has several ports and can take different<00:07:54.960> types<00:07:55.440> of<00:07:55.500> operations<00:07:55.860> at<00:07:56.160> the 00:07:56.330 --> 00:07:56.340 align:start position:0% different types of operations at the 00:07:56.340 --> 00:07:58.309 align:start position:0% different types of operations at the same<00:07:56.460> time<00:07:56.580> so<00:07:57.180> you<00:07:57.360> can't<00:07:57.479> be<00:07:57.660> Computing<00:07:58.080> the 00:07:58.309 --> 00:07:58.319 align:start position:0% same time so you can't be Computing the 00:07:58.319 --> 00:08:00.770 align:start position:0% same time so you can't be Computing the fifth<00:07:58.740> round<00:07:59.220> that<00:07:59.819> is<00:08:00.000> a<00:08:00.180> computation<00:08:00.539> that 00:08:00.770 --> 00:08:00.780 align:start position:0% fifth round that is a computation that 00:08:00.780 --> 00:08:03.170 align:start position:0% fifth round that is a computation that it<00:08:00.960> has<00:08:01.199> to<00:08:01.319> be<00:08:01.500> done<00:08:01.740> on<00:08:02.280> the<00:08:02.460> scalar<00:08:02.819> part<00:08:03.000> of 00:08:03.170 --> 00:08:03.180 align:start position:0% it has to be done on the scalar part of 00:08:03.180 --> 00:08:05.570 align:start position:0% it has to be done on the scalar part of the<00:08:03.300> CPU<00:08:03.660> because<00:08:03.900> it<00:08:04.080> has<00:08:04.259> to<00:08:04.380> be<00:08:04.560> uh<00:08:05.340> it 00:08:05.570 --> 00:08:05.580 align:start position:0% the CPU because it has to be uh it 00:08:05.580 --> 00:08:07.249 align:start position:0% the CPU because it has to be uh it cannot<00:08:05.759> be<00:08:05.940> in<00:08:06.120> parallel 00:08:07.249 --> 00:08:07.259 align:start position:0% cannot be in parallel 00:08:07.259 --> 00:08:09.290 align:start position:0% cannot be in parallel and<00:08:07.800> at<00:08:08.039> the<00:08:08.160> same<00:08:08.220> time<00:08:08.400> you're<00:08:08.759> Computing 00:08:09.290 --> 00:08:09.300 align:start position:0% and at the same time you're Computing 00:08:09.300 --> 00:08:12.110 align:start position:0% and at the same time you're Computing the<00:08:09.599> sixth<00:08:10.259> word<00:08:10.440> in<00:08:11.039> parallel 00:08:12.110 --> 00:08:12.120 align:start position:0% the sixth word in parallel 00:08:12.120 --> 00:08:14.390 align:start position:0% the sixth word in parallel on<00:08:12.720> on<00:08:12.900> Vector<00:08:13.259> registers 00:08:14.390 --> 00:08:14.400 align:start position:0% on on Vector registers 00:08:14.400 --> 00:08:17.990 align:start position:0% on on Vector registers okay<00:08:14.880> so<00:08:15.300> I've<00:08:15.539> covered<00:08:16.259> what<00:08:16.919> is<00:08:17.099> what<00:08:17.639> is<00:08:17.819> a 00:08:17.990 --> 00:08:18.000 align:start position:0% okay so I've covered what is what is a 00:08:18.000 --> 00:08:21.110 align:start position:0% okay so I've covered what is what is a typical<00:08:18.240> uh<00:08:19.080> implementation<00:08:19.800> of<00:08:20.039> hashing<00:08:20.400> and 00:08:21.110 --> 00:08:21.120 align:start position:0% typical uh implementation of hashing and 00:08:21.120 --> 00:08:23.390 align:start position:0% typical uh implementation of hashing and this<00:08:21.720> is<00:08:21.900> the<00:08:22.080> the<00:08:22.319> thing<00:08:22.560> to<00:08:22.860> remember<00:08:22.979> from 00:08:23.390 --> 00:08:23.400 align:start position:0% this is the the thing to remember from 00:08:23.400 --> 00:08:25.730 align:start position:0% this is the the thing to remember from the<00:08:23.580> whole<00:08:23.759> part<00:08:24.000> from<00:08:24.660> the<00:08:24.780> whole<00:08:25.139> purpose<00:08:25.500> of 00:08:25.730 --> 00:08:25.740 align:start position:0% the whole part from the whole purpose of 00:08:25.740 --> 00:08:27.650 align:start position:0% the whole part from the whole purpose of the<00:08:25.860> dot<00:08:26.099> is<00:08:26.460> that<00:08:26.580> the<00:08:26.699> hasher<00:08:27.120> signature<00:08:27.539> is 00:08:27.650 --> 00:08:27.660 align:start position:0% the dot is that the hasher signature is 00:08:27.660 --> 00:08:29.450 align:start position:0% the dot is that the hasher signature is this<00:08:27.900> in<00:08:28.500> either<00:08:28.680> language<00:08:29.039> is<00:08:29.280> something 00:08:29.450 --> 00:08:29.460 align:start position:0% this in either language is something 00:08:29.460 --> 00:08:31.129 align:start position:0% this in either language is something like<00:08:29.699> this<00:08:29.879> it's<00:08:30.240> something<00:08:30.419> that<00:08:30.720> takes<00:08:30.960> an 00:08:31.129 --> 00:08:31.139 align:start position:0% like this it's something that takes an 00:08:31.139 --> 00:08:34.250 align:start position:0% like this it's something that takes an arbitrary<00:08:31.560> length<00:08:32.399> bite<00:08:32.940> slice<00:08:33.360> and<00:08:34.020> it<00:08:34.080> gives 00:08:34.250 --> 00:08:34.260 align:start position:0% arbitrary length bite slice and it gives 00:08:34.260 --> 00:08:37.190 align:start position:0% arbitrary length bite slice and it gives you<00:08:34.440> back<00:08:34.620> 32<00:08:35.039> bytes<00:08:35.459> which<00:08:35.760> is<00:08:35.880> the<00:08:36.120> hash 00:08:37.190 --> 00:08:37.200 align:start position:0% you back 32 bytes which is the hash 00:08:37.200 --> 00:08:39.170 align:start position:0% you back 32 bytes which is the hash so<00:08:37.620> it<00:08:37.800> takes<00:08:37.919> an<00:08:38.219> arbitrary<00:08:38.640> linked<00:08:39.000> message 00:08:39.170 --> 00:08:39.180 align:start position:0% so it takes an arbitrary linked message 00:08:39.180 --> 00:08:40.850 align:start position:0% so it takes an arbitrary linked message and<00:08:39.539> it<00:08:39.659> gives<00:08:39.779> you<00:08:39.959> one<00:08:40.140> it<00:08:40.380> gives<00:08:40.560> you<00:08:40.740> a 00:08:40.850 --> 00:08:40.860 align:start position:0% and it gives you one it gives you a 00:08:40.860 --> 00:08:43.490 align:start position:0% and it gives you one it gives you a digest<00:08:41.279> and<00:08:42.060> this<00:08:42.240> is<00:08:42.300> something<00:08:42.539> that<00:08:43.320> you're 00:08:43.490 --> 00:08:43.500 align:start position:0% digest and this is something that you're 00:08:43.500 --> 00:08:44.930 align:start position:0% digest and this is something that you're not<00:08:43.680> gonna<00:08:43.860> beat<00:08:44.039> you're<00:08:44.459> not<00:08:44.640> going<00:08:44.760> to 00:08:44.930 --> 00:08:44.940 align:start position:0% not gonna beat you're not going to 00:08:44.940 --> 00:08:47.269 align:start position:0% not gonna beat you're not going to implement<00:08:45.240> this<00:08:45.540> better<00:08:45.720> than<00:08:46.020> open<00:08:46.200> SSL<00:08:46.800> no 00:08:47.269 --> 00:08:47.279 align:start position:0% implement this better than open SSL no 00:08:47.279 --> 00:08:48.889 align:start position:0% implement this better than open SSL no one<00:08:47.459> is<00:08:47.580> going<00:08:47.760> to<00:08:47.880> do<00:08:47.940> you<00:08:48.180> perhaps<00:08:48.480> can<00:08:48.600> do<00:08:48.779> it 00:08:48.889 --> 00:08:48.899 align:start position:0% one is going to do you perhaps can do it 00:08:48.899 --> 00:08:50.630 align:start position:0% one is going to do you perhaps can do it for<00:08:49.080> one<00:08:49.320> CPU<00:08:49.740> you're<00:08:50.040> not<00:08:50.220> going<00:08:50.339> to<00:08:50.399> write<00:08:50.519> an 00:08:50.630 --> 00:08:50.640 align:start position:0% for one CPU you're not going to write an 00:08:50.640 --> 00:08:52.250 align:start position:0% for one CPU you're not going to write an implementation<00:08:51.060> faster<00:08:51.540> than<00:08:51.899> what<00:08:52.080> is 00:08:52.250 --> 00:08:52.260 align:start position:0% implementation faster than what is 00:08:52.260 --> 00:08:53.570 align:start position:0% implementation faster than what is already<00:08:52.500> there 00:08:53.570 --> 00:08:53.580 align:start position:0% already there 00:08:53.580 --> 00:08:57.290 align:start position:0% already there however<00:08:54.440> we<00:08:55.440> use<00:08:55.800> hashing<00:08:56.519> in<00:08:57.000> a<00:08:57.120> very 00:08:57.290 --> 00:08:57.300 align:start position:0% however we use hashing in a very 00:08:57.300 --> 00:09:00.350 align:start position:0% however we use hashing in a very restricted<00:08:57.899> scenario<00:08:58.320> we<00:08:58.920> use<00:08:59.100> hashing<00:09:00.000> to 00:09:00.350 --> 00:09:00.360 align:start position:0% restricted scenario we use hashing to 00:09:00.360 --> 00:09:03.170 align:start position:0% restricted scenario we use hashing to Hash<00:09:00.660> Merkle<00:09:01.140> traits<00:09:01.860> a<00:09:02.279> miracle<00:09:02.700> trees<00:09:03.000> are 00:09:03.170 --> 00:09:03.180 align:start position:0% Hash Merkle traits a miracle trees are 00:09:03.180 --> 00:09:07.370 align:start position:0% Hash Merkle traits a miracle trees are not<00:09:03.360> arbitrary<00:09:03.959> length<00:09:04.680> Merkel<00:09:05.399> trees<00:09:05.839> are<00:09:06.839> in 00:09:07.370 --> 00:09:07.380 align:start position:0% not arbitrary length Merkel trees are in 00:09:07.380 --> 00:09:08.930 align:start position:0% not arbitrary length Merkel trees are in the<00:09:07.500> case<00:09:07.680> of<00:09:07.860> the<00:09:08.040> consensus<00:09:08.399> layer<00:09:08.760> for 00:09:08.930 --> 00:09:08.940 align:start position:0% the case of the consensus layer for 00:09:08.940 --> 00:09:10.850 align:start position:0% the case of the consensus layer for example<00:09:09.120> where<00:09:09.420> the<00:09:09.600> nodes<00:09:09.839> are<00:09:10.080> 32<00:09:10.440> bytes 00:09:10.850 --> 00:09:10.860 align:start position:0% example where the nodes are 32 bytes 00:09:10.860 --> 00:09:12.949 align:start position:0% example where the nodes are 32 bytes there's<00:09:11.279> something<00:09:11.459> like<00:09:11.700> this<00:09:11.880> each<00:09:12.660> one<00:09:12.839> of 00:09:12.949 --> 00:09:12.959 align:start position:0% there's something like this each one of 00:09:12.959 --> 00:09:15.949 align:start position:0% there's something like this each one of these<00:09:13.200> nodes<00:09:13.500> represent<00:09:14.100> 32<00:09:14.880> bytes<00:09:15.300> and<00:09:15.779> each 00:09:15.949 --> 00:09:15.959 align:start position:0% these nodes represent 32 bytes and each 00:09:15.959 --> 00:09:18.170 align:start position:0% these nodes represent 32 bytes and each parent<00:09:16.200> node<00:09:16.620> has<00:09:16.860> only<00:09:17.040> two<00:09:17.339> children<00:09:17.580> and 00:09:18.170 --> 00:09:18.180 align:start position:0% parent node has only two children and 00:09:18.180 --> 00:09:20.630 align:start position:0% parent node has only two children and these<00:09:18.480> two<00:09:18.540> children<00:09:18.800> are<00:09:19.800> hashed<00:09:20.279> together 00:09:20.630 --> 00:09:20.640 align:start position:0% these two children are hashed together 00:09:20.640 --> 00:09:24.490 align:start position:0% these two children are hashed together so<00:09:21.240> the<00:09:21.660> two<00:09:22.140> children<00:09:22.399> are<00:09:23.399> 64<00:09:23.820> bytes<00:09:24.180> in 00:09:24.490 --> 00:09:24.500 align:start position:0% so the two children are 64 bytes in 00:09:24.500 --> 00:09:26.990 align:start position:0% so the two children are 64 bytes in concatenated<00:09:25.500> you<00:09:25.980> hash<00:09:26.220> them<00:09:26.459> and<00:09:26.760> you<00:09:26.880> get 00:09:26.990 --> 00:09:27.000 align:start position:0% concatenated you hash them and you get 00:09:27.000 --> 00:09:29.630 align:start position:0% concatenated you hash them and you get the<00:09:27.120> hash<00:09:27.420> of<00:09:27.839> the<00:09:27.959> part<00:09:28.140> you<00:09:28.860> get<00:09:29.040> the<00:09:29.220> what 00:09:29.630 --> 00:09:29.640 align:start position:0% the hash of the part you get the what 00:09:29.640 --> 00:09:31.850 align:start position:0% the hash of the part you get the what what<00:09:29.820> goes<00:09:30.120> in<00:09:30.240> the<00:09:30.360> part<00:09:30.540> so<00:09:31.200> this<00:09:31.380> is<00:09:31.500> what<00:09:31.680> we 00:09:31.850 --> 00:09:31.860 align:start position:0% what goes in the part so this is what we 00:09:31.860 --> 00:09:36.170 align:start position:0% what goes in the part so this is what we we<00:09:32.100> hash<00:09:32.399> typically<00:09:33.680> and<00:09:34.680> we<00:09:35.519> observe<00:09:36.000> two 00:09:36.170 --> 00:09:36.180 align:start position:0% we hash typically and we observe two 00:09:36.180 --> 00:09:37.190 align:start position:0% we hash typically and we observe two things 00:09:37.190 --> 00:09:37.200 align:start position:0% things 00:09:37.200 --> 00:09:39.110 align:start position:0% things well<00:09:37.740> actually<00:09:37.980> the<00:09:38.279> execution<00:09:38.640> layer<00:09:39.000> has 00:09:39.110 --> 00:09:39.120 align:start position:0% well actually the execution layer has 00:09:39.120 --> 00:09:40.430 align:start position:0% well actually the execution layer has something<00:09:39.300> completely<00:09:39.720> different<00:09:39.899> but<00:09:40.260> still 00:09:40.430 --> 00:09:40.440 align:start position:0% something completely different but still 00:09:40.440 --> 00:09:41.690 align:start position:0% something completely different but still the<00:09:40.680> same<00:09:40.860> technique<00:09:41.100> is<00:09:41.220> going<00:09:41.399> to<00:09:41.459> apply 00:09:41.690 --> 00:09:41.700 align:start position:0% the same technique is going to apply 00:09:41.700 --> 00:09:44.150 align:start position:0% the same technique is going to apply what<00:09:42.420> we<00:09:42.660> observe<00:09:43.080> immediately<00:09:43.440> is<00:09:43.860> the 00:09:44.150 --> 00:09:44.160 align:start position:0% what we observe immediately is the 00:09:44.160 --> 00:09:45.769 align:start position:0% what we observe immediately is the following<00:09:44.580> thing<00:09:44.760> first<00:09:45.180> of<00:09:45.300> all<00:09:45.420> we're<00:09:45.660> not 00:09:45.769 --> 00:09:45.779 align:start position:0% following thing first of all we're not 00:09:45.779 --> 00:09:47.570 align:start position:0% following thing first of all we're not hashing<00:09:46.140> arbitrary<00:09:46.560> length<00:09:46.860> we're<00:09:47.100> hashing 00:09:47.570 --> 00:09:47.580 align:start position:0% hashing arbitrary length we're hashing 00:09:47.580 --> 00:09:50.810 align:start position:0% hashing arbitrary length we're hashing every<00:09:47.940> single<00:09:48.240> time<00:09:48.420> 64<00:09:49.019> bytes<00:09:49.980> and<00:09:50.459> second<00:09:50.580> of 00:09:50.810 --> 00:09:50.820 align:start position:0% every single time 64 bytes and second of 00:09:50.820 --> 00:09:53.150 align:start position:0% every single time 64 bytes and second of all<00:09:50.940> we<00:09:51.300> can<00:09:51.480> hash<00:09:51.660> this<00:09:51.899> in<00:09:52.080> parallel<00:09:52.380> you 00:09:53.150 --> 00:09:53.160 align:start position:0% all we can hash this in parallel you 00:09:53.160 --> 00:09:55.009 align:start position:0% all we can hash this in parallel you don't<00:09:53.399> need<00:09:53.519> to<00:09:53.700> you<00:09:54.060> can<00:09:54.180> hash<00:09:54.360> the<00:09:54.600> blue<00:09:54.779> one 00:09:55.009 --> 00:09:55.019 align:start position:0% don't need to you can hash the blue one 00:09:55.019 --> 00:09:58.130 align:start position:0% don't need to you can hash the blue one by<00:09:55.620> knowing<00:09:55.980> the<00:09:56.279> entire<00:09:56.519> left<00:09:57.060> subtree<00:09:57.540> and 00:09:58.130 --> 00:09:58.140 align:start position:0% by knowing the entire left subtree and 00:09:58.140 --> 00:09:59.690 align:start position:0% by knowing the entire left subtree and hash<00:09:58.440> the<00:09:58.680> yellow<00:09:58.860> one<00:09:59.100> completely<00:09:59.519> in 00:09:59.690 --> 00:09:59.700 align:start position:0% hash the yellow one completely in 00:09:59.700 --> 00:10:02.750 align:start position:0% hash the yellow one completely in parallel<00:10:00.000> in<00:10:00.300> the<00:10:00.420> other<00:10:00.540> side<00:10:01.279> of<00:10:02.279> course<00:10:02.399> you 00:10:02.750 --> 00:10:02.760 align:start position:0% parallel in the other side of course you 00:10:02.760 --> 00:10:04.190 align:start position:0% parallel in the other side of course you can<00:10:02.880> do<00:10:03.000> this<00:10:03.120> in<00:10:03.240> parallel<00:10:03.600> if<00:10:03.959> you<00:10:04.019> have<00:10:04.140> a 00:10:04.190 --> 00:10:04.200 align:start position:0% can do this in parallel if you have a 00:10:04.200 --> 00:10:06.170 align:start position:0% can do this in parallel if you have a CPU<00:10:04.500> that<00:10:04.680> if<00:10:05.040> you<00:10:05.100> have<00:10:05.220> two<00:10:05.339> CPUs<00:10:05.760> you<00:10:06.120> can 00:10:06.170 --> 00:10:06.180 align:start position:0% CPU that if you have two CPUs you can 00:10:06.180 --> 00:10:08.269 align:start position:0% CPU that if you have two CPUs you can just<00:10:06.300> use<00:10:06.480> two<00:10:06.720> threads<00:10:07.080> to<00:10:07.260> Hash<00:10:07.440> this<00:10:07.740> and 00:10:08.269 --> 00:10:08.279 align:start position:0% just use two threads to Hash this and 00:10:08.279 --> 00:10:10.550 align:start position:0% just use two threads to Hash this and this<00:10:08.459> is<00:10:08.580> exploited<00:10:09.120> in<00:10:09.720> the<00:10:09.839> consensus<00:10:10.200> layer 00:10:10.550 --> 00:10:10.560 align:start position:0% this is exploited in the consensus layer 00:10:10.560 --> 00:10:12.110 align:start position:0% this is exploited in the consensus layer I<00:10:10.740> don't<00:10:10.860> know<00:10:10.980> if<00:10:11.100> any<00:10:11.279> other<00:10:11.519> besides 00:10:12.110 --> 00:10:12.120 align:start position:0% I don't know if any other besides 00:10:12.120 --> 00:10:14.630 align:start position:0% I don't know if any other besides Lighthouse<00:10:12.480> uses<00:10:12.959> this<00:10:13.560> but<00:10:14.040> I<00:10:14.279> want<00:10:14.399> to<00:10:14.519> use 00:10:14.630 --> 00:10:14.640 align:start position:0% Lighthouse uses this but I want to use 00:10:14.640 --> 00:10:17.090 align:start position:0% Lighthouse uses this but I want to use this<00:10:14.880> I<00:10:15.060> want<00:10:15.240> to<00:10:15.360> say<00:10:15.560> to<00:10:16.560> talk<00:10:16.740> about<00:10:16.860> a 00:10:17.090 --> 00:10:17.100 align:start position:0% this I want to say to talk about a 00:10:17.100 --> 00:10:18.650 align:start position:0% this I want to say to talk about a different<00:10:17.220> kind<00:10:17.399> of<00:10:17.519> parallelization<00:10:18.060> you 00:10:18.650 --> 00:10:18.660 align:start position:0% different kind of parallelization you 00:10:18.660 --> 00:10:20.449 align:start position:0% different kind of parallelization you can<00:10:18.779> parallelize<00:10:19.560> this<00:10:19.860> in<00:10:20.279> different 00:10:20.449 --> 00:10:20.459 align:start position:0% can parallelize this in different 00:10:20.459 --> 00:10:22.670 align:start position:0% can parallelize this in different threads<00:10:21.060> but<00:10:21.660> the<00:10:21.839> point<00:10:22.019> I<00:10:22.200> want<00:10:22.380> to<00:10:22.500> make<00:10:22.560> is 00:10:22.670 --> 00:10:22.680 align:start position:0% threads but the point I want to make is 00:10:22.680 --> 00:10:24.470 align:start position:0% threads but the point I want to make is that<00:10:22.860> you<00:10:22.980> can<00:10:23.100> do<00:10:23.220> this<00:10:23.459> by<00:10:23.820> using<00:10:24.120> Vector 00:10:24.470 --> 00:10:24.480 align:start position:0% that you can do this by using Vector 00:10:24.480 --> 00:10:25.910 align:start position:0% that you can do this by using Vector instructions 00:10:25.910 --> 00:10:25.920 align:start position:0% instructions 00:10:25.920 --> 00:10:28.670 align:start position:0% instructions this<00:10:26.580> is<00:10:26.820> the<00:10:27.120> the<00:10:27.480> typical<00:10:27.899> this<00:10:28.380> is<00:10:28.440> the<00:10:28.560> one 00:10:28.670 --> 00:10:28.680 align:start position:0% this is the the typical this is the one 00:10:28.680 --> 00:10:30.290 align:start position:0% this is the the typical this is the one that<00:10:28.800> is<00:10:28.920> in<00:10:29.100> the<00:10:29.220> specs<00:10:29.580> in<00:10:29.760> the<00:10:29.880> consensus 00:10:30.290 --> 00:10:30.300 align:start position:0% that is in the specs in the consensus 00:10:30.300 --> 00:10:32.269 align:start position:0% that is in the specs in the consensus layer<00:10:30.660> specs<00:10:31.080> this<00:10:31.500> is<00:10:31.560> the<00:10:31.620> implementation<00:10:32.040> I 00:10:32.269 --> 00:10:32.279 align:start position:0% layer specs this is the implementation I 00:10:32.279 --> 00:10:33.650 align:start position:0% layer specs this is the implementation I think<00:10:32.339> this<00:10:32.519> is<00:10:32.580> vitalik's<00:10:33.240> implementation 00:10:33.650 --> 00:10:33.660 align:start position:0% think this is vitalik's implementation 00:10:33.660 --> 00:10:36.710 align:start position:0% think this is vitalik's implementation this<00:10:34.320> is<00:10:34.440> a<00:10:34.560> flat<00:10:34.860> array<00:10:35.240> approach<00:10:36.240> to<00:10:36.480> having 00:10:36.710 --> 00:10:36.720 align:start position:0% this is a flat array approach to having 00:10:36.720 --> 00:10:39.350 align:start position:0% this is a flat array approach to having a<00:10:37.080> miracle<00:10:37.500> tree<00:10:37.680> I<00:10:38.459> highlighted<00:10:39.000> the<00:10:39.180> line 00:10:39.350 --> 00:10:39.360 align:start position:0% a miracle tree I highlighted the line 00:10:39.360 --> 00:10:41.810 align:start position:0% a miracle tree I highlighted the line where<00:10:39.899> you're<00:10:40.260> Computing<00:10:40.680> the<00:10:40.860> hash<00:10:41.160> of<00:10:41.640> the 00:10:41.810 --> 00:10:41.820 align:start position:0% where you're Computing the hash of the 00:10:41.820 --> 00:10:45.050 align:start position:0% where you're Computing the hash of the parent<00:10:41.940> by<00:10:42.779> hashing<00:10:43.800> the<00:10:44.399> two<00:10:44.519> children<00:10:44.760> that 00:10:45.050 --> 00:10:45.060 align:start position:0% parent by hashing the two children that 00:10:45.060 --> 00:10:47.750 align:start position:0% parent by hashing the two children that are<00:10:45.240> concatenated<00:10:45.899> the<00:10:46.560> top<00:10:46.680> one<00:10:46.920> is<00:10:47.339> uh 00:10:47.750 --> 00:10:47.760 align:start position:0% are concatenated the top one is uh 00:10:47.760 --> 00:10:50.690 align:start position:0% are concatenated the top one is uh different<00:10:48.440> memory<00:10:49.440> layout<00:10:49.800> this<00:10:50.220> is<00:10:50.339> Proto 00:10:50.690 --> 00:10:50.700 align:start position:0% different memory layout this is Proto 00:10:50.700 --> 00:10:53.389 align:start position:0% different memory layout this is Proto lambda's<00:10:51.420> implementation<00:10:52.200> on<00:10:52.920> remarkable 00:10:53.389 --> 00:10:53.399 align:start position:0% lambda's implementation on remarkable 00:10:53.399 --> 00:10:55.610 align:start position:0% lambda's implementation on remarkable the<00:10:54.060> point<00:10:54.180> is<00:10:54.360> that<00:10:54.660> it's<00:10:54.959> the<00:10:55.200> same<00:10:55.380> kind<00:10:55.500> of 00:10:55.610 --> 00:10:55.620 align:start position:0% the point is that it's the same kind of 00:10:55.620 --> 00:10:57.470 align:start position:0% the point is that it's the same kind of hashing<00:10:56.040> you<00:10:56.220> take<00:10:56.399> the<00:10:56.579> two<00:10:56.760> children<00:10:56.940> and 00:10:57.470 --> 00:10:57.480 align:start position:0% hashing you take the two children and 00:10:57.480 --> 00:10:59.389 align:start position:0% hashing you take the two children and you<00:10:57.540> hash<00:10:57.779> them<00:10:58.019> and<00:10:58.320> you<00:10:58.440> hash<00:10:58.680> two<00:10:58.860> blocks<00:10:59.279> at 00:10:59.389 --> 00:10:59.399 align:start position:0% you hash them and you hash two blocks at 00:10:59.399 --> 00:11:00.590 align:start position:0% you hash them and you hash two blocks at a<00:10:59.579> time 00:11:00.590 --> 00:11:00.600 align:start position:0% a time 00:11:00.600 --> 00:11:03.650 align:start position:0% a time this<00:11:01.140> is<00:11:01.260> fairly<00:11:01.680> fairly<00:11:02.040> inefficient 00:11:03.650 --> 00:11:03.660 align:start position:0% this is fairly fairly inefficient 00:11:03.660 --> 00:11:05.389 align:start position:0% this is fairly fairly inefficient this<00:11:04.140> is<00:11:04.200> a<00:11:04.320> goal<00:11:04.620> implementation<00:11:04.980> this<00:11:05.339> is 00:11:05.389 --> 00:11:05.399 align:start position:0% this is a goal implementation this is 00:11:05.399 --> 00:11:07.130 align:start position:0% this is a goal implementation this is also<00:11:05.700> Jim's<00:11:06.180> uh<00:11:06.600> this<00:11:06.899> is<00:11:06.959> James 00:11:07.130 --> 00:11:07.140 align:start position:0% also Jim's uh this is James 00:11:07.140 --> 00:11:08.690 align:start position:0% also Jim's uh this is James implementation<00:11:07.680> it's<00:11:07.980> a<00:11:08.100> production<00:11:08.339> a 00:11:08.690 --> 00:11:08.700 align:start position:0% implementation it's a production a 00:11:08.700 --> 00:11:10.490 align:start position:0% implementation it's a production a production<00:11:08.940> implementation<00:11:09.540> and<00:11:10.079> then<00:11:10.260> again 00:11:10.490 --> 00:11:10.500 align:start position:0% production implementation and then again 00:11:10.500 --> 00:11:11.810 align:start position:0% production implementation and then again the<00:11:10.800> same<00:11:10.860> thing<00:11:10.980> is<00:11:11.160> slightly<00:11:11.640> more 00:11:11.810 --> 00:11:11.820 align:start position:0% the same thing is slightly more 00:11:11.820 --> 00:11:12.949 align:start position:0% the same thing is slightly more complicated<00:11:12.180> because<00:11:12.420> it<00:11:12.600> takes<00:11:12.779> into 00:11:12.949 --> 00:11:12.959 align:start position:0% complicated because it takes into 00:11:12.959 --> 00:11:14.810 align:start position:0% complicated because it takes into account<00:11:13.140> other<00:11:13.440> things<00:11:13.680> but<00:11:14.160> the<00:11:14.339> highlighted 00:11:14.810 --> 00:11:14.820 align:start position:0% account other things but the highlighted 00:11:14.820 --> 00:11:16.430 align:start position:0% account other things but the highlighted line<00:11:15.000> is<00:11:15.240> the<00:11:15.420> point<00:11:15.540> is<00:11:15.899> the<00:11:16.079> point<00:11:16.200> where 00:11:16.430 --> 00:11:16.440 align:start position:0% line is the point is the point where 00:11:16.440 --> 00:11:19.430 align:start position:0% line is the point is the point where you're<00:11:16.560> hashing<00:11:17.040> and<00:11:17.459> you<00:11:17.640> hash<00:11:18.000> one<00:11:18.480> node<00:11:18.779> as 00:11:19.430 --> 00:11:19.440 align:start position:0% you're hashing and you hash one node as 00:11:19.440 --> 00:11:22.009 align:start position:0% you're hashing and you hash one node as the<00:11:19.860> hash<00:11:19.980> of<00:11:20.399> the<00:11:20.640> two<00:11:20.880> children<00:11:21.120> and<00:11:21.660> you<00:11:21.839> do 00:11:22.009 --> 00:11:22.019 align:start position:0% the hash of the two children and you do 00:11:22.019 --> 00:11:24.470 align:start position:0% the hash of the two children and you do this<00:11:22.079> on<00:11:22.260> a<00:11:22.380> loop<00:11:23.100> for<00:11:23.700> each<00:11:23.940> pair<00:11:24.120> of<00:11:24.300> children 00:11:24.470 --> 00:11:24.480 align:start position:0% this on a loop for each pair of children 00:11:24.480 --> 00:11:26.690 align:start position:0% this on a loop for each pair of children you<00:11:24.779> have<00:11:25.019> once<00:11:25.500> you<00:11:25.740> call<00:11:25.860> the<00:11:26.040> hasher<00:11:26.459> and 00:11:26.690 --> 00:11:26.700 align:start position:0% you have once you call the hasher and 00:11:26.700 --> 00:11:27.949 align:start position:0% you have once you call the hasher and you<00:11:26.760> get<00:11:26.940> one<00:11:27.120> touch 00:11:27.949 --> 00:11:27.959 align:start position:0% you get one touch 00:11:27.959 --> 00:11:30.230 align:start position:0% you get one touch okay<00:11:28.440> so<00:11:28.920> I<00:11:29.399> want<00:11:29.519> to<00:11:29.640> tell<00:11:29.700> you<00:11:29.820> what<00:11:30.000> is<00:11:30.120> the 00:11:30.230 --> 00:11:30.240 align:start position:0% okay so I want to tell you what is the 00:11:30.240 --> 00:11:31.430 align:start position:0% okay so I want to tell you what is the right<00:11:30.420> way 00:11:31.430 --> 00:11:31.440 align:start position:0% right way 00:11:31.440 --> 00:11:33.470 align:start position:0% right way of<00:11:31.860> hash<00:11:31.980> in<00:11:32.220> a<00:11:32.339> miracle<00:11:32.640> tree<00:11:32.760> and<00:11:33.360> it's 00:11:33.470 --> 00:11:33.480 align:start position:0% of hash in a miracle tree and it's 00:11:33.480 --> 00:11:35.269 align:start position:0% of hash in a miracle tree and it's fairly<00:11:33.899> fairly<00:11:34.200> simple<00:11:34.320> so<00:11:34.620> there's<00:11:34.860> two 00:11:35.269 --> 00:11:35.279 align:start position:0% fairly fairly simple so there's two 00:11:35.279 --> 00:11:37.490 align:start position:0% fairly fairly simple so there's two things<00:11:35.519> that<00:11:36.060> we<00:11:36.420> want<00:11:36.540> to<00:11:36.660> exploit<00:11:37.019> we<00:11:37.320> want 00:11:37.490 --> 00:11:37.500 align:start position:0% things that we want to exploit we want 00:11:37.500 --> 00:11:40.610 align:start position:0% things that we want to exploit we want to<00:11:37.560> exploit<00:11:37.980> the<00:11:38.100> fact<00:11:38.220> that<00:11:38.640> our<00:11:39.420> hash<00:11:40.019> is 00:11:40.610 --> 00:11:40.620 align:start position:0% to exploit the fact that our hash is 00:11:40.620 --> 00:11:43.130 align:start position:0% to exploit the fact that our hash is exactly<00:11:41.040> 64<00:11:41.579> bytes<00:11:42.060> that's<00:11:42.660> the<00:11:42.899> thing<00:11:43.019> that 00:11:43.130 --> 00:11:43.140 align:start position:0% exactly 64 bytes that's the thing that 00:11:43.140 --> 00:11:44.449 align:start position:0% exactly 64 bytes that's the thing that we<00:11:43.260> want<00:11:43.380> to<00:11:43.500> hatch 00:11:44.449 --> 00:11:44.459 align:start position:0% we want to hatch 00:11:44.459 --> 00:11:47.210 align:start position:0% we want to hatch so<00:11:44.940> remember<00:11:45.240> how<00:11:45.779> is<00:11:45.899> it<00:11:46.079> that<00:11:46.320> we 00:11:47.210 --> 00:11:47.220 align:start position:0% so remember how is it that we 00:11:47.220 --> 00:11:50.090 align:start position:0% so remember how is it that we patted<00:11:48.000> our<00:11:48.240> message<00:11:48.480> into<00:11:48.959> a<00:11:49.260> multiple<00:11:49.620> of<00:11:49.740> 64 00:11:50.090 --> 00:11:50.100 align:start position:0% patted our message into a multiple of 64 00:11:50.100 --> 00:11:52.790 align:start position:0% patted our message into a multiple of 64 bytes<00:11:51.000> well<00:11:51.480> our<00:11:51.720> message<00:11:51.899> is<00:11:52.140> already<00:11:52.320> 64 00:11:52.790 --> 00:11:52.800 align:start position:0% bytes well our message is already 64 00:11:52.800 --> 00:11:55.550 align:start position:0% bytes well our message is already 64 bytes<00:11:53.220> but<00:11:53.820> unfortunately<00:11:54.300> unfortunately<00:11:55.200> we 00:11:55.550 --> 00:11:55.560 align:start position:0% bytes but unfortunately unfortunately we 00:11:55.560 --> 00:11:58.250 align:start position:0% bytes but unfortunately unfortunately we need<00:11:55.740> to<00:11:55.860> add<00:11:56.100> one<00:11:56.579> bit<00:11:56.760> to<00:11:57.360> tell<00:11:57.540> the<00:11:58.140> message 00:11:58.250 --> 00:11:58.260 align:start position:0% need to add one bit to tell the message 00:11:58.260 --> 00:12:01.430 align:start position:0% need to add one bit to tell the message I<00:11:58.560> send<00:11:58.740> it<00:11:59.399> so<00:11:59.880> that's<00:12:00.000> the<00:12:00.240> first<00:12:00.420> bit<00:12:00.600> there 00:12:01.430 --> 00:12:01.440 align:start position:0% I send it so that's the first bit there 00:12:01.440 --> 00:12:03.170 align:start position:0% I send it so that's the first bit there the<00:12:01.920> problem<00:12:02.040> is<00:12:02.279> that<00:12:02.459> since<00:12:02.640> we<00:12:02.820> added<00:12:03.120> that 00:12:03.170 --> 00:12:03.180 align:start position:0% the problem is that since we added that 00:12:03.180 --> 00:12:05.269 align:start position:0% the problem is that since we added that bit<00:12:03.360> now<00:12:03.720> we<00:12:03.899> need<00:12:04.019> an<00:12:04.200> entire<00:12:04.440> block 00:12:05.269 --> 00:12:05.279 align:start position:0% bit now we need an entire block 00:12:05.279 --> 00:12:07.670 align:start position:0% bit now we need an entire block and<00:12:05.700> the<00:12:05.880> entire<00:12:06.120> block<00:12:06.360> is<00:12:06.779> all<00:12:07.079> made<00:12:07.500> of 00:12:07.670 --> 00:12:07.680 align:start position:0% and the entire block is all made of 00:12:07.680 --> 00:12:10.850 align:start position:0% and the entire block is all made of zeros<00:12:08.100> except<00:12:09.060> this<00:12:09.720> little<00:12:09.959> bit<00:12:10.200> one<00:12:10.500> here 00:12:10.850 --> 00:12:10.860 align:start position:0% zeros except this little bit one here 00:12:10.860 --> 00:12:13.310 align:start position:0% zeros except this little bit one here and<00:12:11.339> that<00:12:11.519> bit<00:12:11.640> one<00:12:11.880> here<00:12:12.240> is<00:12:12.420> saying<00:12:12.839> the 00:12:13.310 --> 00:12:13.320 align:start position:0% and that bit one here is saying the 00:12:13.320 --> 00:12:15.710 align:start position:0% and that bit one here is saying the message<00:12:13.500> was<00:12:13.860> 64<00:12:14.399> bytes 00:12:15.710 --> 00:12:15.720 align:start position:0% message was 64 bytes 00:12:15.720 --> 00:12:18.889 align:start position:0% message was 64 bytes so<00:12:16.079> that<00:12:16.440> bit<00:12:16.560> there<00:12:16.800> is<00:12:16.980> a<00:12:17.160> 512<00:12:17.880> bits 00:12:18.889 --> 00:12:18.899 align:start position:0% so that bit there is a 512 bits 00:12:18.899 --> 00:12:21.470 align:start position:0% so that bit there is a 512 bits that<00:12:19.500> uh<00:12:19.740> that<00:12:19.920> the<00:12:20.100> message<00:12:20.220> was 00:12:21.470 --> 00:12:21.480 align:start position:0% that uh that the message was 00:12:21.480 --> 00:12:23.210 align:start position:0% that uh that the message was but<00:12:21.779> the<00:12:22.019> point<00:12:22.140> is<00:12:22.320> that<00:12:22.560> this<00:12:22.740> block<00:12:22.920> is 00:12:23.210 --> 00:12:23.220 align:start position:0% but the point is that this block is 00:12:23.220 --> 00:12:25.009 align:start position:0% but the point is that this block is known<00:12:23.519> it's<00:12:23.880> known<00:12:24.180> before<00:12:24.420> we<00:12:24.660> started 00:12:25.009 --> 00:12:25.019 align:start position:0% known it's known before we started 00:12:25.019 --> 00:12:28.310 align:start position:0% known it's known before we started hashing<00:12:25.740> the<00:12:26.339> whole<00:12:26.459> block<00:12:26.700> is<00:12:26.940> known<00:12:27.120> and<00:12:28.079> in 00:12:28.310 --> 00:12:28.320 align:start position:0% hashing the whole block is known and in 00:12:28.320 --> 00:12:31.069 align:start position:0% hashing the whole block is known and in order<00:12:28.440> to<00:12:28.620> compute<00:12:29.040> the<00:12:29.220> scheduled<00:12:29.700> words<00:12:30.079> we 00:12:31.069 --> 00:12:31.079 align:start position:0% order to compute the scheduled words we 00:12:31.079 --> 00:12:32.630 align:start position:0% order to compute the scheduled words we only<00:12:31.200> need<00:12:31.440> to<00:12:31.500> know<00:12:31.680> the<00:12:31.860> message<00:12:32.040> we<00:12:32.519> don't 00:12:32.630 --> 00:12:32.640 align:start position:0% only need to know the message we don't 00:12:32.640 --> 00:12:34.490 align:start position:0% only need to know the message we don't need<00:12:32.760> to<00:12:32.940> know<00:12:33.060> the<00:12:33.300> hash<00:12:33.540> of<00:12:34.200> the<00:12:34.320> previous 00:12:34.490 --> 00:12:34.500 align:start position:0% need to know the hash of the previous 00:12:34.500 --> 00:12:38.690 align:start position:0% need to know the hash of the previous block<00:12:35.339> so<00:12:36.060> whenever<00:12:36.600> we<00:12:36.839> hash<00:12:37.079> 64<00:12:37.680> bytes<00:12:38.160> we 00:12:38.690 --> 00:12:38.700 align:start position:0% block so whenever we hash 64 bytes we 00:12:38.700 --> 00:12:42.230 align:start position:0% block so whenever we hash 64 bytes we can<00:12:38.959> already<00:12:39.959> compute<00:12:40.680> the<00:12:41.579> scheduled<00:12:42.060> words 00:12:42.230 --> 00:12:42.240 align:start position:0% can already compute the scheduled words 00:12:42.240 --> 00:12:44.210 align:start position:0% can already compute the scheduled words for<00:12:42.600> this<00:12:42.779> entire<00:12:43.019> block<00:12:43.260> and<00:12:43.680> we<00:12:43.800> can<00:12:43.920> do<00:12:44.040> this 00:12:44.210 --> 00:12:44.220 align:start position:0% for this entire block and we can do this 00:12:44.220 --> 00:12:46.550 align:start position:0% for this entire block and we can do this now<00:12:44.459> we<00:12:44.940> have<00:12:45.060> them<00:12:45.300> so<00:12:45.600> we<00:12:45.779> can<00:12:45.959> cardboard<00:12:46.320> it 00:12:46.550 --> 00:12:46.560 align:start position:0% now we have them so we can cardboard it 00:12:46.560 --> 00:12:49.430 align:start position:0% now we have them so we can cardboard it in<00:12:46.740> the<00:12:46.920> in<00:12:47.399> the<00:12:47.579> in<00:12:47.820> the<00:12:48.000> hasher<00:12:48.360> itself<00:12:49.079> this 00:12:49.430 --> 00:12:49.440 align:start position:0% in the in the in the hasher itself this 00:12:49.440 --> 00:12:50.690 align:start position:0% in the in the in the hasher itself this is<00:12:49.560> something<00:12:49.680> that<00:12:49.920> the<00:12:50.100> Bitcoin<00:12:50.519> Community 00:12:50.690 --> 00:12:50.700 align:start position:0% is something that the Bitcoin Community 00:12:50.700 --> 00:12:52.850 align:start position:0% is something that the Bitcoin Community learned<00:12:51.360> and<00:12:51.959> they<00:12:52.139> have<00:12:52.260> it<00:12:52.440> in<00:12:52.620> their 00:12:52.850 --> 00:12:52.860 align:start position:0% learned and they have it in their 00:12:52.860 --> 00:12:56.030 align:start position:0% learned and they have it in their standard<00:12:53.639> node<00:12:54.240> so<00:12:54.779> the<00:12:55.079> the 00:12:56.030 --> 00:12:56.040 align:start position:0% standard node so the the 00:12:56.040 --> 00:12:57.290 align:start position:0% standard node so the the I<00:12:56.279> don't<00:12:56.399> know<00:12:56.519> how<00:12:56.579> is<00:12:56.700> it<00:12:56.820> called<00:12:56.940> I<00:12:57.180> think 00:12:57.290 --> 00:12:57.300 align:start position:0% I don't know how is it called I think 00:12:57.300 --> 00:12:59.509 align:start position:0% I don't know how is it called I think it's<00:12:57.360> Bitcoin<00:12:57.839> core<00:12:58.019> the<00:12:58.320> client<00:12:58.620> and<00:12:59.220> it<00:12:59.339> has 00:12:59.509 --> 00:12:59.519 align:start position:0% it's Bitcoin core the client and it has 00:12:59.519 --> 00:13:02.389 align:start position:0% it's Bitcoin core the client and it has it<00:12:59.639> there<00:12:59.820> and<00:13:00.540> no<00:13:00.899> no<00:13:01.200> ethereum<00:13:01.740> client<00:13:02.040> used 00:13:02.389 --> 00:13:02.399 align:start position:0% it there and no no ethereum client used 00:13:02.399 --> 00:13:04.250 align:start position:0% it there and no no ethereum client used this<00:13:02.700> so<00:13:03.240> I<00:13:03.420> was<00:13:03.540> surprised<00:13:03.839> when<00:13:03.959> I<00:13:04.079> found 00:13:04.250 --> 00:13:04.260 align:start position:0% this so I was surprised when I found 00:13:04.260 --> 00:13:06.050 align:start position:0% this so I was surprised when I found this<00:13:04.440> out<00:13:04.620> so<00:13:04.920> we<00:13:05.100> can<00:13:05.220> stop<00:13:05.399> we<00:13:05.760> can<00:13:05.820> steal 00:13:06.050 --> 00:13:06.060 align:start position:0% this out so we can stop we can steal 00:13:06.060 --> 00:13:07.790 align:start position:0% this out so we can stop we can steal this<00:13:06.180> from<00:13:06.300> the<00:13:06.420> Bitcoin<00:13:06.899> community<00:13:07.019> and 00:13:07.790 --> 00:13:07.800 align:start position:0% this from the Bitcoin community and 00:13:07.800 --> 00:13:10.310 align:start position:0% this from the Bitcoin community and indeed<00:13:08.220> uh<00:13:08.940> so<00:13:09.060> prism<00:13:09.480> uses<00:13:09.959> this<00:13:10.139> Library 00:13:10.310 --> 00:13:10.320 align:start position:0% indeed uh so prism uses this Library 00:13:10.320 --> 00:13:12.290 align:start position:0% indeed uh so prism uses this Library loads<00:13:11.220> that<00:13:11.339> are<00:13:11.459> all<00:13:11.579> to<00:13:11.760> implemented<00:13:12.180> this 00:13:12.290 --> 00:13:12.300 align:start position:0% loads that are all to implemented this 00:13:12.300 --> 00:13:14.329 align:start position:0% loads that are all to implemented this and<00:13:12.540> just<00:13:12.660> by<00:13:12.839> implementing<00:13:13.320> this<00:13:13.440> you<00:13:13.680> get<00:13:13.860> at 00:13:14.329 --> 00:13:14.339 align:start position:0% and just by implementing this you get at 00:13:14.339 --> 00:13:17.930 align:start position:0% and just by implementing this you get at least<00:13:14.459> 20<00:13:14.639> gain<00:13:15.420> on<00:13:16.019> your<00:13:16.139> hashing<00:13:16.560> speed<00:13:17.100> just 00:13:17.930 --> 00:13:17.940 align:start position:0% least 20 gain on your hashing speed just 00:13:17.940 --> 00:13:20.389 align:start position:0% least 20 gain on your hashing speed just this<00:13:18.120> no<00:13:18.600> changes<00:13:18.959> on<00:13:19.139> your<00:13:19.260> code<00:13:19.440> just 00:13:20.389 --> 00:13:20.399 align:start position:0% this no changes on your code just 00:13:20.399 --> 00:13:23.990 align:start position:0% this no changes on your code just include<00:13:21.120> this<00:13:21.600> uh<00:13:22.320> hardcode<00:13:22.740> words 00:13:23.990 --> 00:13:24.000 align:start position:0% include this uh hardcode words 00:13:24.000 --> 00:13:26.449 align:start position:0% include this uh hardcode words the<00:13:24.540> second<00:13:24.660> one<00:13:24.839> is<00:13:25.079> vectorization<00:13:25.560> itself 00:13:26.449 --> 00:13:26.459 align:start position:0% the second one is vectorization itself 00:13:26.459 --> 00:13:28.190 align:start position:0% the second one is vectorization itself and<00:13:26.940> vectorization 00:13:28.190 --> 00:13:28.200 align:start position:0% and vectorization 00:13:28.200 --> 00:13:29.629 align:start position:0% and vectorization uh 00:13:29.629 --> 00:13:29.639 align:start position:0% uh 00:13:29.639 --> 00:13:31.009 align:start position:0% uh this<00:13:30.120> is<00:13:30.180> also<00:13:30.480> something<00:13:30.660> that<00:13:30.899> is 00:13:31.009 --> 00:13:31.019 align:start position:0% this is also something that is 00:13:31.019 --> 00:13:33.650 align:start position:0% this is also something that is implemented<00:13:31.560> already<00:13:32.160> if<00:13:32.639> you<00:13:32.880> have<00:13:33.060> several 00:13:33.650 --> 00:13:33.660 align:start position:0% implemented already if you have several 00:13:33.660 --> 00:13:35.870 align:start position:0% implemented already if you have several messages<00:13:34.079> that<00:13:34.620> you<00:13:34.740> need<00:13:34.860> to<00:13:34.980> Hash<00:13:35.160> at<00:13:35.760> the 00:13:35.870 --> 00:13:35.880 align:start position:0% messages that you need to Hash at the 00:13:35.880 --> 00:13:37.610 align:start position:0% messages that you need to Hash at the same<00:13:36.000> time<00:13:36.180> there<00:13:36.480> are<00:13:36.660> libraries<00:13:37.019> to<00:13:37.260> do<00:13:37.440> this 00:13:37.610 --> 00:13:37.620 align:start position:0% same time there are libraries to do this 00:13:37.620 --> 00:13:40.490 align:start position:0% same time there are libraries to do this Intel<00:13:38.399> has<00:13:38.579> a<00:13:38.760> very<00:13:38.880> good<00:13:39.000> library<00:13:39.180> and<00:13:40.079> all<00:13:40.260> I 00:13:40.490 --> 00:13:40.500 align:start position:0% Intel has a very good library and all I 00:13:40.500 --> 00:13:43.009 align:start position:0% Intel has a very good library and all I need<00:13:40.620> is<00:13:40.860> just<00:13:41.160> change<00:13:41.579> the<00:13:42.300> signature<00:13:42.839> of 00:13:43.009 --> 00:13:43.019 align:start position:0% need is just change the signature of 00:13:43.019 --> 00:13:44.750 align:start position:0% need is just change the signature of that<00:13:43.200> library<00:13:43.440> but<00:13:44.160> the<00:13:44.279> point<00:13:44.399> is<00:13:44.519> that<00:13:44.639> you 00:13:44.750 --> 00:13:44.760 align:start position:0% that library but the point is that you 00:13:44.760 --> 00:13:46.730 align:start position:0% that library but the point is that you can<00:13:45.180> be<00:13:45.300> hashing<00:13:45.720> several<00:13:46.019> buffers<00:13:46.380> at<00:13:46.620> the 00:13:46.730 --> 00:13:46.740 align:start position:0% can be hashing several buffers at the 00:13:46.740 --> 00:13:49.370 align:start position:0% can be hashing several buffers at the same<00:13:46.860> time<00:13:47.040> if<00:13:47.820> you<00:13:48.000> have<00:13:48.180> so<00:13:48.839> this<00:13:49.019> is<00:13:49.079> a<00:13:49.260> way 00:13:49.370 --> 00:13:49.380 align:start position:0% same time if you have so this is a way 00:13:49.380 --> 00:13:52.090 align:start position:0% same time if you have so this is a way of<00:13:49.560> like<00:13:49.740> encoding<00:13:50.339> the<00:13:50.700> words<00:13:50.940> again 00:13:52.090 --> 00:13:52.100 align:start position:0% of like encoding the words again 00:13:52.100 --> 00:13:54.410 align:start position:0% of like encoding the words again uh<00:13:53.100> you're<00:13:53.399> gonna<00:13:53.519> take<00:13:53.760> this<00:13:54.000> is<00:13:54.120> the<00:13:54.240> message 00:13:54.410 --> 00:13:54.420 align:start position:0% uh you're gonna take this is the message 00:13:54.420 --> 00:13:56.750 align:start position:0% uh you're gonna take this is the message that<00:13:54.779> you<00:13:54.899> want<00:13:55.019> to<00:13:55.200> you<00:13:55.380> want<00:13:55.560> to<00:13:55.680> hash 00:13:56.750 --> 00:13:56.760 align:start position:0% that you want to you want to hash 00:13:56.760 --> 00:13:58.730 align:start position:0% that you want to you want to hash uh<00:13:57.480> so<00:13:57.660> these<00:13:57.959> are<00:13:58.079> the<00:13:58.139> 16<00:13:58.320> words 00:13:58.730 --> 00:13:58.740 align:start position:0% uh so these are the 16 words 00:13:58.740 --> 00:14:01.009 align:start position:0% uh so these are the 16 words corresponding<00:13:59.639> to<00:13:59.820> this<00:14:00.060> node<00:14:00.360> here<00:14:00.660> the 00:14:01.009 --> 00:14:01.019 align:start position:0% corresponding to this node here the 00:14:01.019 --> 00:14:03.170 align:start position:0% corresponding to this node here the zeroth<00:14:01.380> note<00:14:01.560> then<00:14:02.339> you're<00:14:02.459> gonna<00:14:02.639> have<00:14:02.880> 16 00:14:03.170 --> 00:14:03.180 align:start position:0% zeroth note then you're gonna have 16 00:14:03.180 --> 00:14:05.090 align:start position:0% zeroth note then you're gonna have 16 words<00:14:03.540> corresponding<00:14:04.260> to<00:14:04.380> this<00:14:04.560> one<00:14:04.740> that<00:14:05.040> you 00:14:05.090 --> 00:14:05.100 align:start position:0% words corresponding to this one that you 00:14:05.100 --> 00:14:07.430 align:start position:0% words corresponding to this one that you want<00:14:05.220> to<00:14:05.339> Hash<00:14:05.579> 16<00:14:06.060> words<00:14:06.360> for<00:14:06.779> these<00:14:07.079> two<00:14:07.139> and 00:14:07.430 --> 00:14:07.440 align:start position:0% want to Hash 16 words for these two and 00:14:07.440 --> 00:14:09.110 align:start position:0% want to Hash 16 words for these two and 16<00:14:07.680> words<00:14:07.980> for<00:14:08.220> these<00:14:08.459> two 00:14:09.110 --> 00:14:09.120 align:start position:0% 16 words for these two 00:14:09.120 --> 00:14:10.970 align:start position:0% 16 words for these two and<00:14:09.660> what<00:14:09.899> you're<00:14:10.019> going<00:14:10.139> to<00:14:10.260> do<00:14:10.380> is<00:14:10.560> collect 00:14:10.970 --> 00:14:10.980 align:start position:0% and what you're going to do is collect 00:14:10.980 --> 00:14:13.910 align:start position:0% and what you're going to do is collect the<00:14:11.579> first<00:14:11.700> double<00:14:11.940> word<00:14:12.240> from<00:14:12.600> this<00:14:12.839> one<00:14:13.019> the 00:14:13.910 --> 00:14:13.920 align:start position:0% the first double word from this one the 00:14:13.920 --> 00:14:15.590 align:start position:0% the first double word from this one the first<00:14:14.040> double<00:14:14.220> word<00:14:14.519> from<00:14:14.760> this<00:14:15.060> one<00:14:15.180> the 00:14:15.590 --> 00:14:15.600 align:start position:0% first double word from this one the 00:14:15.600 --> 00:14:16.970 align:start position:0% first double word from this one the first<00:14:15.720> one<00:14:15.899> will<00:14:16.139> work<00:14:16.320> from<00:14:16.500> this<00:14:16.680> one<00:14:16.800> and 00:14:16.970 --> 00:14:16.980 align:start position:0% first one will work from this one and 00:14:16.980 --> 00:14:18.889 align:start position:0% first one will work from this one and the<00:14:17.100> first<00:14:17.220> double<00:14:17.339> word<00:14:17.639> from<00:14:17.880> this<00:14:18.060> one<00:14:18.240> into 00:14:18.889 --> 00:14:18.899 align:start position:0% the first double word from this one into 00:14:18.899 --> 00:14:20.210 align:start position:0% the first double word from this one into one<00:14:19.200> register 00:14:20.210 --> 00:14:20.220 align:start position:0% one register 00:14:20.220 --> 00:14:21.949 align:start position:0% one register that's<00:14:20.639> if<00:14:20.820> you<00:14:20.940> have<00:14:21.120> registers<00:14:21.600> that<00:14:21.839> are 00:14:21.949 --> 00:14:21.959 align:start position:0% that's if you have registers that are 00:14:21.959 --> 00:14:25.310 align:start position:0% that's if you have registers that are 128<00:14:22.800> bits<00:14:23.279> if<00:14:23.820> you<00:14:23.940> had<00:14:24.120> avx12<00:14:24.959> you<00:14:25.200> would 00:14:25.310 --> 00:14:25.320 align:start position:0% 128 bits if you had avx12 you would 00:14:25.320 --> 00:14:27.769 align:start position:0% 128 bits if you had avx12 you would collect<00:14:25.620> eight<00:14:26.160> of<00:14:26.459> them<00:14:26.579> at<00:14:26.760> the<00:14:26.880> same<00:14:27.000> time 00:14:27.769 --> 00:14:27.779 align:start position:0% collect eight of them at the same time 00:14:27.779 --> 00:14:31.490 align:start position:0% collect eight of them at the same time and<00:14:28.200> again<00:14:28.380> if<00:14:28.680> you're<00:14:28.860> a<00:14:29.100> AVX<00:14:30.000> 512<00:14:30.959> you<00:14:31.200> can<00:14:31.320> do 00:14:31.490 --> 00:14:31.500 align:start position:0% and again if you're a AVX 512 you can do 00:14:31.500 --> 00:14:33.110 align:start position:0% and again if you're a AVX 512 you can do 16<00:14:31.800> at<00:14:32.220> a<00:14:32.399> time 00:14:33.110 --> 00:14:33.120 align:start position:0% 16 at a time 00:14:33.120 --> 00:14:36.530 align:start position:0% 16 at a time the<00:14:33.600> point<00:14:33.720> is<00:14:33.959> that<00:14:34.320> you<00:14:34.800> can<00:14:34.920> hash<00:14:35.360> in<00:14:36.360> one 00:14:36.530 --> 00:14:36.540 align:start position:0% the point is that you can hash in one 00:14:36.540 --> 00:14:37.490 align:start position:0% the point is that you can hash in one pass 00:14:37.490 --> 00:14:37.500 align:start position:0% pass 00:14:37.500 --> 00:14:41.629 align:start position:0% pass you<00:14:37.980> can<00:14:38.100> hash<00:14:38.399> up<00:14:39.300> to<00:14:39.480> 16<00:14:39.660> blocks<00:14:40.199> on<00:14:40.440> AVX<00:14:40.920> 512 00:14:41.629 --> 00:14:41.639 align:start position:0% you can hash up to 16 blocks on AVX 512 00:14:41.639 --> 00:14:43.610 align:start position:0% you can hash up to 16 blocks on AVX 512 and<00:14:42.240> you<00:14:42.420> could<00:14:42.540> do<00:14:42.660> the<00:14:42.779> same<00:14:42.899> thing<00:14:43.079> for<00:14:43.380> the 00:14:43.610 --> 00:14:43.620 align:start position:0% and you could do the same thing for the 00:14:43.620 --> 00:14:46.310 align:start position:0% and you could do the same thing for the words<00:14:43.860> themselves<00:14:44.459> so<00:14:45.000> the<00:14:45.240> the<00:14:45.420> digest 00:14:46.310 --> 00:14:46.320 align:start position:0% words themselves so the the digest 00:14:46.320 --> 00:14:49.189 align:start position:0% words themselves so the the digest consists<00:14:47.279> of<00:14:47.399> eight<00:14:47.880> double<00:14:48.180> words 00:14:49.189 --> 00:14:49.199 align:start position:0% consists of eight double words 00:14:49.199 --> 00:14:52.370 align:start position:0% consists of eight double words if<00:14:49.620> you<00:14:49.800> have<00:14:50.100> registers<00:14:50.820> that<00:14:51.060> are<00:14:51.180> 128<00:14:51.959> bits 00:14:52.370 --> 00:14:52.380 align:start position:0% if you have registers that are 128 bits 00:14:52.380 --> 00:14:54.530 align:start position:0% if you have registers that are 128 bits you<00:14:52.920> can<00:14:53.040> put<00:14:53.339> this 00:14:54.530 --> 00:14:54.540 align:start position:0% you can put this 00:14:54.540 --> 00:14:57.110 align:start position:0% you can put this you<00:14:55.139> can<00:14:55.199> put<00:14:55.440> four<00:14:55.800> of<00:14:56.040> them<00:14:56.220> in<00:14:56.579> one<00:14:56.699> register 00:14:57.110 --> 00:14:57.120 align:start position:0% you can put four of them in one register 00:14:57.120 --> 00:14:59.930 align:start position:0% you can put four of them in one register and<00:14:58.019> then<00:14:58.199> you<00:14:58.560> use<00:14:58.740> eight<00:14:59.100> such<00:14:59.399> registers 00:14:59.930 --> 00:14:59.940 align:start position:0% and then you use eight such registers 00:14:59.940 --> 00:15:02.269 align:start position:0% and then you use eight such registers for<00:15:00.420> the<00:15:00.600> eight<00:15:00.720> words<00:15:01.079> of<00:15:01.380> the<00:15:01.500> digest<00:15:01.860> and 00:15:02.269 --> 00:15:02.279 align:start position:0% for the eight words of the digest and 00:15:02.279 --> 00:15:04.310 align:start position:0% for the eight words of the digest and you<00:15:02.399> can<00:15:02.519> hash<00:15:02.760> four<00:15:03.060> blocks<00:15:03.420> at<00:15:03.600> a<00:15:03.720> time<00:15:03.779> so 00:15:04.310 --> 00:15:04.320 align:start position:0% you can hash four blocks at a time so 00:15:04.320 --> 00:15:05.930 align:start position:0% you can hash four blocks at a time so this<00:15:04.500> is<00:15:04.620> again<00:15:04.920> this<00:15:05.279> is<00:15:05.339> a<00:15:05.459> library<00:15:05.639> that 00:15:05.930 --> 00:15:05.940 align:start position:0% this is again this is a library that 00:15:05.940 --> 00:15:07.670 align:start position:0% this is again this is a library that exists<00:15:06.300> so<00:15:06.600> all<00:15:06.899> I'm<00:15:07.139> selling<00:15:07.440> you<00:15:07.560> is 00:15:07.670 --> 00:15:07.680 align:start position:0% exists so all I'm selling you is 00:15:07.680 --> 00:15:09.170 align:start position:0% exists so all I'm selling you is something<00:15:07.860> that<00:15:08.100> already<00:15:08.279> exists<00:15:08.820> it's<00:15:09.000> not 00:15:09.170 --> 00:15:09.180 align:start position:0% something that already exists it's not 00:15:09.180 --> 00:15:11.449 align:start position:0% something that already exists it's not that<00:15:09.300> we're<00:15:09.420> rolling<00:15:09.779> out<00:15:09.959> our<00:15:10.560> own<00:15:10.680> crypto<00:15:11.160> is 00:15:11.449 --> 00:15:11.459 align:start position:0% that we're rolling out our own crypto is 00:15:11.459 --> 00:15:13.490 align:start position:0% that we're rolling out our own crypto is I<00:15:12.060> need<00:15:12.180> to<00:15:12.300> brag<00:15:12.540> about<00:15:12.660> something<00:15:12.959> I<00:15:13.260> was 00:15:13.490 --> 00:15:13.500 align:start position:0% I need to brag about something I was 00:15:13.500 --> 00:15:15.710 align:start position:0% I need to brag about something I was told<00:15:13.680> as<00:15:13.920> a<00:15:14.100> student<00:15:14.220> that<00:15:14.579> if<00:15:15.240> you're<00:15:15.360> given<00:15:15.600> a 00:15:15.710 --> 00:15:15.720 align:start position:0% told as a student that if you're given a 00:15:15.720 --> 00:15:16.790 align:start position:0% told as a student that if you're given a talk<00:15:15.779> you<00:15:16.019> need<00:15:16.139> to<00:15:16.199> brag<00:15:16.440> about<00:15:16.560> something 00:15:16.790 --> 00:15:16.800 align:start position:0% talk you need to brag about something 00:15:16.800 --> 00:15:18.110 align:start position:0% talk you need to brag about something that<00:15:17.040> you<00:15:17.160> did<00:15:17.279> yourself 00:15:18.110 --> 00:15:18.120 align:start position:0% that you did yourself 00:15:18.120 --> 00:15:20.750 align:start position:0% that you did yourself and<00:15:18.660> I<00:15:18.899> guess<00:15:18.959> this<00:15:19.260> is<00:15:19.320> something<00:15:19.500> I<00:15:19.800> did<00:15:19.980> uh 00:15:20.750 --> 00:15:20.760 align:start position:0% and I guess this is something I did uh 00:15:20.760 --> 00:15:23.449 align:start position:0% and I guess this is something I did uh of<00:15:21.180> course<00:15:21.300> the<00:15:21.540> the<00:15:21.800> assembly<00:15:22.800> for<00:15:22.980> Intel<00:15:23.279> and 00:15:23.449 --> 00:15:23.459 align:start position:0% of course the the assembly for Intel and 00:15:23.459 --> 00:15:26.030 align:start position:0% of course the the assembly for Intel and AMD<00:15:23.940> Intel's<00:15:24.720> libraries<00:15:25.139> I'm<00:15:25.620> not<00:15:25.740> gonna<00:15:25.800> beat 00:15:26.030 --> 00:15:26.040 align:start position:0% AMD Intel's libraries I'm not gonna beat 00:15:26.040 --> 00:15:29.509 align:start position:0% AMD Intel's libraries I'm not gonna beat it<00:15:26.579> uh<00:15:27.480> unfortunately<00:15:28.019> Intel<00:15:28.800> or<00:15:29.100> reasonably 00:15:29.509 --> 00:15:29.519 align:start position:0% it uh unfortunately Intel or reasonably 00:15:29.519 --> 00:15:32.930 align:start position:0% it uh unfortunately Intel or reasonably Intel<00:15:30.060> does<00:15:30.480> not<00:15:30.720> produce<00:15:31.079> assembly<00:15:31.620> for<00:15:31.800> arm 00:15:32.930 --> 00:15:32.940 align:start position:0% Intel does not produce assembly for arm 00:15:32.940 --> 00:15:34.449 align:start position:0% Intel does not produce assembly for arm and<00:15:33.240> there's<00:15:33.360> something<00:15:33.540> interesting<00:15:34.019> here 00:15:34.449 --> 00:15:34.459 align:start position:0% and there's something interesting here 00:15:34.459 --> 00:15:37.730 align:start position:0% and there's something interesting here uh<00:15:35.459> arm<00:15:35.699> assembly<00:15:36.300> I<00:15:36.720> told<00:15:36.839> you<00:15:37.019> that<00:15:37.260> every 00:15:37.730 --> 00:15:37.740 align:start position:0% uh arm assembly I told you that every 00:15:37.740 --> 00:15:39.889 align:start position:0% uh arm assembly I told you that every Library<00:15:38.100> out<00:15:38.519> there<00:15:38.699> uses<00:15:39.120> vectorization<00:15:39.600> to 00:15:39.889 --> 00:15:39.899 align:start position:0% Library out there uses vectorization to 00:15:39.899 --> 00:15:41.810 align:start position:0% Library out there uses vectorization to compute<00:15:40.199> the<00:15:40.320> scheduled<00:15:40.680> words<00:15:40.800> that's<00:15:41.519> not 00:15:41.810 --> 00:15:41.820 align:start position:0% compute the scheduled words that's not 00:15:41.820 --> 00:15:44.329 align:start position:0% compute the scheduled words that's not true<00:15:42.060> on<00:15:42.480> arm<00:15:42.720> on<00:15:43.500> arm 00:15:44.329 --> 00:15:44.339 align:start position:0% true on arm on arm 00:15:44.339 --> 00:15:47.210 align:start position:0% true on arm on arm open<00:15:45.060> SSL<00:15:45.720> had<00:15:46.019> a<00:15:46.320> library 00:15:47.210 --> 00:15:47.220 align:start position:0% open SSL had a library 00:15:47.220 --> 00:15:50.569 align:start position:0% open SSL had a library uh<00:15:48.180> using<00:15:48.959> neon<00:15:49.440> instructions<00:15:49.980> Vector 00:15:50.569 --> 00:15:50.579 align:start position:0% uh using neon instructions Vector 00:15:50.579 --> 00:15:52.430 align:start position:0% uh using neon instructions Vector instructions<00:15:51.000> to<00:15:51.180> compute<00:15:51.540> the<00:15:51.660> the<00:15:51.779> hashing 00:15:52.430 --> 00:15:52.440 align:start position:0% instructions to compute the the hashing 00:15:52.440 --> 00:15:54.769 align:start position:0% instructions to compute the the hashing and<00:15:52.860> it<00:15:52.920> turned<00:15:53.220> out<00:15:53.279> to<00:15:53.519> be<00:15:53.760> slower<00:15:54.060> than<00:15:54.360> than 00:15:54.769 --> 00:15:54.779 align:start position:0% and it turned out to be slower than than 00:15:54.779 --> 00:15:58.550 align:start position:0% and it turned out to be slower than than scalar<00:15:55.440> so<00:15:56.399> most<00:15:57.300> libraries<00:15:57.720> for<00:15:57.899> arm<00:15:58.079> do<00:15:58.380> not 00:15:58.550 --> 00:15:58.560 align:start position:0% scalar so most libraries for arm do not 00:15:58.560 --> 00:16:01.550 align:start position:0% scalar so most libraries for arm do not use<00:15:59.060> scalar<00:16:00.060> instructions 00:16:01.550 --> 00:16:01.560 align:start position:0% use scalar instructions 00:16:01.560 --> 00:16:04.129 align:start position:0% use scalar instructions however<00:16:02.399> if<00:16:02.880> you're<00:16:03.000> gonna<00:16:03.240> Hash<00:16:03.540> a<00:16:03.779> miracle 00:16:04.129 --> 00:16:04.139 align:start position:0% however if you're gonna Hash a miracle 00:16:04.139 --> 00:16:05.090 align:start position:0% however if you're gonna Hash a miracle tree 00:16:05.090 --> 00:16:05.100 align:start position:0% tree 00:16:05.100 --> 00:16:07.009 align:start position:0% tree you<00:16:05.639> can<00:16:05.760> hash<00:16:05.940> several<00:16:06.420> blocks<00:16:06.720> at<00:16:06.839> the<00:16:06.959> same 00:16:07.009 --> 00:16:07.019 align:start position:0% you can hash several blocks at the same 00:16:07.019 --> 00:16:08.569 align:start position:0% you can hash several blocks at the same time<00:16:07.139> and<00:16:07.440> that's<00:16:07.620> very<00:16:07.920> very<00:16:08.100> much<00:16:08.339> much 00:16:08.569 --> 00:16:08.579 align:start position:0% time and that's very very much much 00:16:08.579 --> 00:16:11.329 align:start position:0% time and that's very very much much faster<00:16:09.000> than<00:16:09.600> hashing<00:16:10.079> them<00:16:10.199> on<00:16:10.380> scale<00:16:10.620> so<00:16:11.160> the 00:16:11.329 --> 00:16:11.339 align:start position:0% faster than hashing them on scale so the 00:16:11.339 --> 00:16:13.610 align:start position:0% faster than hashing them on scale so the pipelining<00:16:11.880> for<00:16:12.060> arm<00:16:12.180> I<00:16:12.540> think<00:16:12.720> is<00:16:12.899> purely 00:16:13.610 --> 00:16:13.620 align:start position:0% pipelining for arm I think is purely 00:16:13.620 --> 00:16:14.689 align:start position:0% pipelining for arm I think is purely mine 00:16:14.689 --> 00:16:14.699 align:start position:0% mine 00:16:14.699 --> 00:16:16.189 align:start position:0% mine let's<00:16:15.120> say<00:16:15.300> that<00:16:15.420> you<00:16:15.600> want<00:16:15.720> to<00:16:15.839> use<00:16:15.959> this 00:16:16.189 --> 00:16:16.199 align:start position:0% let's say that you want to use this 00:16:16.199 --> 00:16:18.350 align:start position:0% let's say that you want to use this library<00:16:16.320> and<00:16:16.620> you<00:16:16.740> want<00:16:16.860> to<00:16:16.980> implement<00:16:17.339> it<00:16:17.519> so 00:16:18.350 --> 00:16:18.360 align:start position:0% library and you want to implement it so 00:16:18.360 --> 00:16:20.509 align:start position:0% library and you want to implement it so there<00:16:18.600> are<00:16:18.720> changes<00:16:19.079> to<00:16:19.199> your<00:16:19.380> code 00:16:20.509 --> 00:16:20.519 align:start position:0% there are changes to your code 00:16:20.519 --> 00:16:23.269 align:start position:0% there are changes to your code so<00:16:20.940> the<00:16:21.180> changes<00:16:21.540> the<00:16:21.720> following<00:16:22.079> since<00:16:22.860> let 00:16:23.269 --> 00:16:23.279 align:start position:0% so the changes the following since let 00:16:23.279 --> 00:16:26.210 align:start position:0% so the changes the following since let me<00:16:23.399> go<00:16:23.579> back<00:16:23.760> a<00:16:24.000> few<00:16:24.120> slides<00:16:24.899> perhaps<00:16:25.500> ah<00:16:26.040> it 00:16:26.210 --> 00:16:26.220 align:start position:0% me go back a few slides perhaps ah it 00:16:26.220 --> 00:16:29.329 align:start position:0% me go back a few slides perhaps ah it did<00:16:26.399> had<00:16:26.579> a<00:16:26.760> pointer<00:16:27.060> oh 00:16:29.329 --> 00:16:29.339 align:start position:0% did had a pointer oh 00:16:29.339 --> 00:16:32.389 align:start position:0% did had a pointer oh I'm<00:16:29.880> just<00:16:30.120> technology<00:16:30.600> at<00:16:31.079> first<00:16:31.320> okay 00:16:32.389 --> 00:16:32.399 align:start position:0% I'm just technology at first okay 00:16:32.399 --> 00:16:33.889 align:start position:0% I'm just technology at first okay anyways 00:16:33.889 --> 00:16:33.899 align:start position:0% anyways 00:16:33.899 --> 00:16:36.829 align:start position:0% anyways let's<00:16:34.440> see<00:16:34.620> this<00:16:35.040> is<00:16:35.100> the<00:16:35.399> the 00:16:36.829 --> 00:16:36.839 align:start position:0% let's see this is the the 00:16:36.839 --> 00:16:38.930 align:start position:0% let's see this is the the so<00:16:37.320> as<00:16:37.440> I<00:16:37.620> told<00:16:37.740> you<00:16:37.860> I'm<00:16:38.160> gonna<00:16:38.279> hash<00:16:38.579> with 00:16:38.930 --> 00:16:38.940 align:start position:0% so as I told you I'm gonna hash with 00:16:38.940 --> 00:16:40.730 align:start position:0% so as I told you I'm gonna hash with this<00:16:39.120> kind<00:16:39.240> of<00:16:39.360> registers<00:16:39.720> we<00:16:39.959> can<00:16:40.139> hash<00:16:40.320> ah 00:16:40.730 --> 00:16:40.740 align:start position:0% this kind of registers we can hash ah 00:16:40.740 --> 00:16:42.050 align:start position:0% this kind of registers we can hash ah now<00:16:40.860> I<00:16:40.980> have<00:16:41.100> a<00:16:41.220> point 00:16:42.050 --> 00:16:42.060 align:start position:0% now I have a point 00:16:42.060 --> 00:16:43.850 align:start position:0% now I have a point of<00:16:42.300> course<00:16:42.540> since<00:16:43.259> with<00:16:43.500> this<00:16:43.680> kind<00:16:43.800> of 00:16:43.850 --> 00:16:43.860 align:start position:0% of course since with this kind of 00:16:43.860 --> 00:16:45.889 align:start position:0% of course since with this kind of register<00:16:44.160> I<00:16:44.399> can<00:16:44.519> hash<00:16:44.759> four<00:16:45.240> blocks<00:16:45.600> at<00:16:45.779> a 00:16:45.889 --> 00:16:45.899 align:start position:0% register I can hash four blocks at a 00:16:45.899 --> 00:16:47.629 align:start position:0% register I can hash four blocks at a time<00:16:46.019> that<00:16:46.440> means<00:16:46.680> that<00:16:46.800> I<00:16:46.920> can<00:16:47.040> compute<00:16:47.399> this 00:16:47.629 --> 00:16:47.639 align:start position:0% time that means that I can compute this 00:16:47.639 --> 00:16:49.129 align:start position:0% time that means that I can compute this entire<00:16:48.060> layer 00:16:49.129 --> 00:16:49.139 align:start position:0% entire layer 00:16:49.139 --> 00:16:51.710 align:start position:0% entire layer in<00:16:49.620> one<00:16:49.800> pass<00:16:50.040> of<00:16:50.399> the<00:16:50.519> Russia<00:16:50.880> so<00:16:51.420> instead<00:16:51.660> of 00:16:51.710 --> 00:16:51.720 align:start position:0% in one pass of the Russia so instead of 00:16:51.720 --> 00:16:53.329 align:start position:0% in one pass of the Russia so instead of passing<00:16:52.019> on<00:16:52.139> a<00:16:52.259> loop<00:16:52.440> and<00:16:52.560> hash<00:16:52.740> this<00:16:52.920> to<00:16:53.100> Hash 00:16:53.329 --> 00:16:53.339 align:start position:0% passing on a loop and hash this to Hash 00:16:53.339 --> 00:16:55.069 align:start position:0% passing on a loop and hash this to Hash these<00:16:53.579> two<00:16:53.699> hashes<00:16:54.060> to<00:16:54.180> and<00:16:54.360> so<00:16:54.480> forth<00:16:54.660> what 00:16:55.069 --> 00:16:55.079 align:start position:0% these two hashes to and so forth what 00:16:55.079 --> 00:16:57.410 align:start position:0% these two hashes to and so forth what I'll<00:16:55.259> do<00:16:55.500> is<00:16:55.680> and<00:16:56.279> this<00:16:56.399> is<00:16:56.519> a<00:16:56.639> requirement<00:16:57.000> you 00:16:57.410 --> 00:16:57.420 align:start position:0% I'll do is and this is a requirement you 00:16:57.420 --> 00:17:00.350 align:start position:0% I'll do is and this is a requirement you cannot<00:16:57.720> use<00:16:58.019> Proto's<00:16:58.800> implementation<00:16:59.459> that 00:17:00.350 --> 00:17:00.360 align:start position:0% cannot use Proto's implementation that 00:17:00.360 --> 00:17:03.110 align:start position:0% cannot use Proto's implementation that has<00:17:00.540> pointers<00:17:01.079> everywhere<00:17:01.560> and<00:17:02.339> the<00:17:02.699> data<00:17:03.000> can 00:17:03.110 --> 00:17:03.120 align:start position:0% has pointers everywhere and the data can 00:17:03.120 --> 00:17:05.030 align:start position:0% has pointers everywhere and the data can be<00:17:03.300> anywhere<00:17:03.660> if<00:17:04.439> you<00:17:04.500> want<00:17:04.620> to<00:17:04.740> use<00:17:04.860> this 00:17:05.030 --> 00:17:05.040 align:start position:0% be anywhere if you want to use this 00:17:05.040 --> 00:17:06.890 align:start position:0% be anywhere if you want to use this library<00:17:05.220> then<00:17:05.579> you<00:17:05.819> need<00:17:05.939> to<00:17:06.059> have<00:17:06.299> at<00:17:06.780> the 00:17:06.890 --> 00:17:06.900 align:start position:0% library then you need to have at the 00:17:06.900 --> 00:17:08.390 align:start position:0% library then you need to have at the very<00:17:07.020> least<00:17:07.199> this<00:17:07.679> entire<00:17:08.040> layer 00:17:08.390 --> 00:17:08.400 align:start position:0% very least this entire layer 00:17:08.400 --> 00:17:11.150 align:start position:0% very least this entire layer consecutively<00:17:09.360> memory<00:17:09.839> and<00:17:10.620> this<00:17:10.860> entire 00:17:11.150 --> 00:17:11.160 align:start position:0% consecutively memory and this entire 00:17:11.160 --> 00:17:13.370 align:start position:0% consecutively memory and this entire layer<00:17:11.579> consecutively<00:17:12.360> in<00:17:12.540> memory<00:17:12.900> and<00:17:13.199> so 00:17:13.370 --> 00:17:13.380 align:start position:0% layer consecutively in memory and so 00:17:13.380 --> 00:17:14.569 align:start position:0% layer consecutively in memory and so forth 00:17:14.569 --> 00:17:14.579 align:start position:0% forth 00:17:14.579 --> 00:17:16.610 align:start position:0% forth uh<00:17:15.179> vitalik's<00:17:15.780> flat<00:17:15.959> array<00:17:16.199> would<00:17:16.439> work 00:17:16.610 --> 00:17:16.620 align:start position:0% uh vitalik's flat array would work 00:17:16.620 --> 00:17:19.189 align:start position:0% uh vitalik's flat array would work fantastic<00:17:17.040> for<00:17:17.640> this<00:17:17.760> for<00:17:18.059> this<00:17:18.179> thing<00:17:18.419> you 00:17:19.189 --> 00:17:19.199 align:start position:0% fantastic for this for this thing you 00:17:19.199 --> 00:17:20.929 align:start position:0% fantastic for this for this thing you would<00:17:19.380> I<00:17:19.620> would<00:17:19.740> if<00:17:20.040> I<00:17:20.220> were<00:17:20.339> to<00:17:20.459> implement 00:17:20.929 --> 00:17:20.939 align:start position:0% would I would if I were to implement 00:17:20.939 --> 00:17:22.669 align:start position:0% would I would if I were to implement this<00:17:21.179> I<00:17:21.480> would<00:17:21.660> just<00:17:21.780> put<00:17:21.900> everything<00:17:22.140> on<00:17:22.439> one 00:17:22.669 --> 00:17:22.679 align:start position:0% this I would just put everything on one 00:17:22.679 --> 00:17:25.250 align:start position:0% this I would just put everything on one array<00:17:23.040> the<00:17:23.699> point<00:17:23.880> is<00:17:24.000> that<00:17:24.419> what<00:17:25.140> you're 00:17:25.250 --> 00:17:25.260 align:start position:0% array the point is that what you're 00:17:25.260 --> 00:17:27.650 align:start position:0% array the point is that what you're going<00:17:25.439> to<00:17:25.500> pass<00:17:25.679> the<00:17:25.860> hasher<00:17:26.339> is<00:17:26.579> a<00:17:27.179> pointer<00:17:27.480> to 00:17:27.650 --> 00:17:27.660 align:start position:0% going to pass the hasher is a pointer to 00:17:27.660 --> 00:17:29.570 align:start position:0% going to pass the hasher is a pointer to this<00:17:27.900> or<00:17:28.319> this<00:17:28.559> whole<00:17:28.740> slides<00:17:29.220> whatever 00:17:29.570 --> 00:17:29.580 align:start position:0% this or this whole slides whatever 00:17:29.580 --> 00:17:32.450 align:start position:0% this or this whole slides whatever equivalent<00:17:30.440> signature<00:17:31.440> for<00:17:31.620> this<00:17:31.799> is<00:17:31.980> and<00:17:32.280> the 00:17:32.450 --> 00:17:32.460 align:start position:0% equivalent signature for this is and the 00:17:32.460 --> 00:17:34.130 align:start position:0% equivalent signature for this is and the hasher<00:17:32.820> is<00:17:32.880> going<00:17:33.059> to<00:17:33.120> give<00:17:33.240> you<00:17:33.360> back<00:17:33.600> all<00:17:34.020> of 00:17:34.130 --> 00:17:34.140 align:start position:0% hasher is going to give you back all of 00:17:34.140 --> 00:17:35.810 align:start position:0% hasher is going to give you back all of this<00:17:34.260> at<00:17:34.440> the<00:17:34.559> same<00:17:34.679> time<00:17:34.799> so<00:17:35.520> this<00:17:35.700> is 00:17:35.810 --> 00:17:35.820 align:start position:0% this at the same time so this is 00:17:35.820 --> 00:17:37.549 align:start position:0% this at the same time so this is something<00:17:36.000> that<00:17:36.299> you<00:17:36.419> need<00:17:36.539> to<00:17:36.660> change<00:17:36.900> the 00:17:37.549 --> 00:17:37.559 align:start position:0% something that you need to change the 00:17:37.559 --> 00:17:40.010 align:start position:0% something that you need to change the hasher<00:17:38.039> signature<00:17:38.460> for<00:17:38.580> this<00:17:38.880> has<00:17:39.419> this<00:17:39.840> form 00:17:40.010 --> 00:17:40.020 align:start position:0% hasher signature for this has this form 00:17:40.020 --> 00:17:42.950 align:start position:0% hasher signature for this has this form so<00:17:40.500> in<00:17:40.679> go<00:17:40.919> this<00:17:41.340> takes<00:17:41.580> an<00:17:42.120> arbitrary<00:17:42.600> by<00:17:42.780> its 00:17:42.950 --> 00:17:42.960 align:start position:0% so in go this takes an arbitrary by its 00:17:42.960 --> 00:17:44.630 align:start position:0% so in go this takes an arbitrary by its life<00:17:43.140> so<00:17:43.380> this<00:17:43.559> is<00:17:43.679> the<00:17:43.919> layer<00:17:44.220> that<00:17:44.400> you<00:17:44.520> want 00:17:44.630 --> 00:17:44.640 align:start position:0% life so this is the layer that you want 00:17:44.640 --> 00:17:47.990 align:start position:0% life so this is the layer that you want to<00:17:44.700> hash<00:17:44.940> and<00:17:45.419> it<00:17:45.539> gives<00:17:45.720> you<00:17:45.900> back<00:17:46.140> and<00:17:47.280> a 00:17:47.990 --> 00:17:48.000 align:start position:0% to hash and it gives you back and a 00:17:48.000 --> 00:17:50.810 align:start position:0% to hash and it gives you back and a slice<00:17:48.419> of<00:17:48.780> hashes<00:17:49.559> of<00:17:49.860> 32<00:17:50.220> bites<00:17:50.640> these 00:17:50.810 --> 00:17:50.820 align:start position:0% slice of hashes of 32 bites these 00:17:50.820 --> 00:17:52.730 align:start position:0% slice of hashes of 32 bites these hatches<00:17:51.179> all<00:17:51.299> of<00:17:51.419> them<00:17:51.480> at<00:17:51.600> the<00:17:51.720> same<00:17:51.840> time 00:17:52.730 --> 00:17:52.740 align:start position:0% hatches all of them at the same time 00:17:52.740 --> 00:17:55.430 align:start position:0% hatches all of them at the same time uh<00:17:53.640> I<00:17:54.240> think<00:17:54.360> I<00:17:54.660> might<00:17:54.840> have<00:17:54.960> gotten<00:17:55.260> this 00:17:55.430 --> 00:17:55.440 align:start position:0% uh I think I might have gotten this 00:17:55.440 --> 00:17:56.990 align:start position:0% uh I think I might have gotten this correctly<00:17:55.860> in<00:17:55.980> Rust<00:17:56.340> after<00:17:56.580> several 00:17:56.990 --> 00:17:57.000 align:start position:0% correctly in Rust after several 00:17:57.000 --> 00:18:00.470 align:start position:0% correctly in Rust after several iterations<00:17:57.419> uh<00:17:58.320> this<00:17:58.559> would<00:17:58.679> be<00:17:58.799> in<00:17:58.919> Python 00:18:00.470 --> 00:18:00.480 align:start position:0% iterations uh this would be in Python 00:18:00.480 --> 00:18:03.049 align:start position:0% iterations uh this would be in Python uh<00:18:01.260> we<00:18:01.679> have<00:18:01.860> a<00:18:02.039> library 00:18:03.049 --> 00:18:03.059 align:start position:0% uh we have a library 00:18:03.059 --> 00:18:05.750 align:start position:0% uh we have a library uh<00:18:03.780> we<00:18:04.020> have<00:18:04.140> two<00:18:04.320> libraries<00:18:04.679> we<00:18:04.980> have<00:18:05.100> one<00:18:05.400> in 00:18:05.750 --> 00:18:05.760 align:start position:0% uh we have two libraries we have one in 00:18:05.760 --> 00:18:07.850 align:start position:0% uh we have two libraries we have one in go<00:18:05.880> assembly<00:18:06.360> and<00:18:06.480> we<00:18:06.600> have<00:18:06.720> one<00:18:07.020> in<00:18:07.500> user 00:18:07.850 --> 00:18:07.860 align:start position:0% go assembly and we have one in user 00:18:07.860 --> 00:18:09.830 align:start position:0% go assembly and we have one in user assembly<00:18:08.280> with<00:18:08.580> C<00:18:08.760> bindings 00:18:09.830 --> 00:18:09.840 align:start position:0% assembly with C bindings 00:18:09.840 --> 00:18:11.990 align:start position:0% assembly with C bindings if<00:18:10.440> you<00:18:10.679> are<00:18:10.799> going<00:18:10.980> to<00:18:11.100> use<00:18:11.280> that<00:18:11.520> let<00:18:11.700> me<00:18:11.880> know 00:18:11.990 --> 00:18:12.000 align:start position:0% if you are going to use that let me know 00:18:12.000 --> 00:18:13.070 align:start position:0% if you are going to use that let me know because 00:18:13.070 --> 00:18:13.080 align:start position:0% because 00:18:13.080 --> 00:18:15.830 align:start position:0% because crypto<00:18:13.799> extensions<00:18:14.280> on<00:18:14.820> arm<00:18:15.120> is<00:18:15.419> still<00:18:15.660> not 00:18:15.830 --> 00:18:15.840 align:start position:0% crypto extensions on arm is still not 00:18:15.840 --> 00:18:17.330 align:start position:0% crypto extensions on arm is still not implemented<00:18:16.380> it's<00:18:16.620> going<00:18:16.799> to<00:18:16.919> take<00:18:17.039> like<00:18:17.160> 10 00:18:17.330 --> 00:18:17.340 align:start position:0% implemented it's going to take like 10 00:18:17.340 --> 00:18:18.049 align:start position:0% implemented it's going to take like 10 minutes 00:18:18.049 --> 00:18:18.059 align:start position:0% minutes 00:18:18.059 --> 00:18:20.450 align:start position:0% minutes to<00:18:18.539> add<00:18:18.780> this<00:18:19.020> but<00:18:19.620> this<00:18:19.799> is<00:18:19.860> the<00:18:20.039> signature 00:18:20.450 --> 00:18:20.460 align:start position:0% to add this but this is the signature 00:18:20.460 --> 00:18:22.490 align:start position:0% to add this but this is the signature that<00:18:20.880> the<00:18:21.539> library<00:18:21.660> is<00:18:21.900> going<00:18:22.080> to<00:18:22.200> use<00:18:22.320> if 00:18:22.490 --> 00:18:22.500 align:start position:0% that the library is going to use if 00:18:22.500 --> 00:18:25.730 align:start position:0% that the library is going to use if you're<00:18:22.679> using<00:18:22.980> a<00:18:23.160> c<00:18:23.220> bindings 00:18:25.730 --> 00:18:25.740 align:start position:0% you're using a c bindings 00:18:25.740 --> 00:18:27.890 align:start position:0% you're using a c bindings all<00:18:26.160> right<00:18:26.280> so<00:18:26.700> that's<00:18:27.000> all<00:18:27.360> I<00:18:27.539> want<00:18:27.600> to<00:18:27.720> tell 00:18:27.890 --> 00:18:27.900 align:start position:0% all right so that's all I want to tell 00:18:27.900 --> 00:18:31.789 align:start position:0% all right so that's all I want to tell you<00:18:28.080> and<00:18:28.679> that's<00:18:28.799> that's<00:18:29.100> it 00:18:31.789 --> 00:18:31.799 align:start position:0% 00:18:31.799 --> 00:18:33.890 align:start position:0% okay<00:18:32.520> sure<00:18:32.760> I'll<00:18:33.059> just<00:18:33.179> repeat<00:18:33.360> that<00:18:33.600> for<00:18:33.720> the 00:18:33.890 --> 00:18:33.900 align:start position:0% okay sure I'll just repeat that for the 00:18:33.900 --> 00:18:35.210 align:start position:0% okay sure I'll just repeat that for the sake<00:18:34.020> of<00:18:34.140> the<00:18:34.260> recording<00:18:34.500> I<00:18:34.620> was<00:18:34.799> just<00:18:34.860> asking 00:18:35.210 --> 00:18:35.220 align:start position:0% sake of the recording I was just asking 00:18:35.220 --> 00:18:36.230 align:start position:0% sake of the recording I was just asking if 00:18:36.230 --> 00:18:36.240 align:start position:0% if 00:18:36.240 --> 00:18:38.990 align:start position:0% if um<00:18:36.320> uh<00:18:37.320> the<00:18:37.679> the<00:18:37.740> ability<00:18:38.340> to<00:18:38.640> process 00:18:38.990 --> 00:18:39.000 align:start position:0% um uh the the ability to process 00:18:39.000 --> 00:18:40.789 align:start position:0% um uh the the ability to process multiple<00:18:39.660> different<00:18:39.840> hashes<00:18:40.260> in<00:18:40.440> parallel 00:18:40.789 --> 00:18:40.799 align:start position:0% multiple different hashes in parallel 00:18:40.799 --> 00:18:42.230 align:start position:0% multiple different hashes in parallel was<00:18:41.039> already<00:18:41.220> implemented<00:18:41.700> in<00:18:41.880> libraries 00:18:42.230 --> 00:18:42.240 align:start position:0% was already implemented in libraries 00:18:42.240 --> 00:18:45.710 align:start position:0% was already implemented in libraries that<00:18:42.600> was<00:18:42.780> the<00:18:43.440> question<00:18:44.100> to<00:18:44.400> the<00:18:44.640> answer 00:18:45.710 --> 00:18:45.720 align:start position:0% that was the question to the answer 00:18:45.720 --> 00:18:48.350 align:start position:0% that was the question to the answer um<00:18:45.840> so<00:18:46.500> so<00:18:47.220> that<00:18:47.520> so<00:18:47.760> there's<00:18:47.880> two<00:18:48.059> things<00:18:48.179> that 00:18:48.350 --> 00:18:48.360 align:start position:0% um so so that so there's two things that 00:18:48.360 --> 00:18:50.090 align:start position:0% um so so that so there's two things that you<00:18:48.419> have<00:18:48.539> here<00:18:48.660> right<00:18:48.840> one<00:18:49.080> is<00:18:49.320> to<00:18:49.679> like 00:18:50.090 --> 00:18:50.100 align:start position:0% you have here right one is to like 00:18:50.100 --> 00:18:52.549 align:start position:0% you have here right one is to like pre-compute<00:18:50.820> the<00:18:51.059> padding<00:18:51.539> block<00:18:51.720> and<00:18:52.440> then 00:18:52.549 --> 00:18:52.559 align:start position:0% pre-compute the padding block and then 00:18:52.559 --> 00:18:56.029 align:start position:0% pre-compute the padding block and then the<00:18:52.740> second<00:18:52.860> is<00:18:53.220> the<00:18:53.400> ability<00:18:53.640> to<00:18:54.299> to<00:18:55.039> in 00:18:56.029 --> 00:18:56.039 align:start position:0% the second is the ability to to in 00:18:56.039 --> 00:18:58.850 align:start position:0% the second is the ability to to in parallel<00:18:56.600> process<00:18:57.600> multiple<00:18:58.260> different<00:18:58.500> hash 00:18:58.850 --> 00:18:58.860 align:start position:0% parallel process multiple different hash 00:18:58.860 --> 00:19:00.650 align:start position:0% parallel process multiple different hash values<00:18:59.280> and<00:18:59.520> then<00:18:59.640> produce<00:19:00.120> multiple 00:19:00.650 --> 00:19:00.660 align:start position:0% values and then produce multiple 00:19:00.660 --> 00:19:02.690 align:start position:0% values and then produce multiple different<00:19:00.960> hash<00:19:01.320> values<00:19:01.679> right<00:19:01.799> correct<00:19:02.160> and 00:19:02.690 --> 00:19:02.700 align:start position:0% different hash values right correct and 00:19:02.700 --> 00:19:03.730 align:start position:0% different hash values right correct and so 00:19:03.730 --> 00:19:03.740 align:start position:0% so 00:19:03.740 --> 00:19:06.049 align:start position:0% so there<00:19:04.740> are<00:19:04.860> already<00:19:04.980> implementations<00:19:05.700> that 00:19:06.049 --> 00:19:06.059 align:start position:0% there are already implementations that 00:19:06.059 --> 00:19:07.610 align:start position:0% there are already implementations that do<00:19:06.240> multiple<00:19:06.539> different<00:19:06.780> hash<00:19:07.080> values<00:19:07.440> at 00:19:07.610 --> 00:19:07.620 align:start position:0% do multiple different hash values at 00:19:07.620 --> 00:19:09.950 align:start position:0% do multiple different hash values at once<00:19:07.860> and<00:19:08.100> you<00:19:08.280> just<00:19:08.400> modified<00:19:08.760> them<00:19:09.179> to<00:19:09.780> deal 00:19:09.950 --> 00:19:09.960 align:start position:0% once and you just modified them to deal 00:19:09.960 --> 00:19:11.630 align:start position:0% once and you just modified them to deal with<00:19:10.140> the<00:19:10.320> padding<00:19:10.620> block<00:19:10.740> is<00:19:10.980> that<00:19:11.160> right<00:19:11.280> yes 00:19:11.630 --> 00:19:11.640 align:start position:0% with the padding block is that right yes 00:19:11.640 --> 00:19:13.490 align:start position:0% with the padding block is that right yes and<00:19:12.299> yes<00:19:12.660> so<00:19:12.840> so<00:19:13.080> there<00:19:13.260> are<00:19:13.380> two 00:19:13.490 --> 00:19:13.500 align:start position:0% and yes so so there are two 00:19:13.500 --> 00:19:16.070 align:start position:0% and yes so so there are two modifications<00:19:13.860> here<00:19:14.220> one<00:19:14.700> is<00:19:14.940> uh<00:19:15.600> the<00:19:15.900> thing 00:19:16.070 --> 00:19:16.080 align:start position:0% modifications here one is uh the thing 00:19:16.080 --> 00:19:17.390 align:start position:0% modifications here one is uh the thing that<00:19:16.320> you're<00:19:16.500> going<00:19:16.679> to<00:19:16.799> use<00:19:16.860> the<00:19:17.039> padding 00:19:17.390 --> 00:19:17.400 align:start position:0% that you're going to use the padding 00:19:17.400 --> 00:19:20.090 align:start position:0% that you're going to use the padding block<00:19:17.520> uh<00:19:18.480> the<00:19:19.020> padding<00:19:19.500> block<00:19:19.620> has<00:19:19.860> this 00:19:20.090 --> 00:19:20.100 align:start position:0% block uh the padding block has this 00:19:20.100 --> 00:19:21.650 align:start position:0% block uh the padding block has this constants<00:19:20.460> hardcoded<00:19:21.059> and<00:19:21.419> then<00:19:21.539> there's 00:19:21.650 --> 00:19:21.660 align:start position:0% constants hardcoded and then there's 00:19:21.660 --> 00:19:23.630 align:start position:0% constants hardcoded and then there's other<00:19:21.900> modification<00:19:22.440> which<00:19:22.799> is<00:19:22.980> the<00:19:23.520> fact 00:19:23.630 --> 00:19:23.640 align:start position:0% other modification which is the fact 00:19:23.640 --> 00:19:27.770 align:start position:0% other modification which is the fact that<00:19:23.940> you're<00:19:24.480> expected<00:19:25.080> to<00:19:25.320> get<00:19:25.640> a<00:19:26.640> list<00:19:26.940> of<00:19:27.240> 64 00:19:27.770 --> 00:19:27.780 align:start position:0% that you're expected to get a list of 64 00:19:27.780 --> 00:19:30.470 align:start position:0% that you're expected to get a list of 64 bytes<00:19:28.260> chunks<00:19:28.860> and<00:19:29.460> then<00:19:29.580> you<00:19:29.820> pipeline<00:19:30.240> this 00:19:30.470 --> 00:19:30.480 align:start position:0% bytes chunks and then you pipeline this 00:19:30.480 --> 00:19:33.230 align:start position:0% bytes chunks and then you pipeline this so<00:19:30.780> what<00:19:31.140> this<00:19:31.500> Library<00:19:31.620> does<00:19:32.039> is<00:19:32.340> it<00:19:32.820> grabs 00:19:33.230 --> 00:19:33.240 align:start position:0% so what this Library does is it grabs 00:19:33.240 --> 00:19:35.510 align:start position:0% so what this Library does is it grabs all<00:19:33.960> of<00:19:34.080> the<00:19:34.200> blocks<00:19:34.500> consecutively<00:19:35.100> memory 00:19:35.510 --> 00:19:35.520 align:start position:0% all of the blocks consecutively memory 00:19:35.520 --> 00:19:39.890 align:start position:0% all of the blocks consecutively memory it<00:19:36.740> gets<00:19:37.740> a<00:19:37.860> matrix<00:19:38.280> on<00:19:38.940> the<00:19:39.120> vector<00:19:39.480> on<00:19:39.780> the 00:19:39.890 --> 00:19:39.900 align:start position:0% it gets a matrix on the vector on the 00:19:39.900 --> 00:19:41.750 align:start position:0% it gets a matrix on the vector on the vector<00:19:40.140> registers<00:19:40.620> it<00:19:41.039> transposes<00:19:41.580> this 00:19:41.750 --> 00:19:41.760 align:start position:0% vector registers it transposes this 00:19:41.760 --> 00:19:43.909 align:start position:0% vector registers it transposes this Matrix<00:19:42.179> and<00:19:42.660> now<00:19:42.900> you<00:19:43.080> have<00:19:43.200> on<00:19:43.500> all<00:19:43.679> of<00:19:43.799> your 00:19:43.909 --> 00:19:43.919 align:start position:0% Matrix and now you have on all of your 00:19:43.919 --> 00:19:46.250 align:start position:0% Matrix and now you have on all of your registers<00:19:44.400> the<00:19:45.000> different<00:19:45.179> the<00:19:45.780> different 00:19:46.250 --> 00:19:46.260 align:start position:0% registers the different the different 00:19:46.260 --> 00:19:47.390 align:start position:0% registers the different the different messages 00:19:47.390 --> 00:19:47.400 align:start position:0% messages 00:19:47.400 --> 00:19:51.049 align:start position:0% messages so<00:19:47.880> then<00:19:48.059> you<00:19:48.360> can<00:19:48.480> use<00:19:48.780> Intel's<00:19:49.620> Machinery<00:19:50.100> to 00:19:51.049 --> 00:19:51.059 align:start position:0% so then you can use Intel's Machinery to 00:19:51.059 --> 00:19:53.690 align:start position:0% so then you can use Intel's Machinery to Hash<00:19:51.539> those<00:19:51.780> messages<00:19:52.020> in<00:19:52.380> parallel<00:19:52.799> you 00:19:53.690 --> 00:19:53.700 align:start position:0% Hash those messages in parallel you 00:19:53.700 --> 00:19:56.570 align:start position:0% Hash those messages in parallel you print<00:19:54.059> the<00:19:54.660> you<00:19:54.900> output<00:19:55.380> these<00:19:56.100> hashes<00:19:56.400> and 00:19:56.570 --> 00:19:56.580 align:start position:0% print the you output these hashes and 00:19:56.580 --> 00:19:58.850 align:start position:0% print the you output these hashes and then<00:19:56.700> you<00:19:56.880> just<00:19:57.120> Loop<00:19:57.720> it<00:19:57.840> back<00:19:58.020> so<00:19:58.679> that's 00:19:58.850 --> 00:19:58.860 align:start position:0% then you just Loop it back so that's 00:19:58.860 --> 00:20:00.950 align:start position:0% then you just Loop it back so that's cool<00:19:59.160> so<00:19:59.460> you<00:19:59.580> you<00:19:59.700> have<00:19:59.940> so<00:20:00.179> the<00:20:00.299> result<00:20:00.660> is 00:20:00.950 --> 00:20:00.960 align:start position:0% cool so you you have so the result is 00:20:00.960 --> 00:20:03.110 align:start position:0% cool so you you have so the result is basically<00:20:01.260> like<00:20:01.919> so<00:20:02.220> it's<00:20:02.400> inco<00:20:02.700> assembly 00:20:03.110 --> 00:20:03.120 align:start position:0% basically like so it's inco assembly 00:20:03.120 --> 00:20:05.390 align:start position:0% basically like so it's inco assembly what<00:20:03.299> you<00:20:03.480> have<00:20:03.600> right<00:20:03.840> sorry<00:20:04.440> your<00:20:04.980> your 00:20:05.390 --> 00:20:05.400 align:start position:0% what you have right sorry your your 00:20:05.400 --> 00:20:08.210 align:start position:0% what you have right sorry your your implementation<00:20:06.120> is<00:20:06.360> in<00:20:06.539> go<00:20:06.660> assembly<00:20:07.140> I<00:20:07.980> yeah 00:20:08.210 --> 00:20:08.220 align:start position:0% implementation is in go assembly I yeah 00:20:08.220 --> 00:20:10.130 align:start position:0% implementation is in go assembly I yeah so<00:20:08.460> the<00:20:09.059> original<00:20:09.179> implementation<00:20:09.780> was<00:20:10.020> just 00:20:10.130 --> 00:20:10.140 align:start position:0% so the original implementation was just 00:20:10.140 --> 00:20:12.350 align:start position:0% so the original implementation was just purely<00:20:10.500> assembly<00:20:10.919> this<00:20:11.160> is<00:20:11.280> there<00:20:11.760> it<00:20:12.059> has<00:20:12.179> C 00:20:12.350 --> 00:20:12.360 align:start position:0% purely assembly this is there it has C 00:20:12.360 --> 00:20:15.770 align:start position:0% purely assembly this is there it has C bindings<00:20:12.840> but<00:20:13.500> it<00:20:13.740> the<00:20:14.340> C<00:20:14.640> go<00:20:14.940> overhead<00:20:15.480> is 00:20:15.770 --> 00:20:15.780 align:start position:0% bindings but it the C go overhead is 00:20:15.780 --> 00:20:17.750 align:start position:0% bindings but it the C go overhead is horrible<00:20:16.200> so<00:20:16.559> it<00:20:16.679> ended<00:20:17.039> up<00:20:17.100> being<00:20:17.280> slower 00:20:17.750 --> 00:20:17.760 align:start position:0% horrible so it ended up being slower 00:20:17.760 --> 00:20:20.390 align:start position:0% horrible so it ended up being slower than<00:20:18.240> using<00:20:19.020> the<00:20:19.380> go<00:20:19.500> implementation<00:20:20.100> the 00:20:20.390 --> 00:20:20.400 align:start position:0% than using the go implementation the 00:20:20.400 --> 00:20:22.610 align:start position:0% than using the go implementation the standard<00:20:20.940> library<00:20:21.120> and<00:20:21.419> go<00:20:21.539> so<00:20:22.080> we<00:20:22.200> needed<00:20:22.440> to 00:20:22.610 --> 00:20:22.620 align:start position:0% standard library and go so we needed to 00:20:22.620 --> 00:20:24.529 align:start position:0% standard library and go so we needed to write<00:20:22.980> a<00:20:23.220> Google<00:20:23.460> assembly<00:20:23.940> library<00:20:24.120> to<00:20:24.360> use 00:20:24.529 --> 00:20:24.539 align:start position:0% write a Google assembly library to use 00:20:24.539 --> 00:20:27.049 align:start position:0% write a Google assembly library to use ourselves<00:20:24.900> yeah<00:20:25.740> okay<00:20:25.860> cool<00:20:26.220> well<00:20:26.580> done<00:20:26.820> very 00:20:27.049 --> 00:20:27.059 align:start position:0% ourselves yeah okay cool well done very 00:20:27.059 --> 00:20:30.590 align:start position:0% ourselves yeah okay cool well done very good<00:20:27.240> very<00:20:27.539> impressive<00:20:28.320> thanks 00:20:30.590 --> 00:20:30.600 align:start position:0% good very impressive thanks 00:20:30.600 --> 00:20:33.890 align:start position:0% good very impressive thanks uh<00:20:31.380> so<00:20:31.620> I<00:20:32.039> was<00:20:32.220> curious<00:20:32.700> you've<00:20:33.360> shown<00:20:33.720> that 00:20:33.890 --> 00:20:33.900 align:start position:0% uh so I was curious you've shown that 00:20:33.900 --> 00:20:35.450 align:start position:0% uh so I was curious you've shown that there<00:20:34.080> are<00:20:34.200> always<00:20:34.320> optimizations<00:20:34.919> you<00:20:35.280> have 00:20:35.450 --> 00:20:35.460 align:start position:0% there are always optimizations you have 00:20:35.460 --> 00:20:37.430 align:start position:0% there are always optimizations you have to<00:20:35.580> do<00:20:35.760> to<00:20:35.940> get<00:20:36.120> around<00:20:36.240> the<00:20:36.720> general<00:20:36.900> purpose 00:20:37.430 --> 00:20:37.440 align:start position:0% to do to get around the general purpose 00:20:37.440 --> 00:20:40.190 align:start position:0% to do to get around the general purpose nature<00:20:37.919> of<00:20:38.880> these<00:20:39.299> General<00:20:39.480> perfect<00:20:39.780> hash 00:20:40.190 --> 00:20:40.200 align:start position:0% nature of these General perfect hash 00:20:40.200 --> 00:20:42.230 align:start position:0% nature of these General perfect hash functions<00:20:40.679> do<00:20:41.580> you<00:20:41.760> think<00:20:41.880> it<00:20:42.000> would<00:20:42.179> be 00:20:42.230 --> 00:20:42.240 align:start position:0% functions do you think it would be 00:20:42.240 --> 00:20:44.210 align:start position:0% functions do you think it would be possible<00:20:42.539> to<00:20:42.840> design<00:20:42.960> a<00:20:43.320> hash<00:20:43.500> function<00:20:43.860> that 00:20:44.210 --> 00:20:44.220 align:start position:0% possible to design a hash function that 00:20:44.220 --> 00:20:45.890 align:start position:0% possible to design a hash function that was<00:20:44.340> like<00:20:44.460> a<00:20:44.640> special<00:20:44.820> hash<00:20:45.120> function<00:20:45.419> only 00:20:45.890 --> 00:20:45.900 align:start position:0% was like a special hash function only 00:20:45.900 --> 00:20:48.350 align:start position:0% was like a special hash function only for<00:20:46.200> like<00:20:46.440> oh 00:20:48.350 --> 00:20:48.360 align:start position:0% for like oh 00:20:48.360 --> 00:20:50.570 align:start position:0% for like oh oh<00:20:48.419> that's<00:20:49.080> a<00:20:49.260> good<00:20:49.380> question<00:20:49.679> I'm<00:20:50.280> not<00:20:50.400> sure 00:20:50.570 --> 00:20:50.580 align:start position:0% oh that's a good question I'm not sure 00:20:50.580 --> 00:20:53.330 align:start position:0% oh that's a good question I'm not sure if<00:20:50.700> you<00:20:50.820> can<00:20:51.360> do<00:20:51.480> better<00:20:51.720> than<00:20:52.320> the<00:20:52.799> current 00:20:53.330 --> 00:20:53.340 align:start position:0% if you can do better than the current 00:20:53.340 --> 00:20:55.430 align:start position:0% if you can do better than the current implementations<00:20:54.000> like<00:20:54.360> uh<00:20:54.660> you<00:20:54.960> can<00:20:55.020> take<00:20:55.140> the 00:20:55.430 --> 00:20:55.440 align:start position:0% implementations like uh you can take the 00:20:55.440 --> 00:20:57.890 align:start position:0% implementations like uh you can take the sponge<00:20:55.740> type<00:20:56.280> implementations<00:20:57.000> shot<00:20:57.419> 3<00:20:57.660> and 00:20:57.890 --> 00:20:57.900 align:start position:0% sponge type implementations shot 3 and 00:20:57.900 --> 00:21:00.169 align:start position:0% sponge type implementations shot 3 and Company<00:20:58.080> and<00:20:59.039> you<00:20:59.220> can<00:20:59.340> try<00:20:59.460> to<00:20:59.640> adapt<00:21:00.000> this 00:21:00.169 --> 00:21:00.179 align:start position:0% Company and you can try to adapt this 00:21:00.179 --> 00:21:01.669 align:start position:0% Company and you can try to adapt this for<00:21:00.419> this 00:21:01.669 --> 00:21:01.679 align:start position:0% for this 00:21:01.679 --> 00:21:04.070 align:start position:0% for this I<00:21:02.220> don't<00:21:02.340> know<00:21:02.460> I<00:21:02.760> I<00:21:02.940> think<00:21:03.240> yeah<00:21:03.780> your 00:21:04.070 --> 00:21:04.080 align:start position:0% I don't know I I think yeah your 00:21:04.080 --> 00:21:05.690 align:start position:0% I don't know I I think yeah your question<00:21:04.200> is<00:21:04.380> completely<00:21:04.740> open<00:21:04.919> it's<00:21:05.280> it's<00:21:05.520> a 00:21:05.690 --> 00:21:05.700 align:start position:0% question is completely open it's it's a 00:21:05.700 --> 00:21:07.190 align:start position:0% question is completely open it's it's a very<00:21:05.820> good<00:21:05.940> question<00:21:06.120> I<00:21:06.360> think<00:21:06.660> we<00:21:06.840> should 00:21:07.190 --> 00:21:07.200 align:start position:0% very good question I think we should 00:21:07.200 --> 00:21:08.690 align:start position:0% very good question I think we should think<00:21:07.320> about<00:21:07.440> this 00:21:08.690 --> 00:21:08.700 align:start position:0% think about this 00:21:08.700 --> 00:21:11.210 align:start position:0% think about this so<00:21:09.120> if<00:21:09.299> I<00:21:09.539> understood<00:21:09.840> your<00:21:10.080> question<00:21:10.200> as<00:21:10.500> can 00:21:11.210 --> 00:21:11.220 align:start position:0% so if I understood your question as can 00:21:11.220 --> 00:21:13.190 align:start position:0% so if I understood your question as can there<00:21:11.400> be<00:21:11.580> a<00:21:11.820> method<00:21:12.179> that<00:21:12.480> is<00:21:12.660> designed<00:21:13.020> for 00:21:13.190 --> 00:21:13.200 align:start position:0% there be a method that is designed for 00:21:13.200 --> 00:21:15.590 align:start position:0% there be a method that is designed for Merkle<00:21:13.620> trades<00:21:13.980> uh<00:21:14.940> instead<00:21:15.240> of<00:21:15.299> like<00:21:15.480> a 00:21:15.590 --> 00:21:15.600 align:start position:0% Merkle trades uh instead of like a 00:21:15.600 --> 00:21:17.630 align:start position:0% Merkle trades uh instead of like a generic<00:21:15.960> one<00:21:16.140> and<00:21:16.620> your<00:21:16.799> implementation<00:21:17.280> does 00:21:17.630 --> 00:21:17.640 align:start position:0% generic one and your implementation does 00:21:17.640 --> 00:21:19.610 align:start position:0% generic one and your implementation does it<00:21:18.059> so<00:21:18.539> say<00:21:18.840> there's<00:21:19.140> some<00:21:19.320> really<00:21:19.440> big 00:21:19.610 --> 00:21:19.620 align:start position:0% it so say there's some really big 00:21:19.620 --> 00:21:21.289 align:start position:0% it so say there's some really big miracle<00:21:19.980> trades<00:21:20.340> in<00:21:20.460> in<00:21:20.640> the<00:21:20.820> beacon<00:21:21.120> state 00:21:21.289 --> 00:21:21.299 align:start position:0% miracle trades in in the beacon state 00:21:21.299 --> 00:21:23.570 align:start position:0% miracle trades in in the beacon state right<00:21:21.539> yeah<00:21:22.320> um<00:21:22.440> is<00:21:22.919> it<00:21:23.039> the<00:21:23.160> job<00:21:23.280> of<00:21:23.460> the 00:21:23.570 --> 00:21:23.580 align:start position:0% right yeah um is it the job of the 00:21:23.580 --> 00:21:25.130 align:start position:0% right yeah um is it the job of the implementation<00:21:24.059> to<00:21:24.360> kind<00:21:24.539> of<00:21:24.660> split<00:21:24.960> that 00:21:25.130 --> 00:21:25.140 align:start position:0% implementation to kind of split that 00:21:25.140 --> 00:21:27.230 align:start position:0% implementation to kind of split that Merkle<00:21:25.559> tree<00:21:25.679> up<00:21:25.919> into<00:21:26.160> smaller<00:21:26.640> subtrees 00:21:27.230 --> 00:21:27.240 align:start position:0% Merkle tree up into smaller subtrees 00:21:27.240 --> 00:21:30.950 align:start position:0% Merkle tree up into smaller subtrees that<00:21:27.720> fit<00:21:28.260> the<00:21:28.620> size<00:21:28.740> of<00:21:28.919> your<00:21:29.100> CPU<00:21:29.640> no<00:21:30.480> no<00:21:30.659> so 00:21:30.950 --> 00:21:30.960 align:start position:0% that fit the size of your CPU no no so 00:21:30.960 --> 00:21:32.990 align:start position:0% that fit the size of your CPU no no so the<00:21:31.140> the<00:21:31.260> so<00:21:31.860> you're<00:21:32.159> saying<00:21:32.400> if<00:21:32.580> you<00:21:32.880> have 00:21:32.990 --> 00:21:33.000 align:start position:0% the the so you're saying if you have 00:21:33.000 --> 00:21:35.210 align:start position:0% the the so you're saying if you have this<00:21:33.240> large<00:21:33.419> tree<00:21:33.840> if<00:21:34.440> if<00:21:34.620> it's<00:21:34.799> the<00:21:34.980> job<00:21:35.100> of 00:21:35.210 --> 00:21:35.220 align:start position:0% this large tree if if it's the job of 00:21:35.220 --> 00:21:36.590 align:start position:0% this large tree if if it's the job of the<00:21:35.340> implementation<00:21:35.820> in<00:21:36.059> splitting<00:21:36.480> into 00:21:36.590 --> 00:21:36.600 align:start position:0% the implementation in splitting into 00:21:36.600 --> 00:21:38.990 align:start position:0% the implementation in splitting into smaller<00:21:37.080> trees<00:21:37.380> no<00:21:37.559> no<00:21:37.740> this<00:21:38.039> this<00:21:38.220> is<00:21:38.460> this<00:21:38.760> is 00:21:38.990 --> 00:21:39.000 align:start position:0% smaller trees no no this this is this is 00:21:39.000 --> 00:21:41.270 align:start position:0% smaller trees no no this this is this is completely<00:21:39.360> orthogonal<00:21:40.260> to<00:21:40.380> that<00:21:40.500> so<00:21:41.039> what 00:21:41.270 --> 00:21:41.280 align:start position:0% completely orthogonal to that so what 00:21:41.280 --> 00:21:42.770 align:start position:0% completely orthogonal to that so what you<00:21:41.460> guys<00:21:41.520> are<00:21:41.700> doing<00:21:41.820> in<00:21:42.000> Lighthouse<00:21:42.419> is 00:21:42.770 --> 00:21:42.780 align:start position:0% you guys are doing in Lighthouse is 00:21:42.780 --> 00:21:44.930 align:start position:0% you guys are doing in Lighthouse is splitting<00:21:43.200> this<00:21:43.380> into<00:21:43.500> smaller<00:21:43.919> trees<00:21:44.340> and 00:21:44.930 --> 00:21:44.940 align:start position:0% splitting this into smaller trees and 00:21:44.940 --> 00:21:47.510 align:start position:0% splitting this into smaller trees and you<00:21:45.120> send<00:21:45.539> this<00:21:45.720> on<00:21:46.020> different<00:21:46.200> threads<00:21:46.679> to<00:21:47.220> to 00:21:47.510 --> 00:21:47.520 align:start position:0% you send this on different threads to to 00:21:47.520 --> 00:21:49.310 align:start position:0% you send this on different threads to to compute<00:21:48.299> them<00:21:48.360> in<00:21:48.480> parallel<00:21:48.780> this<00:21:49.260> is 00:21:49.310 --> 00:21:49.320 align:start position:0% compute them in parallel this is 00:21:49.320 --> 00:21:51.049 align:start position:0% compute them in parallel this is completely<00:21:49.620> different<00:21:49.799> you<00:21:50.159> just<00:21:50.340> pass<00:21:50.580> the 00:21:51.049 --> 00:21:51.059 align:start position:0% completely different you just pass the 00:21:51.059 --> 00:21:53.450 align:start position:0% completely different you just pass the entire<00:21:51.299> slot<00:21:51.720> so<00:21:51.960> big<00:21:52.200> Trace<00:21:52.620> is<00:21:52.799> the<00:21:53.039> big<00:21:53.220> gain 00:21:53.450 --> 00:21:53.460 align:start position:0% entire slot so big Trace is the big gain 00:21:53.460 --> 00:21:54.830 align:start position:0% entire slot so big Trace is the big gain for<00:21:53.640> this<00:21:53.760> you're<00:21:54.120> going<00:21:54.240> to<00:21:54.360> be<00:21:54.480> at<00:21:54.720> least 00:21:54.830 --> 00:21:54.840 align:start position:0% for this you're going to be at least 00:21:54.840 --> 00:21:56.570 align:start position:0% for this you're going to be at least hashing<00:21:55.260> four<00:21:55.440> times<00:21:55.559> faster<00:21:55.919> than<00:21:56.280> the 00:21:56.570 --> 00:21:56.580 align:start position:0% hashing four times faster than the 00:21:56.580 --> 00:21:58.250 align:start position:0% hashing four times faster than the standard<00:21:56.820> Library<00:21:57.059> you<00:21:57.480> just<00:21:57.600> pass<00:21:57.900> the 00:21:58.250 --> 00:21:58.260 align:start position:0% standard Library you just pass the 00:21:58.260 --> 00:22:00.470 align:start position:0% standard Library you just pass the entire<00:21:58.559> slice<00:21:58.919> like<00:21:59.580> all<00:21:59.820> of<00:21:59.940> the<00:22:00.059> slides<00:22:00.360> that 00:22:00.470 --> 00:22:00.480 align:start position:0% entire slice like all of the slides that 00:22:00.480 --> 00:22:02.930 align:start position:0% entire slice like all of the slides that you<00:22:00.600> of<00:22:00.720> the<00:22:00.900> bottom<00:22:01.140> slice<00:22:01.500> and<00:22:02.100> the<00:22:02.280> what 00:22:02.930 --> 00:22:02.940 align:start position:0% you of the bottom slice and the what 00:22:02.940 --> 00:22:05.270 align:start position:0% you of the bottom slice and the what what<00:22:03.299> this<00:22:03.600> thing<00:22:03.720> is<00:22:03.840> going<00:22:04.020> to<00:22:04.140> do<00:22:04.320> is<00:22:04.980> not 00:22:05.270 --> 00:22:05.280 align:start position:0% what this thing is going to do is not 00:22:05.280 --> 00:22:07.730 align:start position:0% what this thing is going to do is not split<00:22:06.000> into<00:22:06.120> subterris<00:22:06.900> what<00:22:07.320> the<00:22:07.500> thing<00:22:07.620> is 00:22:07.730 --> 00:22:07.740 align:start position:0% split into subterris what the thing is 00:22:07.740 --> 00:22:10.250 align:start position:0% split into subterris what the thing is going<00:22:07.919> to<00:22:07.980> do<00:22:08.100> is<00:22:08.280> grab<00:22:08.640> have<00:22:09.240> as<00:22:09.659> many<00:22:09.840> blocks 00:22:10.250 --> 00:22:10.260 align:start position:0% going to do is grab have as many blocks 00:22:10.260 --> 00:22:12.649 align:start position:0% going to do is grab have as many blocks as<00:22:10.500> possible<00:22:10.860> it<00:22:11.460> can<00:22:11.640> fit<00:22:11.880> in<00:22:12.120> your<00:22:12.240> registers 00:22:12.649 --> 00:22:12.659 align:start position:0% as possible it can fit in your registers 00:22:12.659 --> 00:22:14.630 align:start position:0% as possible it can fit in your registers and<00:22:13.320> then<00:22:13.380> just<00:22:13.559> go<00:22:13.799> to<00:22:13.860> the<00:22:14.039> next<00:22:14.159> chunk<00:22:14.460> and 00:22:14.630 --> 00:22:14.640 align:start position:0% and then just go to the next chunk and 00:22:14.640 --> 00:22:16.130 align:start position:0% and then just go to the next chunk and the<00:22:14.820> next<00:22:14.880> chunk<00:22:15.120> and<00:22:15.299> the<00:22:15.480> next<00:22:15.539> chunk<00:22:15.840> like 00:22:16.130 --> 00:22:16.140 align:start position:0% the next chunk and the next chunk like 00:22:16.140 --> 00:22:18.289 align:start position:0% the next chunk and the next chunk like this<00:22:16.320> so<00:22:16.860> it<00:22:16.980> doesn't<00:22:17.159> split<00:22:17.520> in<00:22:17.700> sub<00:22:17.880> trees 00:22:18.289 --> 00:22:18.299 align:start position:0% this so it doesn't split in sub trees 00:22:18.299 --> 00:22:21.110 align:start position:0% this so it doesn't split in sub trees okay<00:22:18.780> and<00:22:19.080> so<00:22:19.260> you<00:22:19.679> faded<00:22:20.100> all<00:22:20.460> of<00:22:20.640> the<00:22:20.760> leaves 00:22:21.110 --> 00:22:21.120 align:start position:0% okay and so you faded all of the leaves 00:22:21.120 --> 00:22:22.970 align:start position:0% okay and so you faded all of the leaves and<00:22:21.780> then<00:22:21.900> it<00:22:22.020> then<00:22:22.320> it<00:22:22.440> produces<00:22:22.799> the 00:22:22.970 --> 00:22:22.980 align:start position:0% and then it then it produces the 00:22:22.980 --> 00:22:24.669 align:start position:0% and then it then it produces the intermediate<00:22:23.520> level 00:22:24.669 --> 00:22:24.679 align:start position:0% intermediate level 00:22:24.679 --> 00:22:27.110 align:start position:0% intermediate level next<00:22:25.679> layer<00:22:25.980> and<00:22:26.280> then<00:22:26.340> you<00:22:26.460> just<00:22:26.580> feed<00:22:26.760> it<00:22:26.940> the 00:22:27.110 --> 00:22:27.120 align:start position:0% next layer and then you just feed it the 00:22:27.120 --> 00:22:29.149 align:start position:0% next layer and then you just feed it the next<00:22:27.240> layer<00:22:27.600> entirely<00:22:28.140> and<00:22:28.559> it<00:22:28.679> produces<00:22:29.039> the 00:22:29.149 --> 00:22:29.159 align:start position:0% next layer entirely and it produces the 00:22:29.159 --> 00:22:31.549 align:start position:0% next layer entirely and it produces the next<00:22:29.280> one<00:22:29.520> oh<00:22:30.120> yeah<00:22:30.360> okay<00:22:30.600> so<00:22:31.080> for<00:22:31.260> the<00:22:31.380> state 00:22:31.549 --> 00:22:31.559 align:start position:0% next one oh yeah okay so for the state 00:22:31.559 --> 00:22:33.289 align:start position:0% next one oh yeah okay so for the state so<00:22:31.980> you're<00:22:32.100> thinking<00:22:32.340> in<00:22:32.520> the<00:22:32.580> beacon<00:22:33.059> State 00:22:33.289 --> 00:22:33.299 align:start position:0% so you're thinking in the beacon State 00:22:33.299 --> 00:22:35.330 align:start position:0% so you're thinking in the beacon State for<00:22:33.720> the<00:22:33.900> state<00:22:34.020> this<00:22:34.380> is<00:22:34.559> incredibly<00:22:35.159> fast 00:22:35.330 --> 00:22:35.340 align:start position:0% for the state this is incredibly fast 00:22:35.340 --> 00:22:37.130 align:start position:0% for the state this is incredibly fast but<00:22:35.940> this<00:22:36.059> is<00:22:36.179> not<00:22:36.240> how<00:22:36.419> we<00:22:36.600> hash<00:22:36.840> the<00:22:37.020> state 00:22:37.130 --> 00:22:37.140 align:start position:0% but this is not how we hash the state 00:22:37.140 --> 00:22:39.110 align:start position:0% but this is not how we hash the state because<00:22:37.380> we<00:22:37.679> have<00:22:37.860> it<00:22:38.039> in<00:22:38.340> cashier<00:22:38.760> typically 00:22:39.110 --> 00:22:39.120 align:start position:0% because we have it in cashier typically 00:22:39.120 --> 00:22:41.210 align:start position:0% because we have it in cashier typically and<00:22:39.419> you<00:22:39.600> just<00:22:39.659> have<00:22:39.840> a<00:22:40.020> few<00:22:40.140> a<00:22:40.500> few<00:22:40.740> nodes<00:22:40.980> that 00:22:41.210 --> 00:22:41.220 align:start position:0% and you just have a few a few nodes that 00:22:41.220 --> 00:22:43.610 align:start position:0% and you just have a few a few nodes that you're<00:22:41.520> you're<00:22:41.820> you're<00:22:42.179> you're<00:22:42.360> changing<00:22:43.020> so 00:22:43.610 --> 00:22:43.620 align:start position:0% you're you're you're you're changing so 00:22:43.620 --> 00:22:46.850 align:start position:0% you're you're you're you're changing so when<00:22:44.039> you<00:22:44.220> hash<00:22:44.520> the<00:22:44.880> dirty<00:22:45.840> trees<00:22:46.140> the<00:22:46.380> dirty 00:22:46.850 --> 00:22:46.860 align:start position:0% when you hash the dirty trees the dirty 00:22:46.860 --> 00:22:47.630 align:start position:0% when you hash the dirty trees the dirty leaves 00:22:47.630 --> 00:22:47.640 align:start position:0% leaves 00:22:47.640 --> 00:22:50.270 align:start position:0% leaves then<00:22:48.539> well<00:22:49.020> there's<00:22:49.440> two<00:22:49.679> things<00:22:49.860> that<00:22:50.039> happen 00:22:50.270 --> 00:22:50.280 align:start position:0% then well there's two things that happen 00:22:50.280 --> 00:22:51.590 align:start position:0% then well there's two things that happen so 00:22:51.590 --> 00:22:51.600 align:start position:0% so 00:22:51.600 --> 00:22:53.330 align:start position:0% so sometimes<00:22:52.260> you<00:22:52.440> have<00:22:52.559> several<00:22:52.860> vendor 00:22:53.330 --> 00:22:53.340 align:start position:0% sometimes you have several vendor 00:22:53.340 --> 00:22:55.730 align:start position:0% sometimes you have several vendor consecutive<00:22:53.940> and<00:22:54.659> then<00:22:54.780> you<00:22:54.960> can<00:22:55.080> you<00:22:55.440> can<00:22:55.559> be 00:22:55.730 --> 00:22:55.740 align:start position:0% consecutive and then you can you can be 00:22:55.740 --> 00:22:57.950 align:start position:0% consecutive and then you can you can be smart<00:22:55.919> and<00:22:56.400> pass<00:22:56.580> this<00:22:57.000> consecutive<00:22:57.480> layer<00:22:57.840> to 00:22:57.950 --> 00:22:57.960 align:start position:0% smart and pass this consecutive layer to 00:22:57.960 --> 00:22:59.930 align:start position:0% smart and pass this consecutive layer to this<00:22:58.200> hash<00:22:58.500> or<00:22:59.159> you<00:22:59.340> can<00:22:59.400> just<00:22:59.520> use<00:22:59.760> whatever 00:22:59.930 --> 00:22:59.940 align:start position:0% this hash or you can just use whatever 00:22:59.940 --> 00:23:02.270 align:start position:0% this hash or you can just use whatever you're<00:23:00.240> using<00:23:00.539> now<00:23:00.780> just<00:23:01.140> crash<00:23:01.620> two<00:23:01.919> blocks 00:23:02.270 --> 00:23:02.280 align:start position:0% you're using now just crash two blocks 00:23:02.280 --> 00:23:04.010 align:start position:0% you're using now just crash two blocks at<00:23:02.400> a<00:23:02.520> time<00:23:02.640> you're<00:23:03.299> not<00:23:03.480> going<00:23:03.659> to<00:23:03.780> get<00:23:03.900> the 00:23:04.010 --> 00:23:04.020 align:start position:0% at a time you're not going to get the 00:23:04.020 --> 00:23:05.570 align:start position:0% at a time you're not going to get the factorization<00:23:04.559> impact<00:23:04.980> but<00:23:05.220> you're<00:23:05.340> going<00:23:05.460> to 00:23:05.570 --> 00:23:05.580 align:start position:0% factorization impact but you're going to 00:23:05.580 --> 00:23:07.970 align:start position:0% factorization impact but you're going to get<00:23:05.700> at<00:23:05.820> least<00:23:05.940> the<00:23:06.120> 20<00:23:06.299> of<00:23:06.780> the<00:23:06.900> hard<00:23:07.020> quality 00:23:07.970 --> 00:23:07.980 align:start position:0% get at least the 20 of the hard quality 00:23:07.980 --> 00:23:10.070 align:start position:0% get at least the 20 of the hard quality the<00:23:08.400> the<00:23:08.400> padding<00:23:08.940> block 00:23:10.070 --> 00:23:10.080 align:start position:0% the the padding block 00:23:10.080 --> 00:23:11.870 align:start position:0% the the padding block yeah<00:23:10.200> Okay<00:23:10.620> cool<00:23:10.919> so<00:23:11.100> I<00:23:11.400> guess<00:23:11.520> there's<00:23:11.700> maybe 00:23:11.870 --> 00:23:11.880 align:start position:0% yeah Okay cool so I guess there's maybe 00:23:11.880 --> 00:23:13.310 align:start position:0% yeah Okay cool so I guess there's maybe like<00:23:12.120> an<00:23:12.240> argument<00:23:12.480> for<00:23:12.720> maybe<00:23:12.960> it's<00:23:13.140> quite 00:23:13.310 --> 00:23:13.320 align:start position:0% like an argument for maybe it's quite 00:23:13.320 --> 00:23:14.930 align:start position:0% like an argument for maybe it's quite useful<00:23:13.559> for<00:23:13.740> small<00:23:13.919> trades 00:23:14.930 --> 00:23:14.940 align:start position:0% useful for small trades 00:23:14.940 --> 00:23:16.370 align:start position:0% useful for small trades um<00:23:15.000> you<00:23:15.179> don't<00:23:15.299> get<00:23:15.419> a<00:23:15.539> lot<00:23:15.600> from<00:23:15.780> caching<00:23:16.080> yeah 00:23:16.370 --> 00:23:16.380 align:start position:0% um you don't get a lot from caching yeah 00:23:16.380 --> 00:23:18.049 align:start position:0% um you don't get a lot from caching yeah it's<00:23:16.500> it's<00:23:16.679> not<00:23:17.100> really<00:23:17.280> useful<00:23:17.580> for<00:23:17.820> small 00:23:18.049 --> 00:23:18.059 align:start position:0% it's it's not really useful for small 00:23:18.059 --> 00:23:20.149 align:start position:0% it's it's not really useful for small trees<00:23:18.419> you<00:23:18.659> only<00:23:18.780> get<00:23:18.960> the<00:23:19.080> 20<00:23:19.260> of<00:23:19.740> the<00:23:19.860> padding 00:23:20.149 --> 00:23:20.159 align:start position:0% trees you only get the 20 of the padding 00:23:20.159 --> 00:23:23.149 align:start position:0% trees you only get the 20 of the padding block<00:23:20.340> yeah<00:23:20.820> it's<00:23:21.480> very<00:23:21.659> very<00:23:21.840> useful<00:23:22.320> on 00:23:23.149 --> 00:23:23.159 align:start position:0% block yeah it's very very useful on 00:23:23.159 --> 00:23:25.070 align:start position:0% block yeah it's very very useful on large<00:23:23.400> trees<00:23:23.820> and<00:23:24.000> here<00:23:24.120> large<00:23:24.360> means<00:23:24.780> more 00:23:25.070 --> 00:23:25.080 align:start position:0% large trees and here large means more 00:23:25.080 --> 00:23:26.750 align:start position:0% large trees and here large means more than<00:23:25.260> eight<00:23:25.380> blocks<00:23:25.740> so<00:23:26.100> anything<00:23:26.280> that<00:23:26.580> has 00:23:26.750 --> 00:23:26.760 align:start position:0% than eight blocks so anything that has 00:23:26.760 --> 00:23:29.270 align:start position:0% than eight blocks so anything that has depth<00:23:27.299> more<00:23:27.840> than<00:23:27.960> two<00:23:28.200> this<00:23:28.740> is<00:23:28.860> already<00:23:29.039> four 00:23:29.270 --> 00:23:29.280 align:start position:0% depth more than two this is already four 00:23:29.280 --> 00:23:30.410 align:start position:0% depth more than two this is already four times<00:23:29.460> faster 00:23:30.410 --> 00:23:30.420 align:start position:0% times faster 00:23:30.420 --> 00:23:34.579 align:start position:0% times faster [Applause]