SIMD Finance Simulation of Agribusiness in Trinidad using High Performance Computing
Abstract
This report will propose research into the development of an agribusiness model that provides financial simulation information operating on SIMD architecture. The research is conducted as a scientific method to address the problem of increasingly failing agricultural crop farms in Trinidad and Tobago due to uncontrollable financial expenditure by the development of a crop farm model for financial simulations utilizing SIMD architecture implementation.
Introduction
This research is of great importance as total capital to start and manage any sustainable agribusiness is finite and cost of operation quickly becomes the major determining factor to whether the business succeeds or fails. Research show that there is a gap between the use of vector processing as it pertains to agricultural business finance.
This research also may assist in the reduction of cost for consumers by the adoption this method proposed. It may also increase the ability of the sector to provide more local produce partly due to the results of the research. Â It may assist in providing a frame work that permits price prediction. It may assist in understanding what crop output and type of crop to provide that suits market demand trends and the economy.
Objectives
This research project proposed is conducted with the aim of providing the following
Objective# 1- Create process chain/mapping model of a crop farm.
Objective# 2 - Develop a dynamic model from inputs and outputs using MATLAB.
Objective# 3 - Develop financial algorithm for simulating model.
Objective# 4 – Presentation of simulation results
 Literature Review
Cell microprocessor architecture designed in collaboration by Sony, Toshiba and IBM (STI in abbreviation) has been extensively used in off the shelf computer system that provide high performance computing for applications such as multimedia and vector processing, as well as other forms of computing(Hofstee 2008). Vector processing commonly classified as SIMD technique is the executing single instruction on multiple data.
SIMD- Single Instruction Multiple Data provides the capability of one control instruction to operate on data being feed to many processing units that perform the same operation. Data is supplied and received from the processing units by memory of the same data paths as the processing units (Yang et al. 2013). This type of system lends to Processor Arrays that interconnect and provide the ability of controlling the input and output destination on the processor units that is essential for many algorithms.
This system approach for input and output allows for at very high rates the converting of data from formats outside to the local formats of the array, this creates a system that is very application dependent(Bertil 1992).
Such architecture because of its ability to operate on large streams of data at high speed forms the integral backbone in the acceleration of financial simulators and financial computation systems worldwide; widely used for price derivation predicting, portfolio risk analysis and real-time financial stock exchange modeling.
The continuous reduction of transistor sizing and methods of constructing high performance microprocessors have allowed for off-the shelf commercial and consumer platforms that implement SIMD architectures in areas of hardware such as the Graphics processing unit (GPU), in such an application, extremely high graphic video gaming processing on consumer PC’s is done enabling a richer more dynamic experience for computer gamers which drive an highly financially profitable market called the Video Gaming Industry(Giles 2010).
Research shows the application of financial simulations and computations on GPU outperforms the Central Processing unit (CPU) by several factors such as programmability, flexibility and  allows for the usage of hardware acceleration presented in off the shelf computers; as very short spaces of time are required for real-time dynamic situations.
 NVIDIA’s output of new graphic cards claiming from the TESLA is the first C-language high performance computing development environment (High Performance Computing (HPC), 2008) system. It is based on Compute Unified Device Architecture (CUDA) which extends the C programming language libraries; here the technology enables C program utilization without the need of graphical languages.
This enables developers to access to utilize the power of graphical processor units in large scale computations that are completely parallelizable meaning the rate of processing allows for down to the wire financial decisions that mean success or failure in investments and business.
Research was done into the Caribbean Region and food security. Research shows each island is faced with major problems. Regionally, statistical analysis placed the Caribbean as major importer of food and is negatively impacted heavily with the disruption of trade through acts of God, man-made disasters, local, regional and global financial collapses.
Accessible food refers to the average citizen securing sufficient nutrition and is influenced by infrastructure related distribution and economic conditions. Available food, ability to grow one’s own food or import for other sources.
Food security is obtained when a country is capable of producing all food sufficient to the nutritional needs of that country’s population. Being food secure is primarily and directly related to the dynamic financial cost of providing the requirement stated previously.
Research has shown that population increases force agribusiness system to increase in capacity to support rapidly changing demands in food. This presents server financial challenges to local agribusiness as completion from cheaper imports undermine business development, creation and profits coupled with high run away cost of production not to mention consumer demand for specific food crops lead to business failure and an increasingly shrinking agricultural sector, while government and consumers pay above normal prices for their demand on imports slowly bankrupting the economy financially.
This however still presents a problem as business mapping increases efficiency, it is not enough, and financial information and weighting must be obtained to show the benefits of such a process (Scheer 2000).
This proposal tends to solve this problem by utilizing theories presented in financial simulations on GPU based vector processing in the method development of an agribusiness model that simulates sustainable agribusiness through financial control and management, financial forecasting, profitability prediction, risk assessment, better quality control, reduced labor cost, input to output process chain financial planning.
The proposed may also provide scalability and allow for price generation that enables both business profit and consumer price satisfaction.Â
Research was done into agribusiness as it relates to crop farming financing and the ability to provide sustainable ready for market crops at a cost that the average consumer will afford.
(i) crop farm business criteria
This also led to interfacing with agribusiness proprietors and their staff with crop farms ranging from 5 acres to under one acre of land. Monoculture type crop farming is chosen against biodiversity as a bound solely for simplifying analysis as recent literature identifies variations in produce yields that need further research when crop biodiversity is used.  Â
They were selected through national agricultural database, through meetings with farmers and through recent literature on agribusiness research of medium and small crop farm operations. The research also shows criteria set out by International Organization of Standards (ISO) 9001 provide requirements for quality management systems utilizing process flow chain approach.
Business process mapping approach is utilized in this proposal based on its simplicity and ability to extract and intuitively analyses the value added chain (primary chain) and all supporting links (secondary). Value added chain approach or primary chain provides all sustainable activities in stages, raw material must go through that adds or drives value and profitability of the finished product.
Support links or secondary links are not involved directly in production, they increase effectiveness and aid in competitiveness (Scheer 2000).
 These stages are defined, mapped and further developed to extract all inputs and outputs with an approximate weighted (data) financial requirement. Research will be done to provide accurate financial weighted data on support link operations (Lockamy III et al. 2011). Research will be done to provide in both primary and secondary chains financial weighted data on tacit knowledge as a source of value.
MATLAB is also chosen for its test and evaluation of algorithms before being coded in C programming.
 Each module or stage in the primary chain model will accept input data from both previous stage (internal inputs) and independent sources (external inputs) and deliver output data to the next stage with I/O stream sampling abilities. Supporting links are treated the exact same way.
(iii) SIMD Application
The data generated is passed independently to be formatting in C programming language to develop working financial simulation algorithms that take advantage of CUDA platform. Recent literature and benchmark reports show that NVIDIA-Tesla K20X is an excellent off the shelf suited GPU accelerator that delivers 3.52 teraflops of single-precision and 1.17 teraflops of double-precision at peak coupled with a high end PC’s provides the necessary capability to obtain performance required.
Literature presented on financial simulation algorithms is extensive and research into finance simulations employ Monte Carlo option model algorithms. The results will be placed on a simple graphical user interface (GUI) that provides methods to change inputs to the model.Â
General Data Collection
For this proposal, data for the model is obtained mainly from operating crop farms and Ministry of Food Production. This is done to present as real as possible practical data to operate with. The data is collected and separated in the form of inputs and outputs to the crop farm based on the general business process.
(i) General Business Process
Germination, Fertilizer ratio mixing, Weeding, Spraying, Collection & Storage, transport in that order is the general primary chain process respectively. The secondary chain follows marketing & Sales, Pollution Cleanup, Purchasing, Human Resources. From this collection inputs extracted are divided into two (2) main allocation; Physical Inputs and Human inputs.
Data collected uses internationally accepted measures at its rate of cost and given a fluctuation range (Papajorgji et al. 2009). The general output of the model provides two(2) main results: Crop production and money(cost and/or selling price). Each stage in the process dynamically generates output data for the next according to function processes that are developed through the use of MATLAB toolbox and its previous stage requirement.
(ii) General Application process
Data generated from the model is formatted to allow for ready adoption into C programming for SIMD financial simulation. Platforms and development tools may be sourced from NVIDIA development centers ( which provides MATLAB to CUDA implementation for the proposed research.
The proposed method is conducted using MATLAB software ( version R2011 or higher)Â as a primary tool for development and quick evaluation. This is available through the Department of Electrical and Computer Engineering. Software development tools for SIMD application can be obtained from NVIDIA Home ( and may be used on the NVIDIA Tesla K20 or K20X Personal Computer(PC) application built GPU cards(NVIDIA 2013).
Codes and resource tool are freeware and may be used in development of application specific proposal requirements. All processing may be achieved by PC with at least Intel core i7 Sandy Bridge processor and minimum 4GB of RAM. Storage space for data will require a minimum of 750 GB and is expected to increase as proposed research is conducted.
Funding for this proposal is applied through the Electrical &Computer Engineering Department of the University of the West Indies. The work will be supervised by Dr. Ajay Joshi of the University of the West Indies.
Conclusion
This proposal explores the development and implementation of financial simulation using SIMD architecture for agribusiness as it relates to crop farming. A unique opportunity has been created as literature has not yet been attempted to address this area in agriculture. This has the potential to redefine how financial predictions, per unit cost, cost cutting, selling price and profitability is done in agribusiness.
Government may use this approach to standardize market selling prices, provide subsidies that function for the benefit of local producers and consumers.
This leaves avenues open for exploring while project finding may beneficial for further exploration into agribusiness.
Appendix
Figure1: Gantt Chart of Proposal
Bibliography
Bertil, Svensson. 1992. Architectures.pdf.
Giles, Mike. "Fast Finance Calculations." New Hardware for Monte Carlo Calculations. no. 1 (2010): 1-15. www3.imperial.ac.uk/pls/portallive/docs/1/18529701.PDF (accessed February 22, 2013).Giles, Mike. "Fast Finance Calculations." New Hardware for Monte Carlo Calculations. no. 1 (2010): 1-15. www3.imperial.ac.uk/pls/portallive/docs/1/18529701.PDF (accessed February 22, 2013).
Scheer, August- Wilhelm. Aris: Business Process Modeling. New York: Springer, 2000. process chain&ots=aPWHFTbyTN&sig=7tUaVEzv0eCTdmBZNH1zuduz09M&redir_esc=y.
DE
