Inherent Limitations of Hybrid Transactional Memory

Abstract

International audienceSeveralHybridTransactionalMemory(HyTM)schemeshave recently been proposed to complement the fast, but best-effort nature of Hardware Transactional Memory (HTM) with a slow, reliable software backup. However, the costs of providing concurrency between hardware and software transactions in HyTM are still not well understood.In this paper, we propose a general model for HyTM implementations, which captures the ability of hardware transactions to buffer memory accesses. The model allows us to formally quantify and analyze the amount of overhead (instrumentation) caused by the potential presence of software transactions. We prove that (1) it is impossible to build a strictly serializable HyTM implementation that has both uninstrumented reads and writes, even for very weak progress guarantees, and (2) the instrumentation cost incurred by a hardware transaction in any progressive opaque HyTM is linear in the size of the transaction’s data set. We further describe two implementations which exhibit optimal instrumentation costs for two different progress conditions. In sum, this paper proposes the first formal HyTM model and captures for the first time the trade-off between the degree of hardware-software TM concurrency and the amount of instrumentation overhead

The Case for Phase-Based Transactional Memory

“…Besides that, every SW commit aborts all HW transactions due to the eager subscription to the sequence lock. The aforementioned characteristics are inherent limitations of conventional hybrid systems, as also discussed by Shavit et al [17].…”

“…In particular, PhTM* performs on average 1.64x better than PhTM, 2.08x better than HyTM-NOrec and 2.28x better tahn HyCO on the Broadwell machine, and 1.18x, 1.36x and 1.81x on the Power8 machine, respectively (Sections 4.2 and 4.3); It shows that STAMP applications do not exhibit hybrid behavior to justify the use of conventional hybrid systems (HyTM-NOrec [8] and HyCO [13]), therefore making PhTM* a better suited solution. And, for the first time, it shows that conventional hybrid systems do not perform better than phasedbased system in a scenario with hybrid-behaved transactions (Section 4.4); It presents the first experimental results showing the inherent limitations of conventional hybrid systems, as demonstrated by Shavit et al [17] (Section 4.5); It discusses the virtues and limitations of phasebased transactional systems and reports our experience when designing PhTM* (Section 4.6). This article is divided as follows.…”

On the Cost of Concurrency in Hybrid Transactional Memory

“…Instrumentation-optimal progressive HyTM. We describe a HyTM algorithm that is a tight bound for Theorem 3 and the instrumentation cost on the fast-path transactions established in [3]. Pseudocode appears in Algorithm 1.…”

“…In [3], it was shown that hardware transactions in opaque progressive HyTMs must perform at least one metadata access per transactional read and write. In this paper, we show that in opaque progressive HyTMs with invisible reads, software transactions cannot avoid incremental validation.…”

“…The rest of the paper is organized as follows. S 2 presents details of the HyTM model that extends the model introduced in [3]. S 3 presents our main lower bound result on the step-complexity of slow-path transactions in progressive HyTMs while S 4 presents opaque HyTMs that are progressive for a subset of transactions.…”