Transactional locking II
Citations
1,268 citations
934 citations
Cites background from "Transactional locking II"
...The largest read and write set sizes are found in labyrinth+, with values of 783 and 779 cache lines, respectively (24.5 KB and 24.3 KB, respectively)....
[...]
...TM microbenchmarks are typically composed of transactions that execute a few operations on a data structure like a hash table or red-black tree [12, 13, 25, 27, 34]....
[...]
...To benefit from the increasing number of cores per chip, application developers have to develop parallel programs and deal with cumbersome issues such as synchronization tradeoffs, deadlock avoidance, and races....
[...]
525 citations
Cites background from "Transactional locking II"
...2.3 Hybrid transactional memory Kumar et al. [12] recently proposed using HTM to optimize the Dynamic Software Transactional Memory (DSTM) of Herlihy et al....
[...]
...4.4 rand-array benchmark In this section, we report on our simulation studies, in which we used the simple rand-array microbenchmark to evaluate HyTM s ability to mix HTM and STM transactions, and to compare its performance in various con.gurations against standard lock-based approaches....
[...]
...The key idea to achieving correct interaction between software transactions (i.e., those executed using the STM library) and hardware transactions (i.e., those executed using HTM support) is to augment hardware transactions with additional code that ensures that the transaction does not commit if it con.icts with an ongoing software transaction....
[...]
...Apart from boosting the performance of HyTM, best-effort HTM also supports a number of other useful purposes, such as selectively eliding locks, optimizing nonblocking data structures, and optimizing the Dynamic Software Transactional Memory (DSTM) system of Herlihy et al....
[...]
...2 Independently, Harris and Fraser also developed an STM that uses a table of ownership records [8]....
[...]
509 citations
Cites result from "Transactional locking II"
...Recent STM systems, including TL2 [9], are based on optimistic concurrency control and maintain read and write sets similar to those in OCC databases....
[...]
500 citations
Cites background from "Transactional locking II"
...These strategies are usually implemented using the single-writer multiple-readers pattern, with either explicit locks (e.g., TL2) or “virtual”, revocable ones (e.g., obstruction-free TMs, such as DSTM, ASTM and SXM), sometimes with a multi-versioning scheme (e.g., LSA-STM and JVSTM) or specialized optimization strategies....
[...]
...This is confirmed by the already mentioned counterexample TM implementations that have the time complexity either constant or at least independent of k (e.g., RSTM, JVSTM, TL2, etc.)....
[...]
...Most transactional memory systems we know of ensure opacity, including DSTM [14], ASTM [18], SXM [13], JVSTM [5], TL2 [6], LSA-STM [25] and RSTM [19]....
[...]
...However, it is commonly argued that sandboxing is expensive and applicable only to specific run-time environments [6]....
[...]
...It is worth noting that TL2 has a constant time complexity, although it ensures opacity, uses invisible reads, and is singleversion....
[...]
References
7,390 citations
"Transactional locking II" refers methods in this paper
...The transactional load first checks (using a Bloom filter [24]) to see if the load address already appears in the write-set....
[...]
2,406 citations
"Transactional locking II" refers background in this paper
...The transactional memory programming paradigm of Herlihy and Moss [1] is gaining momentum as the approach of choice for replacing locks in concurrent programming....
[...]
1,068 citations
"Transactional locking II" refers methods in this paper
...We present here a set of microbenchmarks that have become standard in the community [25], comparing a sequential red-black tree made concurrent using various algorithms representing state-of-the-art non-blocking [6] and lock-based [5, 18] STMs....
[...]
880 citations
"Transactional locking II" refers methods in this paper
...STM design has come a long way since the first STM algorithm by Shavit and Touitou [12], which provided a non-blocking implementation of static transactions (see [5–8, 15, 9–13])....
[...]
771 citations
"Transactional locking II" refers background in this paper
...Providing large scale transactions in hardware tends to introduce large degrees of complexity into the design [1–4]....
[...]
...There are currently proposals for hardware implementations of transactional memory (HTM) [1–4], purely software based ones, i....
[...]