【正文】 rmance difference between different program implementation, identifying performance bottlenecks of programs, and understanding the hardware resource utilization, and is the important part of development and optimization of highperformance puting programs [7].How to take advantage of program performance analysis technology and bine the architecture characteristics of CPU and GPU to guide the planning and optimization of parallel programs so that a variety of puting resources of CPU and GPU are fully utilized is a problem worthy of study using GPU in generalpurpose puting at ：近幾年來，計(jì)算機(jī)圖形處理器（GPU）比摩爾定律發(fā)展得更迅猛，這種發(fā)展不僅體現(xiàn)在改善圖形處理、虛擬現(xiàn)實(shí)、計(jì)算機(jī)模擬以及相關(guān)運(yùn)用方面，還體現(xiàn)在為使用GPU作圖形以外處理的通用計(jì)算提供了良好的運(yùn)行平臺(tái)。目前，為了減少通用計(jì)算中GPU編程復(fù)雜性，許多GPU制造商和研究機(jī)構(gòu)提出了一系列編程語言和編程模式，這些編程模式類似于傳統(tǒng)德編程方法，但是具有不同的編程風(fēng)格，例如Brook++、CUDA以及OpenCL等等。通用計(jì)算中使用GPU來提高性能很大程度上取決于程序員的硬件知識(shí)和編程技術(shù)。由于軟件程序員往往對(duì)硬件平臺(tái)缺乏深層次的了解，并且沒有相應(yīng)的編寫硬件程序的能力，所以程序植入導(dǎo)致了各種各樣的運(yùn)用程序的增加效果有著明顯的差異，這些運(yùn)用程序加速了使用GPU的通用計(jì)算。如何利用程序性能分析技術(shù)以及如何結(jié)合CPU和GPU的體系特征來引導(dǎo)平行程序的規(guī)劃和優(yōu)化以至CPU和GPU的大量計(jì)算資源得到充分利用，是目前GPU在通用計(jì)算方面一個(gè)值得探討的問題。本論文從GPU內(nèi)部運(yùn)行機(jī)制出發(fā)，分析影響GPU執(zhí)行性能的主要因素，提出基于靜態(tài)程序分析的成本測(cè)試算法。Comment: Cost model has been widely used in the puter field as a way to evaluate whether a program is excellent or quite a number of researchers in puter field has proposed various kinds of cost models for GPU, of which most are under certain this paper give us a new perception of cost model for GPU in generalpurpose parallelization cost model for GPU can be generally applied for many application authors attempt to estimate the cost they take the GPU initialization cost, transmission of data cost as well as the program execution cost into aspect use a special algorithm to calculate the the cost is measured by the time of each this paper ,we can get through the authors’ idea very well ,because they are well organized in form and shown clearly in graphs, charts as well as in parallelization cost model is more accurate, flexible and portable than models of the we should learn the method they study a is to consider a question in a broad if we keep thinking in this way ,our ability of doing scientific research will be greatly the same time, we should keep close watch on the field of it is such an important processor that it is used widely used on modern is even considered the core graphics processor of it develops faster nowadays as the need rises.第三篇：專業(yè)英語論文翻譯嵌入式系統(tǒng)研究專業(yè)英語期末考查第1頁嵌入式系統(tǒng)研究1前言智能軟件Agent是能夠?yàn)橛脩魣?zhí)行特定的任務(wù)、具有一定程度的智能、能夠自主的執(zhí)行部分任務(wù)并以一種合適的方式和環(huán)境相互作用的軟件程序。這使得它適合在高度動(dòng)態(tài)的環(huán)境下做出及時(shí)的響應(yīng)。這些設(shè)備的大都具有嵌入式操作系統(tǒng)的支持, 并運(yùn)行著越來越豐富的應(yīng)用程序。本文將Agent技術(shù)引入嵌入式智能設(shè)備的測(cè)試中，使用目標(biāo)設(shè)備Agent, 測(cè)試控制Agent, 網(wǎng)絡(luò)環(huán)境Agent分別模擬和處理測(cè)試設(shè)備，測(cè)試工程師和測(cè)試環(huán)境的復(fù)雜性，利用Agent自身具有的特點(diǎn)，提出了一種有效的自動(dòng)化測(cè)試的方法。將Agent技術(shù)應(yīng)用于測(cè)試領(lǐng)域已經(jīng)有一些相關(guān)的研究，下面是具體的介紹。但是他們提出的測(cè)試用例選擇技術(shù)僅能在大量已有的測(cè)試用例中選擇最佳的用例，不能減少編寫測(cè)試用例本身的復(fù)雜性。Yu Qi、David Hung 和 Eric Wong [3] 提出了一個(gè)基于Agent 技術(shù)的Web 應(yīng)用程序測(cè)試方法。他們的方法不僅僅適用于Web應(yīng)用程序的測(cè)試，也適合于嵌入式智能設(shè)備的測(cè)試。常見的嵌入式智能設(shè)備測(cè)試工具(比如TestQuest)使用圖像比對(duì)來判斷目標(biāo)設(shè)備的狀態(tài), 這種方法雖然實(shí)現(xiàn)了非侵入性的測(cè)試,但是存在兩個(gè)問題,圖片的抓取和傳送消耗了大量測(cè)試資源,不同手機(jī)的用戶界面風(fēng)格變化很大,實(shí)際的設(shè)備中,當(dāng)重要的事件發(fā)生時(shí),只要操作系統(tǒng)相同,敏感事件的捕獲方式也是一樣,測(cè)試用例不會(huì)因?yàn)榻缑娴淖兓兓?。而測(cè)試控制Agent收到敏感事件后，根據(jù)自己的知識(shí)，采取相應(yīng)的動(dòng)作的機(jī)制（如異常處理，重新調(diào)度測(cè)試等）加以處理。此外，考慮到目標(biāo)設(shè)備處于復(fù)雜的網(wǎng)絡(luò)環(huán)境中，我們利用網(wǎng)絡(luò)環(huán)境Agent來控制目標(biāo)設(shè)備所處的網(wǎng)絡(luò)信號(hào)，從而實(shí)現(xiàn)對(duì)設(shè)備所處網(wǎng)絡(luò)環(huán)境的控制。如圖1所示：測(cè)試平臺(tái)分為四層，包括用戶接口層，測(cè)試控制層，通訊層和設(shè)備Agent層。專業(yè)英語期末考查第3頁用戶接口層測(cè)試管理與配置腳本編輯器虛擬手機(jī)測(cè)試控制層測(cè)試結(jié)果驗(yàn)證測(cè)試資源庫測(cè)試腳步執(zhí)行測(cè)試環(huán)境控制測(cè)試過程監(jiān)控Agent通訊層設(shè) 備agent層Synblan Agent藍(lán)牙 GPRS 3G AT視窗的移動(dòng)Agent定制AgentAT 接口圖1 測(cè)試系統(tǒng)的整體架構(gòu) 系統(tǒng)基本執(zhí)行流程測(cè)試過程是對(duì)真實(shí)用戶使用手機(jī)時(shí)“輸入－反饋”模型的一個(gè)模擬。腳本繼續(xù)運(yùn)行下面的語句，直到運(yùn)行結(jié)束。本系統(tǒng)中它接收從PC機(jī)中接收到的控制命令，然后在智能設(shè)備中進(jìn)行相應(yīng)的操作，包括模擬鍵盤事件，截取屏幕并返回給PC，以及根據(jù)知識(shí)庫中的配置的測(cè)試目標(biāo)，通過推理，有所選擇的將必需的狀態(tài)信息主動(dòng)通知測(cè)試宿主機(jī)。專業(yè)英語期末考查第4頁異常處理規(guī)則（EMRULE）這個(gè)決策規(guī)則決定了當(dāng)Agent發(fā)現(xiàn)測(cè)試中出現(xiàn)異常，要采取的處理方式。⑵ 智能性當(dāng)目標(biāo)設(shè)備Agent獲取到一個(gè)變化的被測(cè)系統(tǒng)狀態(tài)或信息時(shí)，它會(huì)根據(jù)推理規(guī)則，僅僅把和本次目標(biāo)相關(guān)的信息發(fā)送給測(cè)試控制Agent。測(cè)試控制 Agent測(cè)試控制Agent 實(shí)現(xiàn)了對(duì)測(cè)試過程的建模。⑵ 通過和腳本解釋器交互，實(shí)現(xiàn)異步的事件通知和交互測(cè)試用例的執(zhí)行。⑵ 異常處理規(guī)則（EMRULE）測(cè)試控制Agent也遵守異常處理規(guī)則，它能夠檢測(cè)測(cè)試過程中發(fā)生的異常事件，并執(zhí)行相應(yīng)的解決方案，使得測(cè)試能夠順利的進(jìn)行。⑵智能性測(cè)試控制Agent能夠根據(jù)用戶選擇的測(cè)試目標(biāo)，自動(dòng)生成ECA規(guī)則表，并根據(jù)ECA規(guī)則，進(jìn)行推理并采取相應(yīng)的動(dòng)作。 網(wǎng)絡(luò)環(huán)境Agent 網(wǎng)絡(luò)環(huán)境Agent 實(shí)現(xiàn)了對(duì)網(wǎng)絡(luò)環(huán)境的模擬和控制。系統(tǒng)評(píng)估為了評(píng)價(jià)本系統(tǒng)的有效性，我們?cè)O(shè)計(jì)了一個(gè)比較實(shí)驗(yàn)。我們將測(cè)試工程師分為兩組，第一組使用本文的系統(tǒng)MobileTest進(jìn)行測(cè)試，第二組使用業(yè)界著名的TestQuest Pro 進(jìn)行測(cè)試。測(cè)試的內(nèi)容是根據(jù)這兩個(gè)系統(tǒng)各自已經(jīng)建立好的回歸測(cè)試用例對(duì)新的智能手機(jī)進(jìn)行回歸測(cè)試，從而比較這兩個(gè)測(cè)試工具的測(cè)試效率和腳本的可維護(hù)性。從表中可見，雖然TestQuest 在功能測(cè)試上有更高的自動(dòng)化率，在壓力測(cè)試，多狀態(tài)測(cè)試，多任務(wù)測(cè)試，臨界測(cè)試和總計(jì)中，MobileTest有更好的測(cè)試覆蓋率。此外，MobileTest 完成測(cè)試任務(wù)的時(shí)間時(shí)15天，比TestQuest 的測(cè)試效率要高。將來的研究工作中，我們會(huì)進(jìn)一步拓展整個(gè)測(cè)試系統(tǒng)，使之能夠支持整個(gè)測(cè)試的生命周期。來源于：《 Journal of Electronic Science and Technology》專業(yè)英語期末考查第7頁附：英文原文Embedded system research 1 IntroductionIntelligent Software Agent is the ability to perform specific tasks for the user, with a certain degree of intelligence, able to perform some tasks and autonomy in a proper manner and environment interactions software has autonomy, responsiveness, learning and social and other makes it suitable for highly dynamic environment to make a timely technology and the development of new generation mobile munication networks makes the emergence of a large number of embedded intelligent of these devices with embedded operating system support, and run an increasingly rich to test for these applications to be a need to article Intelligent Agent technology into embedded devices test, using the target device Agent, test control Agent, Network Agent and treatment were simulated test equipment, test engineers and test plexity of the environment, the use of Agent has its own characteristics, proposed An effective automated Related research and ideas of this articleAgent possess autonomy, responsiveness, learning and social and other features, makes it very suitable for handling plex problems in the test technology in the test area will already have some relevant research, the following is a specific Choi and Byoungju Choi [1] proposed a testing tool based on Agent technology, which through the use of Agent to handle user interface and test those interactions, the use of Agent to carry out the test case test case selection choices, and use the Agent for regression testing regression testing, so a good automated software they can only be made in a large number of test selection techniques select test cases has been the best use cases, write test cases can not reduce its 第8頁