Message Passing Concurrency in Clojure using Kilim

Kilim-Clojure integration

Message Passing Concurrency in Clojure using
Kilim

Antonio Garrote

Forward

November 7, 2011


Clojure’s Concurrency Model

JVM thread basic autonomous work unit
Java shared memory explicit communication primitives
Immutable data structures + implicit coordination
mechanisms (STM, Refs)


Message Passing Communication

1 % T e r m i t e Scheme b e n c h m a r k s r i n g . e r l
2 m a k e r e l a y ( Next ) − >
3 receive
4 K when K > 0 − >
5 Next ! K − 1 ,
6 m a k e r e l a y ( Next ) ;
7 K− >
8 Next ! K
9 end .
10
11 l o o p (K , C u r r e n t , N) when N > 1 − >
12 l o o p (K , spawn ( r i n g , m a k e r e l a y , [ C u r r e n t ] ) ,N − 1 ) ;
13
14 l o o p (K , C u r r e n t , ) −>
15 s e l f ( ) ! K,
16 make relay ( Current ) .
17
18 % N = number p r o c e s s e s
19 % K = number o f m e s s a g e e x c h a n g e s
20 r i n g (N, K) − >
21 l o o p (K , s e l f ( ) , N ) .



1
2 N
N = 100K , K = 0
518ms, 5µs process creation
3 K
N = 100K , K = 1M
759ms, 0.76µs local msg . send.
4 6
5



How does Erlang do it?

User level processes executing recursive functions
VM preemptive scheduling
One heap/stack per process, no global heap
Tail call optimisation
More predictable GC behaviour


Eﬃcient implementation of message passing concurrency in the
Java Virtual Machine?


Kilim

“Kilim is a message-passing framework for Java that provides
ultra-lightweight threads and facilities for fast, safe, zero-copy
messaging between these threads”

https://github.com/kilim/kilim
2006 Sriram Srinivasan, Opera Group Cambridge University
Current version 0.72


Kilim Tasks

Task

1 p u b l i c c l a s s ExampleTask e x t e n d s Task {
2 execute
3 p u b l i c void execute () throws Pausable {
4 doStuff ( ) ;
5 p u s a b l e S t u f f ( ) ; // t h r o w s P a u s a b l e Worker
6 moreStuff ( ) ; Thread
7 }
8
9 } notify

Scheduler


Kilim Weaving: Bytecode Transformation

Additional step after compilation
Transforms pausable methods
Creates a custom class to store the task state
Associates a Fiber object, tracks task’s stack
Transformed code returns immediately if the task is paused
Worker threads can resume the execution of paused stacks
unwinding the associated ﬁber stack


Kilim Weaving: Bytecode Transformation
1 p u b l i c void execute ( Fiber f i b e r ) throws Pausable {
2 s w i t c h ( f i b e r . pc ) {
3 c a s e 0 : g o t o START ;
4 c a s e 1 : g o t o PAUSABLE CALL ;
5 }
6 START :
7 doStuff ( ) ;
8 PAUSABLE CALL :
9 f i b e r . down ( ) ;
10 p u s a b l e S t u f f ( f i b e r ) ; // t h r o w s P a u s a b l e
11 f i b e r . up ( ) ;
12 switch ( f i b e r . status ) {
13 c a s e NOT PAUSING NO STATE :
14 g o t o RESUME ;
15 c a s e NOT PAUSING HAS STATE :
16 restoresState ( state );
17 g o t o RESUME ;
18 c a s e PAUSING NO STATE :
19 captureState ( state );
20 return ;
21 c a s e PAUSING HAS STATE :
22 return ;
23 }
24 RESUME :
25 moreStuff ( ) ;
26 }


Kilim Tasks: Execution

Fiber.stack

currentState StateObj.(1)
Pausable()
StateObj.(2)
currentState
down()
StateObj.(3)

Fiber Pausable()

currentState
down()

Fiber Pausable()

up()

Pause Normal Return


Kilim Messages

Buﬀered mailboxes
Allows asynchronous communication between tasks
Blocking / not blocking interface
Polling interface


Kilim-Clojure Integration (ﬁrst attempt)

Modiﬁed version of Clojure’s compiler
Clojure’s functions executed as Kilim’s tasks
Weaving performed at run-time

https://github.com/antoniogarrote/clojure/tree/kilim


Kilim-Clojure Integration (second attempt)

Clojure library
[clj-kilim ”0.2.0-SNAPSHOT”]
Doc: http://antoniogarrote.github.com/clojure kilim/
Source: https://github.com/antoniogarrote/clojure kilim


Introduction

Clojure
Compiler

Compiles bytecode

Java Class

Java instrumentation

Kilim Weaved Java
clj-kilim
Weaver Class

Bytecode Transformation Bytecode Transformation

Class
Loader


Kilim-Clojure Integration: API

1 ( d e f − p a u s a b l e y i e l d i n g − f n [ mbox ]
2 ( mbox−get−pausable mbox ) )
3
4
5 ( def−pausable task−fn1 [ ]
6 ...
7 ( c a l l − p a u s a b l e y i e l d i n g − f n ambox )
8 ...)
9
10 ( d e f task−fn2 ( p a u s a b l e [] ...))
11
12
13 ( t a s k − s t a r t task−fn1 )
14 ( t a s k − s t a r t task−fn2 )


Kilim-Clojure Integration

Demo


Performance


Ring example again
1 ( d e f n ˆ { : p a u s a b l e t r u e } make−relay [ ˆ k i l i m . M a i l b o x s e l f
2 ˆ k i l i m . Mailbox next
3 ˆ k i l i m . Mailbox f i n a l ]
4 ( loop [ k ( r e c e i v e s e l f ) ]
5 ( i f (> k 0 )
6 ( do
7 ( ! next ( dec k ) )
8 ( recur ( . get s e l f )))
9 ( do
10 ( ! next ( dec k ) )
11 ( when ( n o t ( n i l ? f i n a l ) )
12 (! f i n a l true ))))))
13
14
15 ( d e f n make−loop [ n k ]
16 ( l e t [ˆ k i l i m . Mailbox f i r s t − m a i l b o x ( k i l i m . Mailbox . )
17 ˆ k i l i m . Mailbox final−mailbox ( k i l i m . Mailbox . ) ]
18 ( loop [ current first−mailbox
19 n n]
20 ( i f (> n 1 )
21 ( recur
22 ( spawn ( p a u s a b l e [ ˆ k i l i m . M a i l b o x s e l f ]
23 ( make−relay s e l f c u r r e n t n i l ) ) )
24 ( dec n ) )
25 ( do ( spawn ( p a u s a b l e [ ˆ k i l i m . M a i l b o x ]
26 ( make−relay f i r s t − m a i l b o x c u r r e n t f i n a l − m a i l b o x ) ) )
27 (! first−mailbox k)
28 ( . getb final−mailbox ) ) ) ) ) )


Performance

N = 100K , K = 0
1264ms, 12µs process creation

N = 100K , K = 1M
3276ms, 3µs local msg . send.


Conclusions

Eﬃcient message passing possible in the JVM thanks to Kilim
Clojure can be a nice match for Kilim (immutable messages,
run-time weaving)
Beneﬁts of asynchronous evented code with a nicer interface
Limitations: non-preemptive
Way to go?
Improve performance
Add distribution

Message Passing Concurrency in Clojure using Kilim

Recommended

More Related Content

Similar to Message Passing Concurrency in Clojure using Kilim (20)

More from Antonio Garrote Hernández (7)

Recently uploaded (20)

Message Passing Concurrency in Clojure using Kilim