1. Memory Models (1) Xinyu Feng University of Science and Technology of China

2. Why Memory Models C1 || C2Compiler Memory Result

3. Why Memory Models C1 || C2Compiler Memory Result Which reads see which write?

4. Sequential Consistent (SC) Model... Memory Interleaving semantics: P1 P2P3Pn [Lamport 1979]

5. The need of weak memory models SC model prohibits many optimization: x = 1; r1 = y; y = 1; r2 = x; r1 = r2 = 0? Initially: x = y = 0 Impossible in SC model, but allowed in x86 or Java. r1 = y; x = 1 y = 1; r2 = x; Weak memory model allow more behaviors.

6. Usability: DRF guarantee – DRF programs have the same behaviors as in SC model Not too strong – Allow common optimization techniques – In some sense hijacked by the mainstream compiler Preserve type-safety and security guarantee – Cannot be too weak Design Criteria Very challenging to satisfy all the requirements!

7. Compiler Optimization Can Be Smart r1 = x; r2 = x; if (r1 == r2) y = 2; r3 = y; x = r3; Initially: x = 0, y = 1 r1 = r2 = r3 = 2? Must be allowed! y = 2; r1 = x; r2 = r1; if (true) Redundant read elim.

8. Efforts for Java Memory Model (JMM) First edition in Java Language Spec – Fatally flawed, not support key optimizations [Pough 2000] Current JMM [Manson et al. 2005] – Based on 5-year discussion and many failed proposals – “very complex” [Adve & Boehm 2010] – Surprising behaviors and bugs [Aspinall & Sevcik 2007] Next generation: JEP 188, Dec. 2013

9. Happens-Before Order Program execution: a set of events, and some orders between them. po: Time T1 T2 rel acqrel acq sw: Happens-before order (hb): transitive closure of po  sw [Lamport 1978]

10. Happens-Before Order Time T1 T2 po: rel acqrel acq sw: Happens-before order (hb): transitive closure of po  sw w1 w2 r w1 hb w2w1 hb r Not: w2 hb rr w2

11. Happens-Before Memory Model (HMM) Time T1 T2 rel acqrel acq w1 w2 r Read can see (1) the most recent write that happens-before it, or (2) a write that has no happens-before relation. r could see both w1 (which happens-before it) and w2 (with which there is no happens-before relation)

12. HMM – Relaxed Ordering x = 1; r1 = y; y = 1; r2 = x; r1 = r2 = 0? Initially: x = y = 0 Allowed in HMM

13. HMM – Examples with Global Analysis r1 = x; r2 = x; if (r1 == r2) y = 2; r3 = y; x = r3; Initially: x = 0, y = 1 r1 = r2 = r3 = 2? Speculation: r3 reads 2 Allowed in HMM! Justified!

14. HMM – Out-of-Thin-Air Read r1 = x; y = r1; r2 = y; x = r2; r1 = r2 = 42? Initially: x = y = 0 Allowed in HMM! Speculation: r1 will get 42 Justified! May break the security and type-safety of Java!

15. r1 = x; r2 = x; if (r1 == r2) y = 2; r3 = y; x = r3; r1 = r2 = r3 = 2? Speculation: r3 reads 2 r1 = x; y = r1; r2 = y; x = r2; r1 = r2 = 42? Speculation: r1 will get 42 Good speculation. Should allow! Bad speculation. Disallow!

16. JMM Take HMM as the core, and try hard to distinguish good speculation from bad speculation! Very complex, with surprising results and bugs. Introduce 9 axioms to constrain causality.

17. JMM – Surprising Results C1; lock l; C2; unlock l; C3 lock l; C1; C2; unlock l; C3 lock l; C1; unlock l; lock l; C2; unlock l; C3 Adding more synchronization may increase behaviors!

18. JMM – Surprising Results (2) C1; C3 C2; C1; C2; C3; Inlining threads may increase behaviors! More: Re-ordering independent operations may change behaviors. Adding/removing redundant reads may change behaviors.

19. Reading http://www.cs.umd.edu/~pugh/java/memoryModel/ http://openjdk.java.net/jeps/188