Player is loading...
Run Scala Faster with GraalVM on any Platform
Vojin Jovanovic, Oracle
Thursday, June 13, 2019 · 10:16 a.m. · 41m 01s
Can we write a Scala program that runs on all mobile and embedded platforms? Can we use Scala to develop cross-platform user interfaces without writing platform-specific code? Yes, and it will run fast. In this talk, we demo a nontrivial Scala GUI that runs on both iOS and Android via GraalVM. We explain how GraalVM runs Scala programs natively on different target platforms with better performance than the JVM. Then, we show how JavaFX allows us to develop consistent cross-platform user interfaces without writing platform-specific code and invite the community to join our effort towards the cross-platform Scala.
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:00
for a while and mining is when you're on it and i'm working or collapse integral renting
00:00:06
um today i'm going to first of all thank you all for coming and
00:00:10
i'm especially glad that my little son is your first time to see me present
00:00:20
sorry
00:00:22
and and today i'm going to talk to you about probably an and collagen grandfather faster without them
00:00:27
um it is kind of known already but i've just come from that with the new results
00:00:32
and then i'll talk to you about the new thing that we did in the
00:00:35
last six months and that is running um scott faster with volume on any platform
00:00:42
so for those who don't know what guardian is this is probably there are slight bit summaries is it best
00:00:48
so i got them is a virtual machine or language wrong thing which is built uh on top of the grout compiler
00:00:57
and it enables us to run a menu languages so it's a particular something
00:01:01
and you're here on the slide you can see you can run languages like java a skyline coupling
00:01:06
but then together with them in the same and then you can look round dynamic languages like doubts which will be are like them
00:01:13
and then it's really cool that can also run all of the native languages because it's all support the l. b. m.
00:01:18
uh interpreter and then on the bottom of the goddamn sentimental there isn't ruddy glow compiler
00:01:24
and then on button you can run that on the top of course but just like you do it's colour or you can
00:01:30
run it as an o. j. s. application you can run it inside
00:01:33
of databases another native but from such as oracle database of my scalp
00:01:37
or you can run it as a standalone shell script shape and we'll have a
00:01:42
bit more today and this this picture so what's really in the core of this uh
00:01:48
probably and so i won't have time to go into the details but the core is the broad compiler write it all revolves around it
00:01:55
and what's the key difference of the grout compiler to the others um is a
00:01:59
few things it's written in java right so the previous compiler see two was written in
00:02:04
c. plus plus and it was essentially very kind of in tangle with the v. m. in the courts but yeah
00:02:10
right and then broke in and it was written in java and get the clear separation appeal interface to the yeah
00:02:15
so it it kind of became very more modular meant the noble and
00:02:18
extensible mexican plugin interface you can defector things so it makes development release
00:02:24
and the crowd compiler comes in two flavours which we'll see in all of the graphs that they show trial though
00:02:30
draw the presentation so there is a community vision or short c. which is essentially the open source a part of the
00:02:37
girl compiler and there's an enterprise edition which as a few
00:02:40
advanced optimisation is which will basically uh add to the performance right
00:02:46
and our general goal is that uh you know the a. p. i. of the grout yam doesn't really change
00:02:52
you better you're in like community doesn't run enterprise edition just in
00:02:56
the performance and then uh because of this nice kind of um
00:03:02
nice interface and you know modularity and extends it it it it was very easy to develop really cool optimisation smell
00:03:08
so what makes crawl faster i would think there is a lot of them right but i would take three all is which i like the most
00:03:16
which is partially state analysis a publication and priority mine act so these two three i think
00:03:22
make rock kind of unique compared to the other compilers and especially how they interact with each other
00:03:29
so if you wanna know more about these you can go and look at the papers about which
00:03:34
ah but in in short you know if you have something like this streams operation uh that you see on the top here
00:03:42
uh essentially what but the grout will the will says that so that's what
00:03:45
everyone programs today specially starts colour you just do you know collection operations high level
00:03:50
uh would grow with the with the with the combination of all these optimisation could use
00:03:54
in the end something like this level right actually measure performance i looked at the grass and measure performance
00:04:00
and it's actually not quite there yet it's there's a few percent
00:04:03
difference but what you would get this is sensitive like writing about
00:04:08
okay so that's the cortical and why is this good for us
00:04:13
while we might be talking to discover community that well this colour language
00:04:18
'cause many obstructions kinda modern any right you know in a very high level of
00:04:23
your you explicitly say what you want to do any captains and in these cases
00:04:28
robin is especially good and so here on this graph the essentially and the
00:04:34
problem with all the benchmarks before uh that kind of show the performance of
00:04:39
a virtual machines and and uh languages was that they were can return for these old
00:04:44
school style programming into other you have like big functions with by loops you know and then
00:04:50
and then if you in line those doesn't really matter much but that's not actually what we can to the soul
00:04:56
uh what we've done is we developed this uh new benchmarking seed called renaissance
00:05:01
um and essentially it's kind of neat three present more with people break today
00:05:06
right and if you look on this graph um on the uh essentially most of
00:05:10
the prosody just speed up on the left and then you have benchmarks on the bottom
00:05:15
and then you have a speed the for the individual and look at what do you mean at
00:05:19
the end okay so that's the do you mean is actually but you really care about big yeah
00:05:23
right and here we have the l. e. e. n. c. which we've seen the beavers like
00:05:29
and on the y. axis is the speed that right so and then you look
00:05:32
at this um renaissance ben franklin represents what people write today i especially is colour
00:05:38
you will see that you get the do you mean of of thirty two percent in problem like so if you run our
00:05:44
your program we'd rather e. e. you will get thirty two percent
00:05:48
approve endangering fit c. r. which would see a shortly how easy this
00:05:54
you will get a six percent right now and then there is the specifics colour the benchmark which is just
00:06:00
first colour and it has a lot of like tools and stuff firms colour ecosystem in here you will again see
00:06:07
uh thirty eight percent for val enterprise and seven percent for c. and
00:06:13
because of this paper is already using it in production so many feet
00:06:17
you know the call that this that is the teacher around is running relatives
00:06:21
generated but got so you know this is a solid technology it won't fail
00:06:27
um and then before you know they were people are always going out um there was one problem that the crowd was written in
00:06:34
java so people would always say valley yeah but then it starts
00:06:37
lawrence on and in the last year but we developed is uh
00:06:41
i like recall the brow we just takes the call grow like roll compiler and compiles
00:06:46
it is a native library right and then put it is uh did compiler in court
00:06:52
so here on the essentially and then repackage it isn't a which is a soon see what it is
00:06:59
and uh every compilation is gonna separate things are just come by the
00:07:02
function and draw a the whole the whole heap when you start or right
00:07:07
and it's actually been able to see on the on the left you will see that you know
00:07:11
on the on the right is what happened before this is a simple streams example that they a compost
00:07:16
you will see that the session the first run it would take a second
00:07:18
to do a first iteration and then it will be like slowly get to that
00:07:22
to the to the performance of on the on the side even see how it kind of you know immediately goes
00:07:29
to the essentially kind of right from the get go it becomes fast and when you run this
00:07:34
would see to is basically pretty much the same numbers right so so now we have the the grout
00:07:39
a compiler you will but whenever you run europe it will run as fast the c. two in it to warm up this past this too
00:07:46
which brings us to the next point um it's essentially using
00:07:51
program for faster computation souls to discover compiler um he's notorious
00:07:57
for essentially being very slow to compile like every scholar these it is it that people would be like
00:08:01
a local that's lot as well um so but if they're done here is we use rather for compilation because
00:08:09
for compilation there is another particularly important uh thing is that it's not only of the number of machines that run
00:08:15
somewhere in the cloud which you don't see it actually or time i compilation whenever you do see a a test
00:08:20
computation is the first thing you do in a static language and because of that
00:08:24
it's your time you wait for it so you on this time to get shorter
00:08:29
that's why i'm this slide lower is but basically is the time essentially the compile and then
00:08:35
on the graphical sissy i took the numbers from the three three big projects one is
00:08:40
the uh uh a leech s. which is written is colour and the font that forgot them
00:08:45
um and then we'll have we compare like running the normal java java eight
00:08:50
running java guardian and here i'm using enterprise edition only know community edition because
00:08:56
in this case i think it's you know your time and your productivity so you know money doesn't really matter
00:09:02
ah and essentially and it's not expensive for these days so essentially and then we used
00:09:08
loop which is a new framework first calls colour compilation uh develop a door from skeleton turf
00:09:14
and essentially we lose lupin bloopers was probably are there other famous speed
00:09:19
ups colour compilation hydra and the numbers for them are pretty much the same
00:09:23
and so if we keep this if you look at here you know compared to normal job i will get
00:09:27
anywhere from like thirty twenty five percent speed and say well i
00:09:32
never get your results faster than it twenty five to thirty five percent
00:09:36
and then also if you have like a little between a perilous is this
00:09:39
colour fill them in this case actually bite size if i pipelines in their license
00:09:43
you will uh if you look at the similar improvements i mean if
00:09:47
you don't see that much because the the you know the first numbers smaller
00:09:50
but you will get pretty much the same speed so any compilation framework to use it quite a bit just commercial blow
00:09:57
um or scholar you will get about thirty percent faster village and that's your back and so um
00:10:05
and so this brings me to the conclusion of this first
00:10:07
part about well known running skull of foster draw here is essentially
00:10:14
uh using c. to make the all the the compiler in java instead of grow
00:10:19
it's just a waste of time and energy and money i saw it was just say say look basting save seven
00:10:25
percent of your performance and all you need to do is basically for jake eleven would be so now come on line
00:10:32
and that's it you know sample sensor fit down right on or you can download got the a. b.
00:10:39
m. c. e. or e. e. and put it in the buttons on the whole it's that these so
00:10:44
and you know transit peter uh people are already using it for compilation all around
00:10:48
so there's absolutely no reason to not use it just drawing drawing money time and so on out the window
00:10:56
um and then it but that's something to get you give you about thirty five
00:10:59
percent speed up which is you know if you have a service running without the machines
00:11:05
ah you know multiply the numbers it's a shepherd people thing to do an r. f.
00:11:12
and then of course for compilation for me this is like a no brainer you know you have a team of a
00:11:16
hundred people waiting thirty percent more for something and you have like one or two or three machines are five thing see i
00:11:22
it's just like a no brainer get get your done stuff done faster
00:11:28
cool um so that was the part the bog roll 'em in
00:11:31
running style of austin well but then the data is also notorious
00:11:36
for it's slow startup right so we develop in in in on the rowdy and we have
00:11:42
this essentially project called native image or people coloured grout yeah make it into just native village
00:11:47
uh which allows us to runs colour a bit fast startup and no memory proper
00:11:52
so so how does this work ah on the on the left side we start with like just like a
00:11:58
dollar you say java their c. p. and the main class with the same right everything of your application libraries
00:12:04
ah java class library and i even have something called substrate the
00:12:08
um reaches a v. m. developed in java that will kind of
00:12:12
chip together with all the lights off and then what we
00:12:15
do inside of a naked image is we do a an iterative
00:12:20
i'm in interactive analysis it's a variation of a point analysis which will
00:12:25
kind of try to see which called from the you know all of these
00:12:28
is reachable right so but but but they're using and even if you're using
00:12:32
not using some feels it would just pick the fields out you don't need that
00:12:36
right and then it runs interactive in another interesting thing it does it re initialise is or stuff so
00:12:42
so essentially you can kind of run parts of your application ahead of time if you need some configuration or
00:12:48
if you need some parsing some files you know which takes a lot of time you can essentially come and uh
00:12:55
you can just do that at build time and then it will produce
00:13:00
uh it would be kind of see lies in the keep right and then maybe a a to compare the
00:13:05
calls and take the snapshot of these he and we'll use the drought compiler of course for the same thing
00:13:11
and for just machine go right so you will get machine code is one object file you
00:13:16
need to keep is another object finding together to get an education this executable or shared library
00:13:22
so and this executable you can run just like any other make your little on pretty much any plot
00:13:27
okay so windows is it in the signal so that i mean really really uh and
00:13:31
also the support a are sixty four and andy for now but more is to come
00:13:37
um and yeah so but you get to this so called this but it's really easy it's colours like a little time hello world
00:13:45
seventy milliseconds to print like online that's it ridiculous ah
00:13:49
and then in the native image you plus the library
00:13:52
we could make this a big maybe we should do something like there's the schools colour to make this nicer
00:13:58
and it's a killer world hello world this is now just like the c. program and it's fun
00:14:03
it's it's okay and now by this interesting is because these days you know it's like micro services
00:14:10
uh along those are becoming very popular and there's a whole ton of
00:14:13
these frameworks essentially you know for micro services and what happens in the clout
00:14:18
is that basically people are um you know the cloud providers
00:14:22
or can remove in your service it's they're not start up right
00:14:26
and so when the new request comes the services to both and then your call the and its first
00:14:30
the current only to start and then your v. m. it's just thought then that takes health problem qualify
00:14:35
and here on this slide we have uh the start time for some of the frameworks is that not all the famous actor could talk about this later
00:14:42
ah but they're very popular in the coming off so for collide and so we'll
00:14:47
have again data cagey to get all and make it today and it gets acolyte
00:14:52
and that's uh the an open source fame and develop an article for micro services
00:14:56
you will get like one second on top of the book time of the crime itself
00:15:00
so it's the brooklyn secondly native image thirty five milliseconds so basically start than it just that
00:15:05
start your answering depressed so this means if your lawn the scolding i get the request
00:15:11
requests will be out before you break your service level agreement right
00:15:14
your customer will have to wait for like a second get the peach
00:15:19
i'm off and similar thing from i cannot and porpoise which is good entirely based on volume you get
00:15:25
sixty milliseconds right so it just you know service up and then another thing which couldn't saves money ah
00:15:34
is basically the memory footprint right because now encouraged compiler human goblin interpreter you don't have
00:15:39
the gloss information it's all kind of burnt into this binary you know very minimal uncompressed they
00:15:45
um you will essentially you know uh the memory footprint for essentially i'm running the first
00:15:52
request for these applications it will be about content twenty max and in seventy under an eighty
00:15:58
and with with native image it'd be about like forks less like to get thirty forty and seventy right
00:16:04
so and now you know you can you can you spend like last memory which means it's cheaper so that's it
00:16:13
um f. good so one thing that i tried before this
00:16:18
conference is getting this so to native image 'cause one limitation
00:16:22
uh in theory and there is a few or no it's still in
00:16:24
active development it's getting quite close to being ready um for why that option
00:16:30
i would say that they try that use colour famous and i notice that the calls colour communities not really trying to put this famous you
00:16:36
only get now does this you get the tone of you shoes and
00:16:39
get up and like ninety five percent of the tissues or java framework so
00:16:44
so essentially i try the fiscal the famous and i found some things that are really not working which
00:16:48
which we can fix in a matter of days um and and i can i'm not is that essentially
00:16:55
oh and of the school community doesn't really not know all this so it doesn't wanna try or doesn't report issues so on you know uh
00:17:01
this this is our priority not just open and you should get up
00:17:05
being if someone is not responding we're very you could've overloaded these these
00:17:09
say hey you know can you fix my favourite show statistics to frame batteries pretty okay and and
00:17:15
others in the t. i. so so it's kind of i think it's important to our report issues
00:17:21
test it in your c. i. if you are framer quarter of play just put native image in your c. i.
00:17:26
because it will also attract users to your framework because other start up in thirty five milliseconds and
00:17:32
you don't so it's better that you also support that so you start in like fifteen seconds or whatever
00:17:37
um so putting your c. i. and that way to keep it's all i can
00:17:41
really be provided all the flags and all the infrastructure can easily download native image
00:17:46
and you can run it i guess i'll just put it in your best um and here we all we are actively
00:17:52
working with the community so just think that's on whatever means like even in a get up whatever you want bill will respond
00:18:01
cool so they did image has another thing so whenever it's confirmatory assay
00:18:06
bad to compile java and dynamic languages in a method of blindly right
00:18:11
so what we provide in a native images something called profit guide optimisation say so
00:18:16
is actually the call brought compiler is built to mulder profiles
00:18:19
you can run in an interpreter get what your application is doing
00:18:23
and then um and then what happens is you basically uh get faster caught on
00:18:29
and we support this also would need to demonstrate it's called pro forgot optimisation so but it requires
00:18:34
it to bill to make it safer to make image feature instrument you'll get an instrument the binary
00:18:40
then you take a few of your workloads if it's of that service just send a few requests
00:18:45
uh you will get your profiles and then feed it to native imagery this flight and we'll get an optimist line
00:18:52
and what this gives us is that funny look essentially the performance of making image at the moment which has
00:18:58
still of a well known be located look too many can locating fruity to get a better
00:19:04
yeah so you if you get like this at that event and you get the cumulative number of requests sent it back to punch
00:19:11
ah you see in in the beginning of course they could images the red line it
00:19:14
will be a beta right because it starts from the very beginning at full speed right
00:19:20
ah and you will see that it needs about like a hundred thousand to five hundred to a million request
00:19:25
for the hotspot of warm up all that stuff and even when it's formed up it's just a tiny bit faster than the than the
00:19:31
naked right and that's mostly because or d. c. is currently not
00:19:35
the best thing in the role and we're working on anyone f. r.
00:19:39
but basically you know the native image and video you will get basically a rhino very soon
00:19:45
you will get faster code when you would get a just and running ah on the gate right
00:19:52
and it's also very important than a divinity school for command line tools so here on the right to give it to video force
00:19:58
colour for the particular because of all the obstruction that remove you
00:20:02
will get actually faster corbin hot it's not only that it starts fast
00:20:06
and his low memory but it will run fast it's of collecting the
00:20:09
the best i think it's like thirty seven percent faster than cut so
00:20:12
it's not like so if you wanna formative files you know in in
00:20:15
college or whatever it would just go to them like five milliseconds okay
00:20:21
both so now now we're standing on the shoulders of giants right because of the drought
00:20:26
compiler which makes stuff really thought we have the native image to compile into native code
00:20:32
and what's new would be been working slide in the last
00:20:34
six months is essentially extend this picture would be uh you know
00:20:41
three new thinks it's like deploring to mobile devices right so
00:20:44
what you're building is the point to iris android and desktop
00:20:50
and you know as a part of this effort uh we tried a few things we have a program that first of all
00:20:55
my ass is kind of the biggest problem because first belligerence it doesn't run on the in the sixty four so
00:21:00
you need like you know and they are compiler so we have to have forts want is to confuse other b. m.
00:21:06
and the other one is to use to growl native compilers that yet so what the
00:21:10
other this just for fun is to see you know how well the seller the um work
00:21:14
uh as the backend for making image annotate done is
00:21:18
we implement the namibian backup program just numbers when it's there
00:21:22
it's kind of still not in the best shape the um uh you cannot you'll see in the demo it works
00:21:27
um and what we need is one of the reasons that up responsibly threatening that it will require
00:21:34
essentially l. b. m. bit code for the applications should to the
00:21:37
i. s. i saw you know then ricochet paladin bit code indirectly
00:21:42
in case you know they pull the plug we can only say that we should yellow your
00:21:45
bit got a um and then another interesting thing with the television gives is that you can
00:21:52
you know relatively easily pork to new on each architectures if you're in the embedded market
00:21:57
like is it automotive or uh intuitive things you know uh you have some some some little
00:22:04
kind of not so well supported uh architecture you can just you know
00:22:09
even if seven million back and which is most likely you're going to generate
00:22:12
that and you should be able to it a few weeks or maybe a
00:22:16
month or two you should be able to get eight inch something's all java
00:22:20
uh and colour programs running on your their butts right but it comes with a price
00:22:26
now this is like not done yet uh moving d. c. in alabama still experimental so
00:22:32
i'll i'll keep you posted right so you can get will probably when it's done when
00:22:35
we know exactly what the story uh with like double post about it and you know talked
00:22:42
so now let's see how this cross reference colour uh and
00:22:46
of course uh it's the same platform look like right so we
00:22:50
talked about culminated image now now you have all the you know
00:22:53
byte code compiling into the uh executable which runs on any platform
00:23:00
ah what we did is now we need to do the front that's right so how does the developer
00:23:04
and which library the views right so for that fee teamed up for this company called go on
00:23:10
and together we work on this tool set it is actually consist you started your i.
00:23:14
d. right and then you have a plugin reaches in our case it's currently greater than maven
00:23:19
but i think as b. t. you know the goal the code is there so just you know someone can pick it up and make them as people again
00:23:25
i'm off and then a committees java effects mobile it essentially ah
00:23:31
contains all the u. i. components that working on all the buttons it will ever come
00:23:35
and write ins and dust up right so also and and all of that is kind of
00:23:41
would be long tools and the static libraries cross compiled for all these buttons
00:23:46
um and compiled it got an eighteen inch and set to the
00:23:50
set to the your device that so so you kind of get um
00:23:55
you know of full full blown stack for development on mobile
00:23:59
uh and of course with these long stuff you will get also um
00:24:03
extensions for all the mobile stuff integration with the modem mobile devices like
00:24:07
obstructions for cameras g. p. s. and all the things like that right um and all of this i i is open source
00:24:16
this part and not part of lawn mobile there is a closed source part that's the bigger the earned money basically um
00:24:22
so that they do some mobile specific connectivity and they do something no libraries for this kind of cross platform stuff
00:24:28
but essentially everything you need is open source to the start and you know today
00:24:33
you can come out and start cracking your like cross platform overlap on the
00:24:37
g. yeah i'm off so i think it was said most of these things
00:24:45
um
00:24:47
yeah so that this was the only way physically before you know the j. v. m. was locked because
00:24:51
because running costs but there's not a lot on on iris norman 'cause developed the proper able to comply
00:24:58
there there were feel like an open source interview commercial
00:25:01
ones but not most built on on solid ground right
00:25:05
ah the the proper team supporting it so for the first time now here proper and look at the time compiler for java
00:25:11
so we can deploy all these devices right and you
00:25:14
can kind of spoken menu ecosystem uh based enjoy life
00:25:19
and then for java facts yeah it because native code but java facts is
00:25:22
not the only library in the world there's others how to to talk about later
00:25:27
um and yeah this is all the things i said and of course this all works only for job eleven twelve and thirty
00:25:33
right so don't wait is not supported so still starting from
00:25:36
java eleven because java facts starting eleven supports all the native stuff
00:25:41
uh and andrei this coming some you're working on that that would be no in them undermines out
00:25:47
ah so let's check this out uh how does this work right
00:25:54
so so here will have this chat at present them with the tree developed in our a. i. d.
00:26:01
and then you know you will have a just here is just a normal rate a project like
00:26:06
anything else and then what we can do is channel parted developer can develop and that's a pretty
00:26:12
so we can say just great to run uh i will not compile here and the reason for that is essentially that
00:26:17
the moment really celery and stuff it corpus wait too long for this presentation to show
00:26:22
um but yeah yeah this is an application now running on the best of
00:26:26
the trap that it's calloused out as carlos colour these no robbie i'm runs faster
00:26:35
and then can send the message yeah so you can just go
00:26:39
back to write the check stuff rather downgrade to run an adaptor kit
00:26:45
but now let's see what we're gonna do for the right
00:26:48
right ah well same project you say cradle uh but native or
00:26:55
right so i of course i can i can i skip the compilation step because
00:27:01
because now i put like they you know i mean it or two minutes to build sub didn't have time for that
00:27:06
so with this bill though is it to fire up a simulator and the simulator will go to the lawn poles and you can see that
00:27:12
it sees this message is actually downloaded from the internet and all this
00:27:16
stuff here i'm using is connected so it's all in the clot right
00:27:21
so you see the message just posted and because this is a simulator you need
00:27:24
to turn it off by clicking on this button on the side and then sliding what
00:27:30
okay and then you know once you're done developing in this environment
00:27:35
you basically want to finally test it on your device to just change one variable in this case a copy the project
00:27:42
and then you say again native around and you say is that the irises it it i as i as there
00:27:47
seem you will just say iris and then what this will do is you know i mean i do wrong with
00:27:55
because see
00:27:58
well you know same two messages right so it's all connected to the same thing and then you can basically start developing right
00:28:20
oh cool
00:28:24
so and that so many especially humans colour they they seen people are using uh
00:28:31
particular trick to get the way up about this block or that you
00:28:34
cannot do j. v. m. on on on mobile and they're using scarlet yes
00:28:39
to compile skyline to j. s. and then coke into this react native
00:28:44
so here i would just like to compare essentially you know what i'm what i'm trying to but they would like to see happening
00:28:51
is that essentially responded j. v. m. community around the mobile right because
00:28:56
it's not that i don't like javascript i think it was good to have you know it's good
00:28:59
to have two things down you know things done into the o. two or more solid base line
00:29:05
so also and i would just like to compare react native and now at this point in time you know we just thought that right so we just
00:29:11
this was the card this quite getting this like demo to run was the hard part of
00:29:15
course it's still body and stuff but you know in a few months it's it's it's fine on
00:29:20
but now and then the then the removal of the all of this noise ah i
00:29:24
want to just it's now compare it to react native and see what they think that happen
00:29:28
so i can now look at the day of course track native has like seventy thousand stars and get covered or whatever
00:29:33
um and you know because of that because of the speaker community it is more components which are developed
00:29:40
um and it has another feature that we still don't have any club called updates
00:29:44
already to make a bugging application you can just you know send new poll essentially
00:29:48
the the cold running on the on the devices unequal functions it to say
00:29:52
well this function abated and it will double the new one from the internet
00:29:56
uh but then it's essentially interpreted code and it's very slow its startup slow
00:30:03
it has essentially for complicated out you won't be able to run them and it burns better and so
00:30:10
uh and for joey facts already in this state you know first of all
00:30:14
the running on the data to get like a humongous ecosystem of java libraries
00:30:18
a readily available right in one of the demos which uh you know because of
00:30:21
the d. c. that doesn't work didn't make it we can manage to discover compiler
00:30:25
on on the thing can show to use and print reasons on ice or can just like
00:30:29
you know you cannot france telecom pilot project needed some it's not number right um and then
00:30:35
you will get a better performance like because you're based on well uh it's all they become pilot
00:30:40
you get like a full day them perform something like that so basically menu
00:30:44
and click on this up actually download some stuff from the internet or my phone
00:30:48
so so it's kind of waiting for the messages within a click on and that just got like you know orchid
00:30:55
our which we've seen also the the microscopes i say get better
00:30:58
performance and process that and what needs to happen now on the
00:31:02
on the j. v. m. is like you know there is to be a community right in it just people started packing stuff up um
00:31:09
and making you like base right and then it will be very better than reacting
00:31:14
right to get you know purely native up running fast with a better commitment and uh
00:31:21
and enough skylights in boston language to start this i'm basing this before and there was this like a dope and then
00:31:28
and then there was far right and and sparkles with install and only and then make a million people like talking else levels part
00:31:35
around the world right there was the same stuff without right there was that occur actors
00:31:39
and they're actors before but in oven at cactus come and there's like graces colour that's fine make
00:31:44
libraries and then like all of the object this is the mobile right and a similar with length breadth
00:31:50
ah and what i suggest to people that you don't want to start this project
00:31:55
i mean this this libraries like flatmate given flutter they all already developed all these needed
00:32:00
components and they're all open source i think so we i think people should just like
00:32:04
borrow that on the j. v. m. and kind of make a really nice library on
00:32:08
top of it so that you can start an ecosystem on the j. v. m. right
00:32:12
uh yeah so you know if you have free time just spun a pet
00:32:17
project you give a great to build start from there at this other plug in
00:32:22
and their ego as the t. plug in you know slowly starts you know how it goes ah it was pretty fast
00:32:28
ah and what they're on next steps on this project uh we still don't support d. c.
00:32:32
for us so you just like to this up a bit to to just like collect your windows
00:32:38
so that's not yet but that's like of equal to work from the problem we
00:32:42
need to the running on andrei that will take a a probably a couple of months
00:32:45
um but we need to do is we need to use the java byte can interpret the for development because at the moment you know and then i was compiling
00:32:52
stella either the um and a java effects look at thousand seconds last
00:32:56
night very fun and so we need to put the java interpreter so that
00:33:00
that's one of the advantages or react is very as you develop you see a stop stuff appearing
00:33:06
we need to make the same thing happens or in the java interpreter an
00:33:08
ipod or on under it it just it should go by corbin seat right away
00:33:13
and of course in into the support for under but these are details you will take care
00:33:17
of this uh it's time to start checking just basically right it's like it this will benefit
00:33:22
in the beginning every project is not perfect but you will be there with some like ha for years
00:33:28
ah off and you know this is kind of it uh i would like to
00:33:33
give special thanks to your can loss and uh uh like work on this project
00:33:37
we can be together ah like is uh actually did
00:33:41
he week using lawn and essentially he worked digital effects
00:33:44
and it's this is this is that what type of contributors that are best in the role he just says i want to continue to this project
00:33:51
and then uh basically the couple months you get you know all this will request
00:33:56
actually for for peaceful rip estimate the metabolic press saying we need an automated
00:34:00
like box that might just pull it was because this is not possible for me
00:34:04
so um so like to give special thanks a lot of this work was basically
00:34:10
uh his stuff and and why couldn't i love him back and you know
00:34:13
must just do than straight from university just put a million back and it works
00:34:17
pretty impressive that's um and yeah so also composed basically you know uh i would just say two things
00:34:26
and don't faced energy time and money just put this on a on
00:34:30
a java applet on a job uh you know they know java flats
00:34:34
and you know it's time to start then you'll come you know in your community
00:34:38
for cross platform development i think scott no skylight is the best place to start that
00:34:43
and you should just start here it's it's this link you follow it you can know the same stuff it it here you can do
00:34:49
on your i. s. device right so and it's now time to start
00:34:53
the earlier could've done with heart out or just cleaning stuff well thank you
00:35:15
so with my son hardy we should online so so uh and they have any questions
00:35:23
faith
00:35:28
hello i'm pretty sure i say to see and then stands it's good to see it again and i'm
00:35:32
working answer using native image damage you doesn't it clearly a suggestion i uh there was a couple slides on
00:35:37
it's less to essentially do remote compilations because we don't have to wait for g.
00:35:41
d. mister knapp each try and you'll be wanting to stick by on a different
00:35:45
i machine that's very nice that's very cool when i appreciate a lot uh i was asked a question
00:35:50
which was a it say for example which you you know one hundred percent like the extreme for men
00:35:55
a p. g. an essence compression image and with that the eco great
00:35:59
concern or that be something that would be yeah exactly but across projects
00:36:03
oh it's a it's a tough question thanks for asking it so easy answer to it it's not that
00:36:08
i don't think i don't think i'm in a position to and so i think it's these two particular things
00:36:14
uh yeah i cannot i cannot answer this publicly uh yeah i yeah and
00:36:21
mm yes expressed similar i think he's joking transistor things in particular um
00:36:30
i think they're not so hard to do various things which are very hard to do and yeah these things are
00:36:35
moderately hard right and so so it would probably spawn opponent
00:36:40
discussion but there is a chance of this going to say
00:36:43
it's really good you think you had mentioned there is a chance but you know i cannot
00:36:46
of course i'm not completely not in a position to say is just like mike that feeling because
00:36:51
because it's not really a now it's uh it's not really the
00:36:54
core thing which is car which is you're keeping also sits there
00:36:58
think francine i'm even done very question but really we shouldn't unfair question but it's
00:37:02
i give you might you know this is a completely personal answer at that feeling
00:37:08
thank you so much
00:37:20
but
00:37:27
i i didn't understand how the native image decision memory footprint
00:37:32
uh_huh so probably the the what what happens when you start the java a
00:37:37
minister java this as it does a lot that it's that loading classes from disk
00:37:43
uh which are kind of not in an optimum formant then you know then you start interpreting those
00:37:48
classes right for you need an interpreter and an interpreter generates a ton of garbage and stuff uh
00:37:54
then you need to profile that as well right so you need a profile or any store all of this in memory for every method that you run
00:38:01
then in the c. one compiler which compiles so you could get past the warm up to compile once
00:38:07
and then after ten thousand times in in the c. two compiler which compiles the final version record which is up also need
00:38:13
you know to compilers one interpreter one profile or by cordless
00:38:17
loading and it's an open role second and you know it's all
00:38:21
i don't know what's where's it gonna need to keep like
00:38:24
massive data structures for that right so that's essentially whites right
00:38:30
yes yes that you can basically done in advance all you need to do is draw a
00:38:36
dynamic lost all the right so you can not the download the jar from the internet and running from it it right
00:38:41
so the this this is exclude some things like as the t. for example if you wanna make the image of that that's impossible right it will
00:38:47
in no way to download plug is a different versions it becomes a combinatorial explosion of like jars that you need
00:38:54
right but if you carry your mike the service if you carry your common line tool uh if you cover more well uh
00:39:01
fine i just put the jars on the task off you do a bit of configuration on your lap
00:39:09
that's a good question
00:39:24
you quickly and messing about actually specifically and the weed intensive been able to generated images dynamically
00:39:29
from niger is the downloaded and then buttons in images suggesting done any pressing gently isn't there intentionally
00:39:36
yeah but we can alter granted until two level so it actually
00:39:38
does change without any open source that's that that's yes yes that in
00:39:42
on so yeah it's this l. right that's what we need i think it would make us other compilers for your macros on your project
00:39:49
and though p. g. on top of it you would get this colour compiler that's faster than
00:39:53
hotspot and thoughts immediately that doesn't doing that's what we need but that's only seems holidays definition
00:39:58
of program we're using it directly says thank you there whoever did that to doing the hard
00:40:02
work yes i was hoping last year and i gave a presentation this will happen and see whether
00:40:08
fall amazing thanks
00:40:11
sure question new i didn't send the waist and
00:40:15
eating she slow here so then this easy she
00:40:19
it's just so not to mention active for yeah there's a there's a lot of organisations
00:40:25
in enterprise that that that uh not that and and they're not easy to do uh
00:40:29
it's like a yeah it's a lot of optimisation slight yes it it's
00:40:37
uh_huh
00:40:45
okay
Conference Program
A Tour of Scala 3
Martin Odersky, Professor EPFL, Co-founder Lightbend
June 11, 2019 · 5:15 p.m.
8335 views
A story of unification: from Apache Spark to MLflow
Reynold Xin, Databricks
June 12, 2019 · 9:15 a.m.
1267 views
In Types We Trust
Bill Venners, Artima, Inc
June 12, 2019 · 10:15 a.m.
1569 views
Creating Native iOS and Android Apps in Scala without tears
Zahari Dichev, Bullet.io
June 12, 2019 · 10:16 a.m.
2231 views
Techniques for Teaching Scala
Noel Welsh, Inner Product and Underscore
June 12, 2019 · 10:17 a.m.
1295 views
Future-proofing Scala: the TASTY intermediate representation
Guillaume Martres, student at EPFL
June 12, 2019 · 10:18 a.m.
1156 views
Metals: rich code editing for Scala in VS Code, Vim, Emacs and beyond
Ólafur Páll Geirsson, Scala Center
June 12, 2019 · 11:15 a.m.
4695 views
Akka Streams to the Extreme
Heiko Seeberger, independent consultant
June 12, 2019 · 11:16 a.m.
1552 views
Scala First: Lessons from 3 student generations
Bjorn Regnell, Lund Univ., Sweden.
June 12, 2019 · 11:17 a.m.
577 views
Cellular Automata: How to become an artist with a few lines
Maciej Gorywoda, Wire, Berlin
June 12, 2019 · 11:18 a.m.
386 views
Why Netflix ❤'s Scala for Machine Learning
Jeremy Smith & Aish, Netflix
June 12, 2019 · 12:15 p.m.
5026 views
Massively Parallel Distributed Scala Compilation... And You!
Stu Hood, Twitter
June 12, 2019 · 12:16 p.m.
958 views
Polymorphism in Scala
Petra Bierleutgeb
June 12, 2019 · 12:17 p.m.
1113 views
sbt core concepts
Eugene Yokota, Scala Team at Lightbend
June 12, 2019 · 12:18 p.m.
1655 views
Double your performance: Scala's missing optimizing compiler
Li Haoyi, author Ammonite, Mill, FastParse, uPickle, and many more.
June 12, 2019 · 2:30 p.m.
837 views
Making Our Future Better
Viktor Klang, Lightbend
June 12, 2019 · 2:31 p.m.
1682 views
Testing in the postapocalyptic future
Daniel Westheide, INNOQ
June 12, 2019 · 2:32 p.m.
498 views
Context Buddy: the tool that knows your code better than you
Krzysztof Romanowski, sphere.it conference
June 12, 2019 · 2:33 p.m.
393 views
The Shape(less) of Type Class Derivation in Scala 3
Miles Sabin, Underscore Consulting
June 12, 2019 · 3:30 p.m.
2321 views
Refactor all the things!
Daniela Sfregola, organizer of the London Scala User Group meetup
June 12, 2019 · 3:31 p.m.
514 views
Integrating Developer Experiences - Build Server Protocol
Justin Kaeser, IntelliJ Scala
June 12, 2019 · 3:32 p.m.
551 views
Managing an Akka Cluster on Kubernetes
Markus Jura, MOIA
June 12, 2019 · 3:33 p.m.
735 views
Serverless Scala - Functions as SuperDuperMicroServices
Josh Suereth, Donna Malayeri & James Ward, Author of Scala In Depth; Google ; Google
June 12, 2019 · 4:45 p.m.
936 views
How are we going to migrate to Scala 3.0, aka Dotty?
Lukas Rytz, Lightbend
June 12, 2019 · 4:46 p.m.
709 views
Concurrent programming in 2019: Akka, Monix or ZIO?
Adam Warski, co-founders of SoftwareMill
June 12, 2019 · 4:47 p.m.
1974 views
ScalaJS and Typescript: an unlikely romance
Jeremy Hughes, Lightbend
June 12, 2019 · 4:48 p.m.
1376 views
Pure Functional Database Programming‚ without JDBC
Rob Norris
June 12, 2019 · 5:45 p.m.
6374 views
Why you need to be reviewing open source code
Gris Cuevas Zambrano & Holden Karau, Google Cloud;
June 12, 2019 · 5:46 p.m.
484 views
Develop seamless web services with Mu
Oli Makhasoeva, 47 Degrees
June 12, 2019 · 5:47 p.m.
785 views
Implementing the Scala 2.13 collections
Stefan Zeiger, Lightbend
June 12, 2019 · 5:48 p.m.
810 views
Introduction to day 2
June 13, 2019 · 9:10 a.m.
250 views
Sustaining open source digital infrastructure
Bogdan Vasilescu, Assistant Professor at Carnegie Mellon University's School of Computer Science, USA
June 13, 2019 · 9:16 a.m.
374 views
Building a Better Scala Community
Kelley Robinson, Developer Evangelist at Twilio
June 13, 2019 · 10:15 a.m.
245 views
Run Scala Faster with GraalVM on any Platform
Vojin Jovanovic, Oracle
June 13, 2019 · 10:16 a.m.
1340 views
ScalaClean - full program static analysis at scale
Rory Graves
June 13, 2019 · 10:17 a.m.
463 views
Flare & Lantern: Accelerators for Spark and Deep Learning
Tiark Rompf, Assistant Professor at Purdue University
June 13, 2019 · 10:18 a.m.
380 views
Metaprogramming in Dotty
Nicolas Stucki, Ph.D. student at LAMP
June 13, 2019 · 11:15 a.m.
1250 views
Fast, Simple Concurrency with Scala Native
Richard Whaling, data engineer based in Chicago
June 13, 2019 · 11:16 a.m.
624 views
Pick your number type with Spire
Denis Rosset, postdoctoral researcher at Perimeter Institute
June 13, 2019 · 11:17 a.m.
245 views
Scala.js and WebAssembly, a tale of the dangers of the sea
Sébastien Doeraene, Executive director of the Scala Center
June 13, 2019 · 11:18 a.m.
661 views
Performance tuning Twitter services with Graal and ML
Chris Thalinger, Twitter
June 13, 2019 · 12:15 p.m.
2003 views
Supporting the Scala Ecosystem: Stories from the Line
Justin Pihony, Lightbend
June 13, 2019 · 12:16 p.m.
163 views
Compiling to preserve our privacy
Manohar Jonnalagedda and Jakob Odersky, Inpher
June 13, 2019 · 12:17 p.m.
301 views
Building Scala with Bazel
Natan Silnitsky, wix.com
June 13, 2019 · 12:18 p.m.
565 views
Diversity and Inclusion Panel: Bring the Thunder!
Panel
June 13, 2019 · 2:30 p.m.
244 views
Asynchronous streams in direct style with and without macros
Philipp Haller, KTH Royal Institute of Technology in Stockholm
June 13, 2019 · 3:45 p.m.
304 views
Interactive Computing with Jupyter and Almond
Sören Brunk, USU Software AG
June 13, 2019 · 3:46 p.m.
681 views
Scala best practices I wish someone'd told me about
Nicolas Rinaudo, CTO of Besedo
June 13, 2019 · 3:47 p.m.
2702 views
High performance Privacy By Design using Matryoshka & Spark
Wiem Zine El Abidine and Olivier Girardot, Scala Backend Developer at MOIA / co-founder of Lateral Thoughts
June 13, 2019 · 3:48 p.m.
753 views
Immutable Sequential Maps – Keeping order while hashed
Odd Möller
June 13, 2019 · 4:45 p.m.
276 views
All the fancy things flexible dependency management can do
Alexandre Archambault, engineer at the Scala Center
June 13, 2019 · 4:46 p.m.
389 views
ScalaWebTest - integration testing made easy
Dani Rey, Unic AG
June 13, 2019 · 4:47 p.m.
468 views
Mellite: An Integrated Development Environment for Sound
Hanns Holger Rutz, Institute of Electronic Music and Acoustics (IEM), Graz
June 13, 2019 · 4:48 p.m.
213 views
Closing panel
Panel
June 13, 2019 · 5:54 p.m.
400 views