![: Built on Truffle
11
cnt
1
+
cnt:
=
if
cnt:
=
0
cnt
1
+
cnt:
=if cnt:
=
0
NS
Truffle’s Self-Optimization Approach:
[1] Würthinger, T.; Wöß, A.; Stadler, L.; Duboscq, G.; Simon, D. & Wimmer, C. (2012),
Self-Optimizing AST Interpreters, Proceedings of the 8th Dynamic Languages Symposium.
JIT Compiled
Native Code
Self-Optimized AST](https://image.slidesharecdn.com/opt-cel-languages-151101183300-lva1-app6891/85/Optimizing-Communicating-Event-Loop-Languages-with-Truffle-11-320.jpg)
![vs. JVM Actor Libraries
13
NS
Chameneos
ConcurrentSorted
LinkedList
Counting ForkJoinThroughput PingPong RadixSort ThreadRing Geo Mean
0
1
2
3
4
5
6
7
8
RuntimefactoroverScalaz
lowerisbetter
Akka Jetlang SOMns Scalaz
RuntimeFactor
NormalizedtoScalaz
Lower
Is
Better
Savina Benchmarks
[2] Imam, S. M. & Sarkar, V. (2014), Savina - An Actor Benchmark
Suite: Enabling Empirical Evaluation of Actor Libraries, Proceedings of
the 4th AGERE! Workshop, ACM.](https://image.slidesharecdn.com/opt-cel-languages-151101183300-lva1-app6891/85/Optimizing-Communicating-Event-Loop-Languages-with-Truffle-13-320.jpg)
![Enforcing Isolation
16
if (isMutableObject(arg[i])) {
return farReference(arg[i]);
} else if (isValueObject(arg[i]) {
return arg[i];
} else if (isFarReference(arg[i]) && toCurrentActor(arg[i])) {
...
} else if (isFarReference...
} else if (isPromise(arg[i])...
...
public traverse: t col: start to: end = (
(* ... *)
)
worker <-: traverse: table col: 1 to: 10](https://image.slidesharecdn.com/opt-cel-languages-151101183300-lva1-app6891/85/Optimizing-Communicating-Event-Loop-Languages-with-Truffle-16-320.jpg)
![Impact on Microbenchmarks
25
Speedup Factor over Unoptimized Version
public count = (
cnt := cnt + 1.
cnt = iterations
ifTrue: [ completionPP resolve: cnt ]
ifFalse: [ self <-: count ]
)
lookup in cls
lookup in
5th supercls
1.0 1.5 2.0 2.5 3.0
Benchmark](https://image.slidesharecdn.com/opt-cel-languages-151101183300-lva1-app6891/85/Optimizing-Communicating-Event-Loop-Languages-with-Truffle-25-320.jpg)
![Impact on Microbenchmarks
26
Speedup Factor over Unoptimized Version
public calc: a and: b = (
| r |
r := a * b + b + b – a.
r := r - (a * a * b).
^ r
)
lookup in cls
lookup in
5th supercls
with
int or double
1.0 1.5 2.0 3.0 10.0 50.0 100.0
Benchmark
lookup in cls
lookup in
5th supercls
with
int or double
1.0 1.5 2.0 3.0 10.0 50.0 100.0
Benchmark
public benchmark = (
1 to: numIter do: [:i |
self <-: calc: 2 and: 4.
self <-: calc: 1.2 and: 3.3.
].
)](https://image.slidesharecdn.com/opt-cel-languages-151101183300-lva1-app6891/85/Optimizing-Communicating-Event-Loop-Languages-with-Truffle-26-320.jpg)
Optimizing Communicating Event-Loop Languages with Truffle
- 1. Stefan Marr, Hanspeter Mössenböck AGERE! Workshop October 26, 2015 Optimizing Communicating Event-Loop Languages with Truffle Research funded by NS
- 2. Initial Goals • Safety – Guaranteed Isolation – No Low-Level Data Races • Deadlock Freedom • Performance Competitive with Java 2 NS: A Platform For Concurrency Research
- 3. Communicating Event Loops 3 E Programming Language à la
- 4. Communicating Event Loops 4 Actor A Actor B Actor Principle
- 5. Communicating Event Loops 5 Actor A Actor B But, Actor Not First-Class
- 6. Communicating Event Loops 6 Actor A Actor B Actors Contain Objects
- 7. Communicating Event Loops 7 Actor A Actor B Objects Can Have Far-References
- 8. public class PingPong new = Benchmark <: Value ( class Ping new: cnt with: pongAct = ( private pingsLeft ::= cnt. (* mutable slot *) private pongAct = pongAct. (* immutable slot *) )( 8 Newspeak A Class-based Language Dynamically Typed No Global/Static State
- 9. public class PingPong new = Benchmark <: Value ( class Ping new: cnt with: pongAct = ( private pingsLeft ::= cnt. (* mutable slot *) private pongAct = pongAct. (* immutable slot *) )( 9 Newspeak public ping = ( pongAct <-: ping: self. pingsLeft := pingsLeft - 1. ) Communicating Event-Loop Actors
- 10. public class PingPong new = Benchmark <: Value ( class Ping new: cnt with: pongAct = ( private pingsLeft ::= cnt. (* mutable slot *) private pongAct = pongAct. (* immutable slot *) )( 10 Newspeak Newspeak Programming Language Draft Specification Version 0.095 http://bracha.org/newspeak-spec.pdf With Spec: public ping = ( pongAct <-: ping: self. pingsLeft := pingsLeft - 1. )
- 11. : Built on Truffle 11 cnt 1 + cnt: = if cnt: = 0 cnt 1 + cnt: =if cnt: = 0 NS Truffle’s Self-Optimization Approach: [1] Würthinger, T.; Wöß, A.; Stadler, L.; Duboscq, G.; Simon, D. & Wimmer, C. (2012), Self-Optimizing AST Interpreters, Proceedings of the 8th Dynamic Languages Symposium. JIT Compiled Native Code Self-Optimized AST
- 12. : A Fast Newspeak 12 NS SOMNS versus Java (Graal Compiler) On average 1.65x slower (min. −3%, max. 2.6x) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Bounce BubbleSort DeltaBlue Fannkuch Json Mandelbrot NBody PageRank Permute Queens QuickSort Richards Sieve Storage Runtimefactornormalized toJava RuntimeFactor NormalizedtoJava Lower Is Better
- 13. vs. JVM Actor Libraries 13 NS Chameneos ConcurrentSorted LinkedList Counting ForkJoinThroughput PingPong RadixSort ThreadRing Geo Mean 0 1 2 3 4 5 6 7 8 RuntimefactoroverScalaz lowerisbetter Akka Jetlang SOMns Scalaz RuntimeFactor NormalizedtoScalaz Lower Is Better Savina Benchmarks [2] Imam, S. M. & Sarkar, V. (2014), Savina - An Actor Benchmark Suite: Enabling Empirical Evaluation of Actor Libraries, Proceedings of the 4th AGERE! Workshop, ACM.
- 14. TWO OPTIMIZATIONS Enforcing Isolation Asynchronous Sends 14
- 16. Enforcing Isolation 16 if (isMutableObject(arg[i])) { return farReference(arg[i]); } else if (isValueObject(arg[i]) { return arg[i]; } else if (isFarReference(arg[i]) && toCurrentActor(arg[i])) { ... } else if (isFarReference... } else if (isPromise(arg[i])... ... public traverse: t col: start to: end = ( (* ... *) ) worker <-: traverse: table col: 1 to: 10
- 17. Optimistic AST Specialization 17 async send WrapArg WrapArg WrapArgWrapArg ReadVar worker ReadVar table worker <-: traverse: table col: 1 to: 10 Literal 1 Literal 10
- 18. Optimistic AST Specialization 18 async send WrapArg WrapArg WrapArg Unwrap FarRef ReadVar worker ReadVar table worker <-: traverse: table col: 1 to: 10 Literal 1 Literal 10
- 19. Optimistic AST Specialization 19 async send IsValue WrapArg WrapArg Unwrap FarRef ReadVar worker ReadVar table worker <-: traverse: table col: 1 to: 10 Literal 1 Literal 10
- 20. Optimistic AST Specialization 20 async send IsValue Unwrap FarRef ReadVar worker ReadVar table worker <-: traverse: table col: 1 to: 10 Literal 1 Literal 10
- 21. Impact on Microbenchmarks 21 Speedup Factor over Unoptimized Version public class With10Args new = Benchmark ( private aValue = Value new. private obj ::= Object new. public benchmark = ( self <-: a1: aValue a2: Object new a3: obj a4: Benchmark a5: aValue a6: 0 a7: 7 a8: 8 a9: #eee a0: '33'. ) ) with 10 arguments 1.0 1.5 2.0 2.5 3.0 Benchmark
- 22. Method Lookup for Asynchronous Sends 22 A1 Event-Loop: Single Point of Reception B A2 C do get set do a1 <-: do b <-: get a2 <-: set c <-: do Megamorphic Method Invocations while (true) { msg = mailbox.receive() mthd = msg.obj.getClass(). lookup(msg.selector()) mthd.invoke(obj, msg.args) }
- 23. Optimization: Send-site Caching 23 a1 <-: do actor(a1) <-: fun(o) { o.do() } Code Transformation Introduces Inline Cache
- 24. With Send-site Caching 24 A1 B A2 C do get set do a1 <-: do b <-: get a2 <-: set c <-: do while (true) { msg = mailbox.receive() msg.fun. invoke(obj, msg.args) }fun1(.) {…} fun2(.) {…} fun3(.) {…} fun4(.) {…}
- 25. Impact on Microbenchmarks 25 Speedup Factor over Unoptimized Version public count = ( cnt := cnt + 1. cnt = iterations ifTrue: [ completionPP resolve: cnt ] ifFalse: [ self <-: count ] ) lookup in cls lookup in 5th supercls 1.0 1.5 2.0 2.5 3.0 Benchmark
- 26. Impact on Microbenchmarks 26 Speedup Factor over Unoptimized Version public calc: a and: b = ( | r | r := a * b + b + b – a. r := r - (a * a * b). ^ r ) lookup in cls lookup in 5th supercls with int or double 1.0 1.5 2.0 3.0 10.0 50.0 100.0 Benchmark lookup in cls lookup in 5th supercls with int or double 1.0 1.5 2.0 3.0 10.0 50.0 100.0 Benchmark public benchmark = ( 1 to: numIter do: [:i | self <-: calc: 2 and: 4. self <-: calc: 1.2 and: 3.3. ]. )
- 27. : Fast And Scalable • Platform for Concurrency Research • Initial optimizations – Send-site Caching + Isolation 27 NS Chameneos ConcurrentSorted LinkedList Counting ForkJoinThroughput PingPong RadixSort ThreadRing Geo Mean 0 1 2 3 4 5 6 7 8 RuntimefactoroverScalaz lowerisbetter Akka Jetlang SOMns ScalazLower Is Better RuntimeFactor NormalizedtoScalaz