Speculative Storage Research Articles

Exploiting reference idempotency to reduce speculative storage overflow

Recent proposals for multithreaded architectures employ speculative execution to allow threads with unknown dependences to execute speculatively in parallel. The architectures use hardware speculative storage to buffer speculative data, track data dependences and correct incorrect executions through roll-backs. Because all memory references access the speculative storage, current proposals implement speculative storage using small memory structures to achieve fast access. The limited capacity of the speculative storage causes considerable performance loss due to speculative storage overflow whenever a thread's speculative state exceeds the speculative storage capacity. Larger threads exacerbate the overflow problem but are preferable to smaller threads, as larger threads uncover more parallelism.In this article, we discover a new program property called memory reference idempotency . Idempotent references are guaranteed to be eventually corrected, though the references may be temporarily incorrect in the process of speculation. Therefore, idempotent references, even from nonparallelizable program sections, need not be tracked in the speculative storage, and instead can directly access nonspeculative storage (i.e., conventional memory hierarchy). Thus, we reduce the demand for speculative storage space in large threads. We define a formal framework for reference idempotency and present a novel compiler-assisted speculative execution model. We prove the necessary and sufficient conditions for reference idempotency using our model. We present a compiler algorithm to label idempotent memory references for the hardware. Experimental results show that for our benchmarks, over 60% of the references in nonparallelizable program sections are idempotent.

Price, Production, Storage and Futures Markets

We model seasonal, uncertain production of a commodity, with speculative storage. We allow agents to be risk averse, and we allow planned production to respond to price prospects. We also explicitly consider the presence or absence of a futures market in the commodity. Our technique involves a two phase extension of the usual dynamic program, to solve for the equilibrium price and inventory held by storers. In our extra phase, we numerically solve the First Order Conditions for the planned production and futures position taken by the producer. Our main conclusions are: First, we confirm that storage reduces the price volatility, but the availability of futures is important in making storage feasible. Second, the producer takes a longer futures position, in many cases positive (i.e. he hedges by going long the future), than he would in the absence of storage. Third, futures increase average production, but this effect is generally less when storage is available, and might be quite marginal in cases when the producer's optimal futures position is near zero.

Reference idempotency analysis

Recent proposals for multithreaded architectures allow threads with unknown dependences to execute speculatively in parallel. These architectures use hardware speculative storage to buffer uncertain data, track data dependences and roll back incorrect executions. Because all memory references access the speculative storage, current proposals implement this storage using small memory structures for fast access. The limited capacity of the speculative storage causes considerable performance loss due to speculative storage overflow whenever a thread's speculative state exceeds the storage capacity. Larger threads exacerbate the overflow problem but are preferable to smaller threads, as larger threads uncover more parallelism. In this paper, we discover a new program property called memory reference idempotency . Idempotent references need not be tracked in the speculative storage, and instead can directly access non-speculative storage (i.e., the conventional memory hierarchy). Thus, we reduce the demand for speculative storage space. We define a formal framework for reference idempotency and present a novel compiler-assisted speculative execution model. We prove the necessary and sufficient conditions for reference idempotency using our model. We present a compiler algorithm to label idempotent memory references for the hardware. Experimental results show that for our benchmarks, over 60% of the references in non-parallelizable program sections are idempotent.

Estimating the rational expectations model of speculative storage: A Monte Carlo comparison of three simulation estimators

The non-negativity constraint on inventories imposed on the rational expectations theory of speculative storage implies that the conditional mean and variance of commodity prices are non-linear in lagged prices and have a kink at a threshold point. In this paper, the structural parameters of this model are estimated using three simulation-based estimators. In a Monte Carlo experiment, the finite sample properties of the simulated methods of moments estimator of Duffie and Singleton (1993, Econometrica 61 (4), 929–952) the indirect inference estimator of Gourieroux et al. (1993, Journal of Applied Economterics 8, S85–S118) and the efficient method of moments estimator of Gallant and Tauchen (1996, Econometric Theory 12, 657–681) are assessed. Exploiting the invariant distribution implied by the theory allows us to evaluate the error induced by simulations. Our results show that the estimators differ in their sensitivity to the sample size, the number of simulations, choice of auxiliary models, and computation demands. For some estimators, the test for overidentifying restrictions exhibit significant size distortions in small samples. Overall, while the simulation estimators have small bias, they are less efficient than pseudo-maximum likelihood (PMLE). Hence for the small sample sizes considered, the simulation estimators are still inferior to the PMLE estimates in a mean-squared sense.

Food Price Variability and Economic Reform: An ARCH Approach for Ghana

AbstractChanges in maize price levels and variability in Ghana are investigated. A model of wholesale price determination is reviewed in which grain stocks are held for speculative storage as well as export to neighboring countries in the Sahel. To test the model, an Autoregressive Conditionally Heteroskedastic (ARCH) regression is applied to monthly maize data for two markets over the period 1978–93. The regression is used to measure changes in maize price volatility in Ghana, and to infer the importance of past prices, domestic and regional production, and commodity storage and trade in explaining these changes.

Looking for evidence of speculative stockholding in commodity markets

Abstract The theory of commodity price with speculative storage predicts that prices are a two-regime process depending on whether or not inventories are held. The price process is nonlinear in that it is nondifferentiable at some p ∗ which separates the data into a history independent regime and an autoregressive process. This paper looks for evidence of nonlinearity in the price data and tests the theory in the context of threshold autoregressive models under the assumption that shocks to harvest are i.i.d. While we find evidence for regime-specific behavior, we also find the degree of persistence in the stockout regime to be much stronger than that predicted by theory.

Competitive Speculative Storage and the Cost of Refinery Product Supply

In a recent article in thisjournal, Charles Tiplitz (1986) presented estimates of the cost-minimizing level of seasonal distillate fuel oil inventories in the US primary sector. The costs he considered are for storage and the extra cost of operating refineries at extreme distillate yield. The "effect of price expectations" on seasonal inventories is deliberately excluded from the analysis. The Tiplitz study has interesting and useful results. However, it raises (without answering) the question of how cost-minimizing stocks differ from stocks held for price reasons. Some readers may conclude that price-responsive stocks are somehow different from or additional to cost-minimizing stocks. This interpretation runs counter to economic intuition. n the classic theory of the firm, the profit-maximizing plan for variable inputs also minimizes the variable cost of the most profitable output level. When there is an opportunity to store, supply can be accomplished through storage as well as input processing. Stocks are a variable input in the supply process. If producers are expected-profit maximizers, they should therefore manage stocks like other inputs to minimize the expected variable cost of supply.

Econometric testing for risk averse behaviour in agriculture

The time–series behaviour of commodity prices and consumption expenditures are examined under the assumption of risk averse speculative storage. An asset pricing model of commodity storage is specified with constant relative risk aversion utility and lognormally distributed commodity prices. Testing the overidentifying restrictions constitutes a joint test that the model is valid and that agents have rational expectations. The risk premium on commodity storage markets is characterized in terms of the relative risk aversion parameter of a representative agent and estimates of this parameter can be obtained using maximum likelihood methods. In an application to US corn, wheat and soybean storage, a risk premium is found and the resulting estimates of the relative risk aversion parameter compare favourably with prior information about farmer risk attitudes.