On 15/09/2015 22:34, Caldarale, Charles R wrote: >> From: Mark Thomas [mailto:ma...@apache.org] >> Subject: Re: RV-Predict bugs > >> st_200 is non-volatile > >> L1 if (st_200 == null ) { >> L2 st_200 = sm.getString("sc.200"); >> L3 } >> L4 return st_200; > >> Ln=Line n >> Tx=Thread x >> Rn=Read at line n >> Wn=Write at line n > >> So T2 has two reads, T2R1 and T2R4. > >> Depending on the value read for T2R1 there may be a write at T2W2. >> If there is a write there is a happens before relationship between T2R1 >> and T2R4. > > There is also a happens-before relationship between T2R1 and T2R4 due to JLS > Section 14.2 (Blocks), not Section 17.4 (Memory Model). The JMM cannot be > considered in a vacuum. > >> Consider the following sequence >> T2R4 (out of order read returns null) >> T1R1 (returns null) >> T1W2 (writes non-null value) >> T1R4 (reads new non-null value) >> T2R1 (reads new non-null value) > > The above is not permitted due to Section 14.2 (although it would be by > 17.4). The compiler must not move T2R4 in front of T2R1 because of the > potential of T2W2 occurring. Correct operation is dependent on the CPU > observing the same value for T2R1 and T2R4, which is implemented in all > current hardware I'm aware of by not permitting reordering of accesses to the > same memory location (e.g., see Chapter 8 of the Intel 64 Architecture SDM, > Volume 3 - System Programming Guide). Such accesses are usually just > combined when they occur. > >> Because T2R1 reads a non-null value there is no write in T2. > >> Therefore there is no happens-before relationship between T2R1 and T2R4 >> because there is no intervening write in that thread (the write happens >> in T1). >> Therefore the re-ordering is allowed to happen. >> And we get the unexpected result. > > If the JMM were the only controlling specification, that would be true. > Since the JVM must also conform to Section 14, the happens-before order > exists, and we don't have a problem. > >> The write at T1W2 is sufficient to enforce the happens before >> relationship between T2R1 and T2R4. > > Yes, the potential of the write occurring is enough.
To make sure I understand this it is the possibility of the write T2W2 that is sufficient to ensure T2R1 and T2R4 are not re-ordered not the possibility of a write in some other thread? Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
