<headius>
sam2000: it might be good for us to have a closer look at your numbers...if you're seeing that much allocation after warmup, and forcing GC doesn't clear it all out, we might have something to fix
<headius>
jensnockert: nailgun is kinda deprecated at this point; the persistent process doesn't behave well after a few uses because resources don't get cleaned up (threads, IO, etc)
<headius>
TBPhd: we've managed to get a basic scaffolded app to do CRUD things but that's about it...we have work to do and plan to have it all good by RubyConf
<GitHub53>
jruby/master 0ca3184 Charles Oliver Nutter: Add a JIT spec for #4148, testing attr reader call with *args.
<GitHub32>
[jruby] headius closed issue #4148: Java::JavaLangInvoke::WrongMethodTypeException when using invokedynamic https://git.io/viucs
lance|afk is now known as lanceball
claudiuinberlin has quit []
zacts has joined #jruby
temporalfox has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
pilhuhn is now known as pil-afk
cprice has joined #jruby
cprice404 has quit [Ping timeout: 264 seconds]
cprice has quit [Client Quit]
<chrisseaton>
headius: have you thought about how to implement Process#groups on Java 9? It looks like com.sun.security.auth.module.UnixSystem isn't available. That's the main failure I see running Truffle on Java 9
pawnbox has quit [Remote host closed the connection]
<headius>
chrisseaton: nope, haven't thought about it
<headius>
I imagine we'd just do it with ffi
cprice has joined #jruby
<chrisseaton>
I might try that - we could both use the same code
nicksieg_ has joined #jruby
nicksieger has quit [Read error: Connection reset by peer]
cprice has quit [Quit: Konversation terminated!]
jensnockert has joined #jruby
nicksieg_ has quit [Read error: Connection reset by peer]
nicksieger has quit [Read error: Connection reset by peer]
cprice404 has joined #jruby
nicksieger has joined #jruby
zph_ has joined #jruby
nicksieger has quit [Read error: Connection reset by peer]
nicksieger has joined #jruby
lanceball is now known as lance|afk
Puffball has joined #jruby
camlow325 has quit [Ping timeout: 244 seconds]
camlow325 has joined #jruby
Puffball has quit [Remote host closed the connection]
emakris has quit [*.net *.split]
rtyler has quit [*.net *.split]
subbu has quit [*.net *.split]
rtyler has joined #jruby
emakris has joined #jruby
nicksieger has quit [Read error: Connection reset by peer]
nicksieger has joined #jruby
nicksieger has quit [Read error: Connection reset by peer]
nicksieger has joined #jruby
nicksieger has quit [Read error: Connection reset by peer]
nicksieger has joined #jruby
subbu has joined #jruby
nicksieger has quit [Remote host closed the connection]
Puffball has joined #jruby
Aethenelle has quit [Quit: Aethenelle]
Puffball has quit [Remote host closed the connection]
Puffball has joined #jruby
enebo has quit [Quit: enebo]
Puffball has quit [Ping timeout: 248 seconds]
bga57 has joined #jruby
<donValentin>
chrisseaton: Hi! Do you have any good resources benchmarking the memory usage of JRuby+Truffle vs the other Ruby implementations?
<donValentin>
chrisseaton: I am using Ruby on some devices with memory restrictions.
<chrisseaton>
We have tools for measuring minimum heap and total allocation. I don't have numbers off the top of my head.
<chrisseaton>
donValentin: ^
<donValentin>
chrisseaton: Thanks! Are there any benchmarks measuring memory usage on rubybench.org? There are so many benchmarks there, and I have no idea if any of them target memory usage.
<donValentin>
chrisseaton: Hey! I actually found it!
<donValentin>
chrisseaton: Looks like _all_ the benchmarks also measure RSS memory allocation.
<chrisseaton>
RSS is a questionable way to measure for Java though
<donValentin>
OK? What is better?
<chrisseaton>
I'm not sure any scalar metric is good - performance varies as available heap size
<chrisseaton>
You can talk about one without the other
<chrisseaton>
A Java a app will run in one heap size at a given speed, and slower in less heap
<chrisseaton>
The two things are functions of each other
prasunanand has quit [Quit: Leaving]
<donValentin>
chrisseaton: Yeah, I have devices with 512MB (ish) total memory, and total memory usage is critical. Heap is only part of the story.
<donValentin>
In this case setting lower heap size increases speed since non-heap memory triggers paging to disk.
<donValentin>
chrisseaton: Running JRuby 9K with the same setting as JRuby 1.7.x results in vastly higher total memory usage and GC pauses of many minutes.
<chrisseaton>
Truffle may not run well in 512 - I think we can safely say that
<chrisseaton>
It's probably not our target environment
<donValentin>
chrisseaton: OK, thanks! Very good to know?
<donValentin>
not a question :)
<chrisseaton>
Oh wait but SVM should run well there
<donValentin>
Cool.
<donValentin>
chrisseaton: Do you have an opionion (or measurements) on the memory usage of JRuby+Truffel vs JRuby divided into startup cost vs runtime cost? I realize it is a complex question :)
<donValentin>
sidenote: rubybench.org only measures MRI.
<chrisseaton>
Too complex to answer on my phone - ask me again later
<donValentin>
:)
<donValentin>
chrisseaton: Thanks for the info. Any news on SSL (and RubyGems) support?
<chrisseaton>
We ran a useful part of the openssl C extension successfully last week
<donValentin>
Sounds promising. :)
<chrisseaton>
donValentin: I can't really say anything definitive about memory usage. The idea we have is that a little more memory for startup and compilation can mean less memory for actual requests, because of things like scalar replacement (malloc removal)
<chrisseaton>
'GC pauses of many minutes' - do you mean literally?
<donValentin>
Yes. :(
<donValentin>
With tuning we are down to 4 seconds or less.
<donValentin>
That has actually been accepted by the customer since we are introducing new hardware with 4GB memory over the next year or so.
<donValentin>
Must sleep soon :)
<donValentin>
chrisseaton: Why are you still up?! Don’t you need to sleep?
<chrisseaton>
I'm in San Francisco this week
<donValentin>
Ah! Explains it :)
<donValentin>
Well, I need some zzzzzzz
<donValentin>
Bye all!
donValentin has quit [Quit: donValentin]
eregon has quit [Quit: No Ping reply in 180 seconds.]
eregon has joined #jruby
eregon has quit [Quit: No Ping reply in 180 seconds.]