TW201140313A - Virtual performance monitoring apparatus and method for using to monitor system performance - Google Patents

Virtual performance monitoring apparatus and method for using to monitor system performance Download PDF

Info

Publication number
TW201140313A
TW201140313A TW99114377A TW99114377A TW201140313A TW 201140313 A TW201140313 A TW 201140313A TW 99114377 A TW99114377 A TW 99114377A TW 99114377 A TW99114377 A TW 99114377A TW 201140313 A TW201140313 A TW 201140313A
Authority
TW
Taiwan
Prior art keywords
virtual
performance
performance monitoring
monitoring
simulator
Prior art date
Application number
TW99114377A
Other languages
Chinese (zh)
Inventor
Shi-Hao Hong
jia-heng Hong
Original Assignee
Univ Nat Taiwan
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Univ Nat Taiwan filed Critical Univ Nat Taiwan
Priority to TW99114377A priority Critical patent/TW201140313A/en
Publication of TW201140313A publication Critical patent/TW201140313A/en

Links

Landscapes

  • Debugging And Monitoring (AREA)

Abstract

This invention relates to a virtual performance monitoring apparatus and method for using to monitor system performance. The apparatus includes a performance monitoring and managing module for monitoring a command flow executed on a virtual platform and receiving a performance monitoring command; at least a simulator for simulating a specific hardware; and at least a counter for storing related performance data of the command flow. The method includes: step A: monitoring a command flow executed on a virtual platform and receiving a performance monitoring command; step B: simulating specific software and hardware for obtaining performance data; step C: judging whether the performance data needs to be calculated for determining whether to enter step D or step E; step D: performing calculation on the performance data to generate a post-calculated performance data and entering into step E; step E: store the performance data.

Description

201140313 六、發明說明: 【發明所屬之技術領域】 本發明係關於一種用於系統效能監測之虛擬效能監控裝置及其方法, 特別是一種利用模擬器模擬出目標硬體平台,進而評估應用軟體與硬體之 互動與整能以及珊該顧倾之全部或部分程賴段效能的虛擬效 能監控裝置及其方法。 【先前技術】201140313 VI. Description of the Invention: [Technical Field] The present invention relates to a virtual performance monitoring device and method for system performance monitoring, in particular to simulating a target hardware platform by using a simulator, thereby evaluating application software and A virtual performance monitoring device and method for hardware interaction and versatility, and the performance of all or part of the process. [Prior Art]

近年來,在軟體與硬體的互動與開發愈來愈複雜的情況下,用來剖析 程式執行指令、追職定事件及找料統效驗_效能分私具即應運 而生。然而’在早期的軟駿計階段,設計者必須在軟體所對應之一目標 硬體平台上進行開發設計,因此在該目標硬體平台完成職之前,習知之 軟體效能分析工具並無法於該目標硬體平台上進行軟體效能之分析評估, 造成了在該目標硬體平台職階段,無法同時進行評估系統效能、找出效 能瓶頸以及料軟硬體制設計等,耻延長了贿軟種關發時間, 進而降低了產品的競爭力。 為了解決上述之問題,用來模擬該目標硬體平台之虛擬平台即娜 出來/模㈣目標硬體平台魏之方式,提供—個«環境給軟體歡 者錢計_即可切虛擬的目標魏平台上,進行健之操作與評估 疋般虛擬平台在執行軟體時,並沒有提供使用者關鍵的軟硬體效矣 影_’如私_型、絲_、_潛錢_人之潛伏等 因此使仔在該虛擬平台上無法有效的同時進行軟體硬體共同最佳4 (co-optimization)。 201140313 另一方面,習知之平台效能分析工具,係用準確性高但相對緩慢的指 令集架構模擬器(instruction set architecture simulator)或電子化系統層級模擬 . 工具(electronic system-level simulation tool)來評估單一應用軟體之效能,此 方式僅著重在模擬硬體,而沒有辦法分析應用軟體部分程式區段之效能, 所以在使用此種方式來作軟體效能評估時,只能夠提供某一程式區段在執 行時之整體效能數據,而無法單獨對某一特定程式區段進行剖析(piOfiie)。 此外’由於習知之平台效能分析工具的執行速度不能及時模擬較複雜 籲之硬體與軟體之間的互動,所以只能夠適用於簡化後之嵌入式作業系統, 故而限制了所開發出之軟體的硬體應用範圍。 因此,如何開發出-在虛擬平台上評估一複雜系統的整體效能之裝 置’且該裝置亦可義於評估該軟體單—特定程式區段之效能,即成了相 關軟硬體設計廠商努力的目標。 【發明内容】 本發明人有鑑於習知效能評估方式在虛擬平台上系統效能評估與軟體 #行為分析之問題’乃積極著手進行·,以期可以改進上舰有之缺點, 經過不斷地試驗及努力,終於開發出本發明。 本發明之主要目的在於提供-_於_效測之虛擬效能監控裝 置,其係操作於—虛擬平台’該虛擬平台係在—操作平台上以—中介軟體 來模擬-目㈣,細虛㈣上顯-目標娜統與—效能 監控軟體’該裝置包括-效能監控f理模組、至少—模觸及至少一計數 器。 該效能監控管理模組係用以接收並 監控在該虛擬平台上執行之指令威s ] 4 201140313 以及接受該霞幽概驗料;W恤難標軟體 系統之指令與該效能監控軟體之指令。 該至少一模擬器係連結於該效能監控管理模組,其係依據該效能監控 管理模組所接收之效能監控命令來模擬該目標硬體平台之部分或全部硬 體。 該至少-雜ϋ猶結機魏監控管理池,制⑽存在該虛擬 平台上執行之指令流的相關效能資訊。 • 本發明之另一目的在於提供一種用於系統效能監測之虛擬效能監控之 方法’該方法係操作於-虛擬平台,職擬平台係在—操作平台上以一中 介軟體來模擬-目標硬體平台,並在該虛擬平台上模擬一目標軟體系統與 一效能監控軟體,該方法包括: 步驟A ··接收並監控在該虛擬平台上執行之指令流及接受該效能監控 軟體之效能監控命令; 步驟B :依據該效能監控軟體之效能監控命令及該指令流,來模擬特 # 定之軟、硬體以取得一效能資訊; 步驟C :判斷該效能資訊是否需要進一步的計算處理,若是,則進入 步驟D ’若否,則直接進入步驟e ; 步驟D :對該效能資訊加以計算處理,並產生一經過計算後之效能資 . 訊’進入步驟E;以及 . 步驟E:儲存該效能資訊。 藉由上述之裝置與方法,可計算、處理並儲存在虛擬平台上所執行之 指令流的效能資訊,進而讓設計者可以利用該效能資訊評估不同硬體元 5 201140313 在整體线巾對軟體之蚊程式區段 生之影響。 單—软體或軟硬體之_效能所產 【實施方式】 &配合圖式將本發明之較 為使熟悉該項技藝人士瞭解本發明之目的 佳實施例詳細說明如下。 她__咐崎監控裝置In recent years, in the context of the increasingly complex interaction and development of software and hardware, it has been developed to analyze program execution instructions, track down events, and find the system. However, in the early stage of soft computing, the designer must develop and design on the target hardware platform corresponding to the software. Therefore, before the target hardware platform is completed, the conventional software performance analysis tool cannot be used for the target. The analysis and evaluation of software performance on the hardware platform has caused the evaluation of system performance, the identification of performance bottlenecks, and the design of soft and hard systems at the same stage of the hardware platform. , which reduces the competitiveness of the product. In order to solve the above problems, the virtual platform used to simulate the target hardware platform is the way out of the hardware platform of the target hardware platform, providing a "environment to the software winners" _ can cut the virtual target Wei On the platform, the virtual operation and evaluation of the virtual platform does not provide users with key software and hardware effects when executing software _'such as private _ type, silk _, _ 潜 钱 _ people lurking, etc. Enables the co-optimization of software hardware at the same time that it is not effective on the virtual platform. 201140313 On the other hand, the well-known platform performance analysis tool is evaluated with an accurate but relatively slow instruction set architecture simulator or electronic system-level simulation tool. The performance of a single application software, this method only focuses on the simulation hardware, and there is no way to analyze the performance of the application software part of the program section. Therefore, when using this method for software performance evaluation, only a certain program section can be provided. The overall performance data at the time of execution, and it is not possible to separately parse a particular program section (piOfiie). In addition, because the implementation speed of the platform performance analysis tool can not simulate the interaction between the hardware and software, it can only be applied to the simplified embedded operating system, thus limiting the software developed. The range of hardware applications. Therefore, how to develop a device for evaluating the overall performance of a complex system on a virtual platform' and the device can also be used to evaluate the performance of the software-specific program segment, which becomes the effort of the relevant software and hardware design vendors. aims. SUMMARY OF THE INVENTION The present inventors have in view of the problem that the system performance evaluation and the software #behavior analysis on the virtual platform are actively carried out in view of the conventional performance evaluation method, in order to improve the shortcomings of the ship, and to continuously test and work hard. Finally, the invention has been developed. The main purpose of the present invention is to provide a virtual performance monitoring device for operating in a virtual platform. The virtual platform is on the operating platform, and the intermediary software is used to simulate - (4), fine virtual (4) Display-target Natong and - performance monitoring software 'The device includes - performance monitoring module, at least - mode touch and at least one counter. The performance monitoring management module is configured to receive and monitor the instructions executed on the virtual platform, and to accept the instructions of the software system and the instructions of the performance monitoring software. The at least one simulator is coupled to the performance monitoring management module, and simulates part or all of the hardware of the target hardware platform according to the performance monitoring command received by the performance monitoring management module. The at least the hodgepodge machine monitors the management pool, and (10) has relevant performance information of the instruction stream executed on the virtual platform. Another object of the present invention is to provide a method for virtual performance monitoring for system performance monitoring. The method is operated on a virtual platform, and the platform is simulated on an operating platform by an intermediary software. a platform, and simulating a target software system and a performance monitoring software on the virtual platform, the method comprising: Step A: receiving and monitoring an instruction flow executed on the virtual platform and accepting a performance monitoring command of the performance monitoring software; Step B: According to the performance monitoring command of the performance monitoring software and the instruction stream, simulate the software and hardware to obtain a performance information; Step C: determine whether the performance information needs further calculation processing, and if so, enter Step D 'If no, go directly to step e; Step D: Calculate the performance information and generate a calculated performance resource. Enter 'Step E; and. Step E: Store the performance information. With the above device and method, the performance information of the instruction stream executed on the virtual platform can be calculated, processed and stored, so that the designer can use the performance information to evaluate different hardware elements 5 201140313 in the overall wire towel to software The impact of the mosquito section. The present invention is described in detail below with reference to the drawings. She __咐崎监控装置

(圖中未示)以一中介軟體(如 ’、° 種^•杈擬整個電腦系統,包括 中央處理ϋ及其他财設備的自域體,圖 Α 术模擬—目標硬體平 。(圖中㈣;細虛㈣墙—目標峨晴 能監控倾(财未示),_於純聽監測之虛擬效能監控裝置⑴ 包括一效能監控管理模組(i 〇 )、至少—模擬器(2 Q)及 (3 0)。 /效能驗管理· ( i Q )係賴概域控在該麵平台上執行 之指令流以及接受該效能監控軟體之效能監控命令。 該至少-模擬器(2 〇 )係連結於該效能監控管理模組(丄〇 ),其係 依據該效能馳管理池(! G )所減之魏監控命令來模擬該目標硬 體平台之部分或全部硬體。 該至少一計數器(3 〇 )係連結於該效能監控管理模組(1 〇 ),係用 以儲存在該虛擬平台上執行之指令流的相關效能資訊,進哪該效能資訊 傳至外部軟體作進一步之分析與應用。 其中’在本實施例中,該效能資訊包括:指令數(instruetiQn e_t)] 6 201140313 虛擬執行時間、時脈職數、快取存取數以及錄失誤數等。 - 叫參考第二圖所示,其中該效能監控管理模組(1〇)更包括一事件 -監控目錄(1 1)以及一效能事件收集介面(1 2 ),其中該事件監控目錄 (11)係與該至少-模擬器(2 0)連結,其係用以列出被監視之事件, 該各別被監視之事件係各自連結對應之模擬器;該效能事件收集介面(工 2)係與該事件監控目錄(丄χ)連結,其制以計算被監視之事件的資 料、更新该至少-計數器(3 〇)的資料以及交換該效能監控管理模組(工 籲 0 )、至少一模擬器(2 0)及至少一計數器(3 〇)之間的資訊,其更包 括-指令記錄緩衝器(! 2 1 )、-效能資料緩衝!| ( i 2 2 )及一管理函 數(123)。 其中該指令記錄緩衝器(121)係與該至少一模擬器(2 〇)連結, 其係用以保持指令流之追從(instructi〇ntrace)以驅動該至少一模擬器(2 0)。 該效能資料緩衝器(12 2)係與該至少一模擬器(2 〇)及該至少 隹一計數器(3 0)連結,其係用以儲存該至少一模擬器(2 〇)產生之效 能資訊,並將該效能資訊傳至該至少一計數器(3 〇 )。 該管理函數(12 3)係與該至少一模擬器(2 〇)及該至少一計數 器(3 0 )連結’其包括一組事件相依回呼函數(event_depen(jentcall-back • 加咖118)以管理該至少一模擬器(2 ◦),其包括有初始化函數(initialization(not shown) is an intermediary software (such as ', ° ° ^ ^ simulation of the entire computer system, including the central processing ϋ and other financial equipment self-domain, graphic simulation - target hardware flat. (In the picture (4); imaginary (four) wall - target 峨 能 能 能 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( And (3 0) / Performance Test Management ( i Q ) is the command flow executed on the platform and the performance monitoring command of the performance monitoring software. The at least - simulator (2 〇) system Linked to the performance monitoring management module (丄〇), which simulates part or all of the hardware of the target hardware platform according to the Wei monitoring command reduced by the performance management pool (! G). The at least one counter ( 3 〇) is linked to the performance monitoring management module (1 〇), which is used to store relevant performance information of the instruction stream executed on the virtual platform, and the performance information is transmitted to the external software for further analysis and application. Where 'in this embodiment, the performance The information includes: the number of instructions (instruetiQn e_t)] 6 201140313 Virtual execution time, clock number, cache access number and number of errors, etc. - Refer to the second figure, where the performance monitoring management module (1) 〇) further includes an event-monitoring directory (1 1) and a performance event collection interface (1 2 ), wherein the event monitoring directory (11) is linked to the at least-simulator (20), and is used to column The monitored events are respectively connected to the corresponding simulators; the performance event collection interface (Work 2) is linked to the event monitoring directory (丄χ), which is configured to calculate the monitored events. Information, updating the at least-counter (3 〇) data and exchanging information between the performance monitoring management module (work call 0), at least one simulator (20), and at least one counter (3 〇) Further includes an instruction record buffer (! 2 1 ), an performance data buffer!| (i 2 2 ), and a management function (123), wherein the instruction record buffer (121) is associated with the at least one simulator (2) 〇) link, which is used to keep track of instruction flow (instru Cti〇ntrace) to drive the at least one simulator (20). The performance data buffer (12 2) is coupled to the at least one simulator (2 〇) and the at least one counter (30), And storing the performance information generated by the at least one simulator (2 〇), and transmitting the performance information to the at least one counter (3 〇). The management function (12 3) is associated with the at least one simulator (2 〇 And the at least one counter (30) is linked 'which includes a set of event-dependent callback functions (event_depen (jentcall-back)) to manage the at least one simulator (2 ◦), which includes an initialization function ( Initialize

Unction)、最終化函數(finalization function)、推動指令記錄緩衝器(1 2 1 )之i料傳送函數(PushtoTraceBuf data transfer fbnction)與用以更新該Unction), finalization function, PushtoTraceBuf data transfer fbnction and the update of the instruction transfer buffer (1 2 1)

a „ [ S 至少一 s十數器(3 0)之資料傳送函數(UpdatetoPMC data transfer 201140313 fiinction )。 該初始化函數(initialization fimction)係用以初始化該至少一計數器(3 0)° 該最終化函數(finalization function )係用以清除該至少一計數器(3 0)内之紀錄。 該推動4曰令s己錄緩衝器(1 2 1 )之資料傳送函數(pusht〇TraceBufdata transferfUnction)係用以從指令流中得到記憶體存取指令流,以及將該記憶 體存取指令流放進該指令記錄緩衝器(1 2 1 )。 該更新至少-計數器(3 0 )之資料傳送函數(Updatet〇pMC她 tmnsferfimction)可在該效能監控軟體欲讀取該至少一計數器(3 〇 )時, 更新該至少一計數器(3 0)。 該至少一模擬器(2 0)包括: -微型架構模擬器(2 i),包括至少—用以模擬該目標硬體平台之微 型架構元件的模擬器;及 週邊㈣細(2 2),包括至少 邊裝置元件的模擬器。 在本實施例中,該微型架構模擬器( ,,ώ1)係與該效能監控管理模組 (1 0 )連結,其包括一管線模擬器(2 9 ν .t . )、一資料快取模擬器(2 1 4 ) Γ ( 2 1 3 }( 2 1 4)及一匯流排模擬器(2丄5 )。 丄 該管線模擬HUH ),係與該事件 於一被龄邾之拿株# . 皿&quot;·目錄(11)連結,並對應 件,其係用以模擬該目標硬體平台之中央處理器的管^ 201140313 (pipelme)執行,利用上述之模擬功能,&lt;進一步分析與估算出虛擬執行 , 時間,並將該虛擬執行時間回傳至該效能監控管理模組(1〇 )。 該貧料快取模擬器(2 1 2)係與該事件監控目錄(1 1)連結,並 對應於-被監視之事件’其侧以模擬該目標硬體平台之巾央處理器的資 料快取記憶體(data cache memory);該指令快取模擬器(2 1 3 )係與該 事件監控目錄(1丨)連結,並職於—被監視之事件,其_以模擬該 目心硬體平C»之指令快取記憶體(instructi〇n cache mem〇ry );該動態隨機存 鲁取記憶體模擬器(214)係與該事件監控目錄(1!)連結,並對應於 一被監視之事件,其係用以模擬該目標硬體平台之動態隨機存取記憶體; 該匯流排模娜(2 1 5 )係與該事件監控目錄(丨丨)連結,並對應於 一被監視之事件,其係用以模擬該目標硬體平台之匯流排。 該微型架構模擬器(21)可模擬該目標硬體平台之特定元件。舉例 來說,該資料快取模擬器(212)僅需要記憶體參考記錄以執行資料快 取記憶體之模擬,該資料快取模擬器(2 i 2)具有可調整參數,該可調 鲁整參數包括:快取容量(cache capacity)、區塊大小(block Size)、龄人性 (associativity)以及替換政策(repiacementp〇1icy)等。在模擬工作—成之 後,該微型架構模擬器(2 1)將該記憶體參考記錄所造成之快取命中數 與快取失誤數回報給該效能監控管理模組(1 〇 )。 -該週邊裝置模擬器(2 2)係與該效能監控管理模組(丄〇 )連結 .其包括一_記憶體模擬器(2 2 i )、磁碟模擬器(2 2 2 )、網路模擬 器(2 2 3)及直接記憶體存取模擬器(2 2 4 )。 該快閃記憶體模擬器(2 2 1 )係與該事件監控目錄(i丄)連結[s-j 9 201140313 之快閃記憶體; 並對應於-被監視之事件,其係用以模擬該目標硬體平台 目錄(11)連結,並對應於一 該磁碟模擬器(2 2 2 )係與該事件監控 被監視之事件,其制峨_目標硬體料之岭賴频擬器(2 2 3)係與該事件監控目錄(11)連結,並對應於-被監視之事件,其 係用以模擬該目標硬體平台之網路系統;該直接記憶體存取模擬器(2 2 4)係與該事件監控目錄(11)連結,並對應於-被監視之事件,其係 用以模擬该目標硬體平台之直接記憶體存取器。a „ [ S at least one s ten (3 0) data transfer function (UpdatetoPMC data transfer 201140313 fiinction). The initialization function (initialization fimction) is used to initialize the at least one counter (3 0) ° the finalization function (finalization function) is used to clear the record in the at least one counter (30). The push data transfer function (pusht〇TraceBufdata transferfUnction) of the s recording buffer (1 2 1) is used to slave instructions. The memory access instruction stream is obtained in the stream, and the memory access instruction stream is put into the instruction record buffer (1 2 1 ). The update is at least - the data transfer function of the counter (3 0 ) (Updatet〇pMC her tmnsferfimction The at least one counter (30) may be updated when the performance monitoring software wants to read the at least one counter (3 。). The at least one simulator (20) comprises: - a micro-architecture simulator (2 i) , including at least - a simulator for simulating the micro-architectural components of the target hardware platform; and a peripheral (four) thin (2 2), including a simulator of at least the side device components. In this embodiment, The micro-architecture simulator ( ,, ώ 1) is coupled to the performance monitoring management module (10), and includes a pipeline simulator (2 9 ν .t . ) and a data cache simulator (2 1 4 ) Γ ( 2 1 3 } ( 2 1 4) and a bus simulator (2丄5). 丄 The pipeline simulates HUH), and the event is in the same age as the one of the squats. #皿&quot;·Table of Contents (11 Linking, and corresponding, which is used to simulate the CPU of the target hardware platform ^ 201140313 (pipelme) execution, using the above simulation function, & < further analysis and estimation of virtual execution, time, and The virtual execution time is transmitted back to the performance monitoring management module (1〇). The poor material cache simulator (2 1 2) is linked to the event monitoring directory (11) and corresponds to the monitored event. 'The side is to simulate the data cache memory of the target central processor of the target hardware platform; the instruction cache simulator (2 1 3) is linked to the event monitoring directory (1丨), Companion to the -monitored event, its _ to simulate the hard disk of the hard disk C» cache memory (instructi〇n ca Che mem〇ry ); the dynamic random access memory simulator (214) is linked to the event monitoring directory (1!) and corresponds to a monitored event, which is used to simulate the target hardware platform. Dynamic random access memory; the bus modulo (2 15) is linked to the event monitoring directory (丨丨) and corresponds to a monitored event, which is used to simulate the target hardware platform Bus bar. The micro-architecture simulator (21) simulates specific components of the target hardware platform. For example, the data cache simulator (212) only needs a memory reference record to perform simulation of the data cache memory, and the data cache simulator (2 i 2) has adjustable parameters, which can be adjusted. Parameters include: cache capacity, block size, associativity, and replacement policy (repiacementp〇1icy). After the simulation work, the micro-architecture simulator (2 1) reports the number of cache hits and cache misses caused by the memory reference record to the performance monitoring management module (1 〇 ). - the peripheral device simulator (2 2) is connected to the performance monitoring management module (丄〇2). It includes a memory simulator (2 2 i ), a disk simulator (2 2 2 ), and a network. Simulator (2 2 3) and direct memory access simulator (2 2 4). The flash memory emulator (2 2 1 ) is coupled with the event monitoring directory (i丄) [sj 9 201140313 flash memory; and corresponds to the monitored event, which is used to simulate the target hard The platform catalog (11) is linked and corresponds to an event that the disk simulator (2 2 2 ) is monitored and monitored by the event monitoring device, and the target device is a ridge target device (2 2 3) Linked to the event monitoring directory (11) and corresponding to the monitored event, which is used to simulate the network system of the target hardware platform; the direct memory access simulator (2 2 4) Linked to the event monitoring directory (11) and corresponding to the monitored event, which is used to simulate the direct memory accessor of the target hardware platform.

本發明之用於系統效能監測之虛擬效能監控裝置⑴更包括一使用 者介面(圖巾未示),使用者可藉由該制者麵難所要啟敗模擬器數 目,-般來說’啟用衫賴擬器不但越提昇效能分析之準確性,更能 夠得到更麵效歸訊。然而,本發明之麟魏聽監測之虛擬效能監 控裝置(1 )係並行且分開執行多麵擬器(M鑛^遍,其 整體模擬速度相等於速度最慢之模擬㈣速度,目此相較於_次執行一個 模擬器的方式,本發明可有賴^執行彳後個模擬糾之整體執行速度。 該至少一汁數器(3 0 )包括一虛擬時間標籤計數器(3 1 )、-時脈 週期計數器(3 2)及一事件計數器(3 3)。 該虛擬時間標籤計數^ ( 3 i )制靖存虛擬執行_ ;該時脈週 期·》十數器(3 2 )係用以儲存該指令流所細之時脈週期數:事件計數器 (3 3)係用以儲存與效能有關之事件。 此外本發明更同時在該目標硬體平台與該虛擬平自上執行複數標準 檢查程式,本發明m統效能監測之虛擬效能監控裝置更包括一計時 裝置(圖未不)’係用以取得該指令流之實際與虛擬執行時間,利用該計時 201140313 可以下列計算式 ER = Ο 100% 其中ER代表該計時裝置之準確率,P為預測值, 該標準檢查程式在該虛擬平台 上執行之一虛擬執行時間 其為該計時裝置取得 時間 …十時裝置之平解轉則以下列計算式求出:The virtual performance monitoring device (1) for system performance monitoring of the present invention further includes a user interface (not shown), and the user can defeat the number of simulators by the manufacturer, and is generally enabled. Not only does it improve the accuracy of performance analysis, but it also enables more effective returning. However, the virtual performance monitoring device (1) of the Lin Wei listening monitoring of the present invention performs the multi-faceted simulator in parallel and separately, and the overall simulation speed is equal to the slowest analog (four) speed. In the manner of executing one simulator _ times, the present invention may depend on the overall execution speed of the subsequent simulation correction. The at least one juice counter (30) includes a virtual time tag counter (3 1 ), - clock a cycle counter (3 2) and an event counter (3 3). The virtual time tag count ^ ( 3 i ) is stored in the virtual execution _; the clock cycle · "10" is used to store the The number of clock cycles of the instruction stream: the event counter (3 3) is used to store performance-related events. In addition, the present invention simultaneously executes a plurality of standard check programs on the target hardware platform and the virtual platform. The virtual performance monitoring device for inventing the m system performance monitoring further includes a timing device (not shown) for obtaining the actual and virtual execution time of the instruction stream. With the timing 201140313, the following calculation formula ER = Ο 100% can be used. On behalf of the timing Home of accuracy, P is the predicted value of the standard inspection program executed on one of the virtual time of the virtual platform execution timing means for acquiring time ten o'clock ... Solutions level of transfer apparatus places the following calculation equation:

K 其中顧代表該計時裝置之平解確率,κ _縣檢查程式之總 數,而ERi為第1個標準檢查程式之準確率。 請參考第三圖所示,本發明之用於系統效能監測之虛擬效能監控之方 法’該方法係操作於-虛擬平…該虛擬平台係在—操作平台上以一中介 軟體(如QEMU’圖中未不)來模擬—目標硬體平台(圖中未示);並在該 虛擬平台上模擬-目標軟體系統(圖中未示)與—效能監控軟體(圖中未 示),該方法包括: 步驟A ( 4 0 1 ).接收並監控在該虛擬平台上執行之指令流及接受該 效能監控軟體之效能監控命令; (4 0 2 ).依據能監控軟體之效能監控命令及·令流來 模擬特技軟、硬體以取得—效能資訊; 々驟C ( 4 0 3 ).判斷該效能資訊是否需要進—步輯算處理,若是⑶ 11 05); 201140313 並產生一經過計算後 則進入步驟D(4 0 4) ’若否,則直接進入步驟e(4 步驟D ( 4 0 4 ):對該效能資訊加以計算處理, 之效能資訊,進入步驟E(4 〇 5);及 步驟E ( 4 0 5 ):儲存該效能資訊。 此外,該步驟E(4 〇 5)執行完成後,可回到執行步驟a(4◦丄). 該效能資1_ :齡數(instruetl_unt)、虛_辦間、時脈週期數、 快取存取數以及快取失誤數等。K where Gu represents the average resolution of the timing device, κ _ county inspection program total, and ERi is the accuracy of the first standard inspection program. Please refer to the third figure, the method for virtual performance monitoring of system performance monitoring of the present invention. The method is operated on a virtual platform. The virtual platform is on an operating platform with an intermediary software (such as QEMU'. In the simulation-target hardware platform (not shown); and on the virtual platform, the simulation-target software system (not shown) and the performance monitoring software (not shown), the method includes : Step A ( 4 0 1 ). Receive and monitor the instruction flow executed on the virtual platform and accept the performance monitoring command of the performance monitoring software; (4 0 2 ). According to the performance monitoring command and the flow of the monitoring software To simulate the soft and hardware of the stunt to obtain the performance information; Step C (4 0 3 ). Determine whether the performance information needs to be processed in advance, if it is (3) 11 05); 201140313 and generate a calculated Step D (4 0 4) 'If no, go directly to step e (4 step D (4 0 4 ): calculate and process the performance information, the performance information, enter step E (4 〇 5); and step E ( 4 0 5 ): Store the performance information. In addition, this step E (4 〇5) After the execution is completed, you can go back to the execution step a (4◦丄). The performance resource 1_: age (instruetl_unt), virtual_interaction, number of clock cycles, cache accesses, and cache Number of mistakes, etc.

請參考第四圖所示,本發明之用於系統效能監測之虛擬效能監控之方 法更包括: 之後’傳送該效能資訊 步驟F ( 4 0 6 ):係執行於步驟E ( 4 〇 5 ) 至該效能監控軟體赠進—步的監控及效能評估的參考資訊。 此外該步驟F ( 4 0 6)執行完之後,可回到步驟A ( 4 0 1 )。 藉由上述之結構與方法,可計算、處理並儲存在虛擬平台上所執行之 產生之影響 旨令流的效能資訊,進而讓設計者可以_該效蹄訊評估於不同硬體元 件在整體純中對軟體之特定程式區段、單__軟^錄硬體之間的效能所 —睛參考第二至四圖所示,為了讓審查委員更容易瞭解本發明之創作内 容,特舉-實施方式說明如下: 田使用者欲得到-在虛擬平台上執行之指令流的效能資訊(如快取失 誤數)時’首先先透過紐能監控軟體下—魏監控命令; 〜效此孤控管理模組(丨Q )依據該效能監控命令,啟動對應之至少 擬器(如該快取模擬器(2丄2》,使該至少一模擬器開始模擬所^ 12 201140313 應之該目標硬體平台之元件(如令央處理器的快取記憶體(切匕 memory)) ’同時取得該指令流的一效能資訊; • 該效能監控管理模組(1 〇)判斷該效能資訊是否需要進一步的叶算 若是,則對該效能資訊加以計算,並產生一經過計算後之效能資訊(如陕 取失誤處罰時間),最後再以該計算後之效能資訊更新該至少一計數器(3 〇 )(如將快取失誤數儲存至該事件計數器(3 3)或利用該快取失誤處罰 時間更新該虛擬時間標籤計數器(3 χ )内之該虛擬執行時間);若否,則 • 直接將該效能資訊儲存至該至少一計數器(3 〇 )。 最後將該更新後的效能資訊回傳至該效能監控軟體,以供使用者作進 一步的分析。 顯然地,依社面實施财的描述,本發明可能有許多的修正與差異。 因此需要在其附加的權利要求項之範_加以理解,除了上述詳細的描述 外,本發明還可以叙地在其他的實施射施行。上述僅為本發明之較佳 實施_已,並非用雜定本發明之㈣翻細;凡其絲麟本發明 • ^不之精神下所完成的等效改變或修飾,均應包含在下述申請專利範圍 内〇Please refer to the fourth figure, the method for virtual performance monitoring of the system performance monitoring of the present invention further includes: then transmitting the performance information step F (460): performing step E (4 〇 5) to The performance monitoring software provides reference information for monitoring and performance evaluation. In addition, after the step F (460) is performed, it can return to step A (4 0 1). Through the above structure and method, the performance information generated by the virtual platform can be calculated, processed and stored, so that the designer can evaluate the different hardware components in the overall pure In the specific program section of the software, the performance between the single __software hardware, as shown in the second to fourth figures, in order to make it easier for the reviewing committee to understand the creative content of the present invention, the special implementation The mode is as follows: The field user wants to get - the performance information of the instruction stream executed on the virtual platform (such as the number of cache misses) - first through the Newton monitoring software - Wei monitoring command; ~ effect this orphan management mode The group (丨Q) starts the corresponding at least emulator according to the performance monitoring command (such as the cache simulator (2丄2), so that the at least one simulator starts to simulate the target hardware platform Component (such as the cache memory of the central processor) to obtain a performance information of the instruction stream; • the performance monitoring management module (1 〇) determines whether the performance information needs further leaves If yes, the performance information is calculated, and a calculated performance information (such as the penalty time of the acquisition error) is generated, and finally the at least one counter (3 〇) is updated with the calculated performance information (if it will be fast) The error is stored in the event counter (3 3) or the virtual execution time in the virtual time tag counter (3 χ ) is updated by using the cache error penalty time; if not, the performance information is directly stored to The at least one counter (3 〇) finally transmits the updated performance information back to the performance monitoring software for further analysis by the user. Obviously, the invention may have many Modifications and differences are therefore intended to be understood by the scope of the appended claims, and in addition to the detailed description above, the invention may be practiced in other embodiments. The foregoing is only a preferred embodiment of the invention. It has not been used to (4) to refine the invention; the equivalent changes or modifications made by the Silken in the spirit of the invention shall be included in the following application. Within the scope billion

13 201140313 【圖式簡單說明】 第一圖係本發明之架構圖。 ' 第二圖係本發明之細部架構圖。 ' 第三圖係本發明之方法流程圖。 第四圖係本發明之詳細方法流程圖。 【主要元件符號說明】 1用於系統效能監測之虛擬效能監控裝置13 201140313 [Simplified description of the drawings] The first figure is an architectural diagram of the present invention. The second figure is a detailed diagram of the present invention. The third diagram is a flow chart of the method of the present invention. The fourth figure is a flow chart of the detailed method of the present invention. [Main component symbol description] 1 Virtual performance monitoring device for system performance monitoring

Device

1 0效能監控管理模組 1 2效能事件收集介面 1 2 2效能資料緩衝器 2 0模擬器 2 1 1管線模擬器 2 1 3指令快取模擬器 2 2週邊裝置模擬器 2 2 2磁碟模擬器 2 2 4直接記憶體存取模擬器 3 1虛擬時間標籤計數器 3 3事件計數器1 0 performance monitoring management module 1 2 performance event collection interface 1 2 2 performance data buffer 2 0 simulator 2 1 1 pipeline simulator 2 1 3 instruction cache simulator 2 2 peripheral device simulator 2 2 2 disk simulation 2 2 4 direct memory access simulator 3 1 virtual time tag counter 3 3 event counter

4 Ο 1步驟A 4 0 2步驟B 4 0 3步驟C 1 1事件監控目錄 1 2 1指令記錄緩衝器 1 2 3管理函數 2 1微型架構模擬器 2 1 2資料快取模擬器 2 1 4動態隨機存取記憶體模擬 2 1 5匯流排模擬器 2 2 1快閃記憶體模擬器 2 2 3網路模擬器 3 0計數器 3 2時脈週期計數器 14 2011403134 Ο 1Step A 4 0 2Step B 4 0 3Step C 1 1 Event Monitoring Directory 1 2 1 Instruction Record Buffer 1 2 3 Management Function 2 1 Micro Architecture Simulator 2 1 2 Data Cache Simulator 2 1 4 Dynamics Random Access Memory Analog 2 1 5 Bus Simulator 2 2 1 Flash Memory Simulator 2 2 3 Network Simulator 3 0 Counter 3 2 Clock Cycle Counter 14 201140313

4 〇 4步驟D 4 Ο 5步驟E 4 Ο 6步驟F4 〇 4 Step D 4 Ο 5 Step E 4 Ο 6 Step F

Claims (1)

201140313 七、申請專利範圍: 1,-種用於系統效能監測之虛擬效能監控裝置,其係操作於—虛擬平台,該 虛擬平台係在-操作平台上以-中介軟體來模擬—目標硬體平台,並在該 虛擬平台上模擬-目標軟齡贿—效能監控健,朗於纽效能監測 之虛擬效能監控裝置包括: -效能監控管理模組’係用以接收並監控在該虛擬平台上執行之指令 流以及接受該效能監控軟體之效能監控命令; • 至少一模擬器,係連結於該效能監控管理模組,其係依據該效能監控 管理模組所接收之效能監控命令來模擬該目標硬體平台之部分或全部硬 體;以及 至少一計數器’係連結於該效能監控管理模組,係用以儲存在該虛擬 平台上執行之指令流的效能資訊。 2. 如申請專利範圍第丨項所述之用於系統效能監測之虛擬效能監控装置,其 中該效能監控管理模組包括: • —事件監控目錄’係與該至少一模擬器連結,係用以列出被監視之事 件’該各別被監視之事件係各自連結對應之模擬器;以及 -效能事件收集介面,係與該事件監控目錄連結,係用以計算被監視 之事件的資料、更新該至少一計數器的資料以及交換該效能監控管理模 組、至少一模擬器及至少一計數器之間的資訊。 3. 如t w利細第2項所述m敝能監測之虛擬效能驗裝置,其 中該效能事件收集介面包括: -指令記錄緩衝器,係、與該至少_模擬器連結,其係用以保持指令ά s] 201140313 之追踨以驅動該至少—模擬器; • _效能貝料緩衝器,係與該至少一模擬器及該至少-計數器連結,其 侧_存該至少-模擬ϋ產生之效能資訊,並_效能f訊傳至該至^ 一計數器;以及 -管理函數’係、與該至少—模擬器及該至少—計數器連結,其包括— 組事件相依回呼函數以管理該至少一模擬器。 4. 如申明專利|巳圍第3項所述之用於系統效能監測之虛擬效能監控裝置,其 φ 中該至少一模擬器包括: &quot; -微型架構模擬器’包括至少一用以模擬該目標硬體平台之微型架構 元件的模擬器;以及 一週邊裝置模擬器,包括至少一用以模擬該目標硬體平台之週邊裝置 元件的模擬器。 5. 如申4專她圍第4撕述之驗系統效能制之虛擬魏監控裝置其 中該微型架構模擬器係與該效能監控管理模組連結,用以模擬該目標硬體 鲁平台之特定元件,其包括: 一管線模擬器,係與該事件監控目錄連結,並對應於一被監視之事件, 其係用以換擬該目標硬體平台之中央處理器之管線執行,利用上述之模擬 力食 b進步分析與估算出虛擬執行時間並將該虛擬執行時間回傳至該 . 效能監控管理模組; 一貝料快取模擬器,係與該事件監控目錄連結,並對應於一被監視之 事件’其係用以模擬該目標硬體平台之中央處理器的資料快取記憶體; 一指令快取模擬器’係與該事件監控目錄連結,並對應於一被監視之 17 201140313 事件’其她娜_恤巾输雜梅記憶體; 、動祕機存取記鐘模擬器,係與該事件監控目錄連結,並對應於 見之事件其伽吨擬該目標硬體平台之觸隨機存取記憶體; 以及 一匯流顺擬n,係與解件監控目料結,麟胁-被監視之事 件,其係肋模擬該目標硬體平台之匯流排。 6‘如申請補細第4或5項所狀用於线效紐敬虛触能監控裝 置’其中該週邊裝置模擬器係與該效能監控管理模組連結,直包括: 一快閃記賴模擬器,倾該事件監控目錄連結,並對絲一被監視 之事件,其係用以模擬該目標硬體平台之快閃記憶體; 一磁碟模擬器,係與該事件監控目錄連結,並對應於-被監視之事件, 其係用以模擬該目標硬體平台之磁碟,· -網賴擬H,係無事件驗目職結,並制於—被監視之事件, 其係用以模擬該目標硬體平台之網路系統;以及 一直接記憶體存取模擬器,係與該事件監控目錄連結,並對應於一被 監視之事件,其_輯擬該目標硬體平台之直接記隨存取器。 7.如申請專利範圍第6項所述之用於系統效能監測之虛擬效能監控裝置,該 至少一計數器包括: -虛擬時間標籤計數器,侧靖存虛擬執行時間; -時脈聊収H,铜靖存該齡流树脈聊數;以及 事件。十數器’係用以儲存與效能有關之事件。 8.如申請專利範圍第 7項所述之驗_絲_之虛贼雛控裝置,更[ 201140313 包括-使用者介面’使用者藉由該使用者介面調整所要啟狀模擬器數目。 9. 如中#專鄕圍第1項所述m驗能朗之虛擬效能監控裝置,其 巾’該效監控命令為—高階語言命令,該效能資訊包括指令數(instruction count)、虛擬執行時間、時脈週嫌、快取存取數以及快取失誤數。 10. 如申β專赚g第3項所述之用於系統魏監測之虛擬效能監控裝置, 其中該管理函數包括: 一初始化函數,係用以初始化該至少一計數器; • 一最終化函數,係用以清除該至少-計數器内之紀錄; 一推動指令記錄_ϋ之龍傳送純,_峨指令記鱗到記憶 體存取指令流’以及_記舰存取齡献職指令__器;以及 更新至少-計數器之資料傳送函數,係肋在該效能監控軟體欲讀 取β亥至少一計數器時,更新該至少一計數器。 1.如申β專利範圍第丨項所述m統效冑n狀虛擬效能監控裝置, 係並行且分開執行多個模擬器(mu驗eaded simulati〇n),其整體模擬速度相 隹等於速度最慢之模擬器的速度。 12.如申請專利範圍第!項所述之用於系統效能監測之虛擬效能監控裝置, 更包括-計時裝置,係用以取得該指令流之虛擬執行時間,利用該計時裝 置取得-標準檢查程式分別執行於該目標硬體平台與該虛擬平台上之實際 與虛擬執行時間後,以下列計算式求出該計時裝置之準確率·· ER=^^*i〇0% 其中ER代表該計時裝置之準確率,p為預測值,其為該計時裝置取得 該標準檢查程式在該虛擬平台上執行之—虛擬執行咖;◦為朗值其[S 19 201140313 為該計時㈣取得該標準檢查程式在該 s 時間 標硬體平纟上齡之—實際執行 * ’ 其::置==::=:虛__ ^£7? = Xi=〇ll5il Κ 其中aER代表断時裝置之平解確率,κ _標準檢查程式之總 φ ,而ERi為第1個標準檢查程式之準確率。 14.一種用於系統效能監測之虛擬效能監控之方法,係操作於—虛擬平台, 該虛擬平台係在—操作平台上以一中介軟體來模擬-目標硬體平台,I在 該虛擬平台上模擬-目標軟體系統與一效能監控軟體,該方法包括·· 步驟A.接收並監控在該虛擬平台上執行之指令流及接受該效能監控 軟體之效能監控命令; 步驟B :依據該效能監控軟體之效能監控命令及該指令流,來模擬特 φ 定之軟、硬體以取得一效能資訊; 步驟C :判斷該效能資訊是否需要進一步的計算處理,若是,則進入 步驟D,若否,則直接進入步驟e ; 步驟D :對該效能資訊加以計算處理,並產生一經過計算後之效能資 訊’進入步驟E;以及 步驟E:儲存該效能資訊。 15.如申請專利範圍第14項所述之用於系統效能監測之虛擬效能監控之方 法’更包括: f 20 201140313 步驟F .係執行於步驟E之後,傳送該效能資訊至該效能監控軟體以 * 作進一步的監控及效能評估的參考資訊。 • 16·如申吻專利範圍第14或15項所述之用於系統效能監測之虛擬效能監控 之方法,、中,該效成資訊包括:指令數(丨耶加以丨加c⑽扯)、虛擬執行時 間、時脈週期數、快取存取數以及快取失誤數等。 17.如申請專娜圍帛14 _述之祕祕魏監測之虛擬效能監控之方 法,其中該步驟E執行完成後,回到步驟A。 φ I8.如申請專利範圍第b項所述之用於系統效能監測之虛擬效能監控之方 法,其中該步驟F執行完成後,回到步驟A。201140313 VII. Patent application scope: 1. A virtual performance monitoring device for system performance monitoring, which operates on a virtual platform, which is simulated on the operating platform by an intermediary software - target hardware platform And on the virtual platform, the simulation-target soft-bribery-performance monitoring, the virtual performance monitoring device for the performance monitoring of the New Zealand includes: - the performance monitoring management module is used to receive and monitor the execution on the virtual platform The command stream and the performance monitoring command of the performance monitoring software; • at least one simulator connected to the performance monitoring management module, which simulates the target hardware according to the performance monitoring command received by the performance monitoring management module Part or all of the hardware of the platform; and at least one counter is coupled to the performance monitoring management module for storing performance information of the instruction stream executed on the virtual platform. 2. The virtual performance monitoring device for system performance monitoring as described in the scope of the patent application, wherein the performance monitoring management module comprises: - an event monitoring directory is coupled to the at least one simulator, List the monitored events 'The respective monitored events are connected to the corresponding simulators; and the performance event collection interface is linked to the event monitoring directory to calculate the data of the monitored events and update the And at least one counter data and information exchanged between the performance monitoring management module, the at least one simulator, and the at least one counter. 3. The virtual performance checking device as described in item 2 of the tw, wherein the performance event collecting interface comprises: - an instruction recording buffer, coupled to the at least _ simulator, which is used to maintain The command ά s] 201140313 is tracked to drive the at least one simulator; • the _ performance bedding buffer is coupled to the at least one simulator and the at least counter, and the side _ stores the at least the performance of the analog ϋ Information and _ efficiencies are passed to the counter and the management function is coupled to the at least one emulator and the at least counter, including a set of event dependent callback functions to manage the at least one emulation Device. 4. The virtual performance monitoring device for system performance monitoring according to claim 3, wherein the at least one simulator comprises: &quot;a micro-architecture simulator&quot; comprising at least one for simulating the An emulator of a micro-architectural component of the target hardware platform; and a peripheral device emulator comprising at least one simulator for simulating peripheral device components of the target hardware platform. 5. For the virtual Wei monitoring device of the system performance system, the micro-architecture simulator is connected with the performance monitoring management module to simulate the specific components of the target hardware platform. The method includes: a pipeline simulator connected to the event monitoring directory and corresponding to a monitored event, which is used to replace the pipeline execution of the central processor of the target hardware platform, and utilizes the above simulated force The food b progress analysis and estimate the virtual execution time and return the virtual execution time to the performance monitoring management module; a billing cache simulator, linked to the event monitoring directory, and corresponding to a monitored The event 'is used to simulate the data cache of the central processor of the target hardware platform; an instruction cache simulator is linked to the event monitoring directory and corresponds to a monitored 17 201140313 event' Her Na _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Touch random access memory; and a quasi bus cis n, the monitoring system and the solution mesh material junction element, Lin threat - events are monitored, its ribs analog bus lines of the target hardware platform. 6' If you apply for the 4th or 5th item, you can use it to connect to the line-effect monitoring device. The peripheral device simulator is connected to the performance monitoring management module, including: The event monitors the directory link, and the event that is monitored by the wire is used to simulate the flash memory of the target hardware platform; a disk simulator is linked to the event monitoring directory and corresponds to - the monitored event, which is used to simulate the disk of the target hardware platform, - the network is not H, is the event-free inspection, and is used to - the monitored event, which is used to simulate the a network system of the target hardware platform; and a direct memory access simulator connected to the event monitoring directory and corresponding to a monitored event, which directly records the target hardware platform Take the device. 7. The virtual performance monitoring device for system performance monitoring according to claim 6, wherein the at least one counter comprises: - a virtual time tag counter, a side execution virtual execution time; - a clock chat H, copper Jing Cun’s number of chattering sessions; and events. The tensor is used to store performance-related events. 8. If the application of the invention is as described in item 7 of the patent application, the [201140313 includes-user interface] user adjusts the number of simulators to be activated by the user interface. 9. If the virtual performance monitoring device of the m-inspection of the first item mentioned in Item 1 is used, the effect monitoring command is a high-order language command, and the performance information includes an instruction count (instruction count) and a virtual execution time. , clocks, cache accesses, and cache misses. 10. The virtual performance monitoring device for system monitoring according to item 3 of claim 3, wherein the management function comprises: an initialization function for initializing the at least one counter; • a finalization function, Used to clear the record in the at least-counter; a push command record _ ϋ 龙 传送 传送 传送 传送 传送 传送 传送 传送 传送 传送 传送 到 到 到 到 到 到 到 到 到 到 到 到 到 记忆 记忆 记忆 记忆 记忆 记忆 记忆 记忆 记忆 记忆 记忆And updating the data transfer function of the at least-counter, the rib is updating the at least one counter when the performance monitoring software wants to read at least one counter. 1. As described in the scope of the patent scope of the patent, the m-effect virtual performance monitoring device is implemented in parallel and separately to execute multiple simulators (mu test eaded simulati〇n), the overall simulation speed is equal to the speed The speed of the slow simulator. 12. If you apply for a patent scope! The virtual performance monitoring device for system performance monitoring further includes a timing device for obtaining a virtual execution time of the instruction stream, and the timing device is used to execute the standard inspection program on the target hardware platform. After the actual and virtual execution time on the virtual platform, the accuracy of the timing device is obtained by the following formula: ER=^^*i〇0% where ER represents the accuracy of the timing device, and p is the predicted value For the timing device, the standard inspection program is executed on the virtual platform - the virtual execution coffee; the ◦ is the value of it [S 19 201140313 for the timing (four) to obtain the standard inspection program in the s time standard hardware The old age - the actual execution * 'It:: set ==::=: virtual __ ^£7? = Xi=〇ll5il Κ where aER represents the flat solution rate of the device, κ _ standard check program total φ And ERi is the accuracy of the first standard inspection program. 14. A method for virtual performance monitoring for system performance monitoring, operating on a virtual platform, the virtual platform is simulated on an operating platform by an intermediary software-target hardware platform, and I simulates on the virtual platform a target software system and a performance monitoring software, the method comprising: Step A. receiving and monitoring an instruction flow executed on the virtual platform and accepting a performance monitoring command of the performance monitoring software; Step B: monitoring the software according to the performance The performance monitoring command and the instruction stream are used to simulate the soft and hardware of the specific φ to obtain a performance information; Step C: determine whether the performance information needs further calculation processing, and if yes, proceed to step D, if not, directly enter Step e: Step D: calculating the performance information, and generating a calculated performance information 'Entering step E; and Step E: storing the performance information. 15. The method for virtual performance monitoring for system performance monitoring as described in claim 14 further includes: f 20 201140313 Step F. After performing step E, transmitting the performance information to the performance monitoring software * Reference information for further monitoring and performance evaluation. • 16· The method of virtual performance monitoring for system performance monitoring as described in claim 14 or 15 of the patent scope, wherein the effect information includes: the number of instructions (丨耶丨加c(10)), virtual Execution time, number of clock cycles, number of cache accesses, and number of cache misses. 17. If you want to apply for the virtual performance monitoring method of the secret monitoring system, the step E is returned to step A after the execution of the step E is completed. φ I8. The method for virtual performance monitoring for system performance monitoring as described in item b of the patent application, wherein after the step F is completed, the process returns to step A. 21twenty one
TW99114377A 2010-05-05 2010-05-05 Virtual performance monitoring apparatus and method for using to monitor system performance TW201140313A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW99114377A TW201140313A (en) 2010-05-05 2010-05-05 Virtual performance monitoring apparatus and method for using to monitor system performance

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW99114377A TW201140313A (en) 2010-05-05 2010-05-05 Virtual performance monitoring apparatus and method for using to monitor system performance

Publications (1)

Publication Number Publication Date
TW201140313A true TW201140313A (en) 2011-11-16

Family

ID=46760260

Family Applications (1)

Application Number Title Priority Date Filing Date
TW99114377A TW201140313A (en) 2010-05-05 2010-05-05 Virtual performance monitoring apparatus and method for using to monitor system performance

Country Status (1)

Country Link
TW (1) TW201140313A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103186449A (en) * 2011-12-28 2013-07-03 英业达股份有限公司 Method for managing hardware efficiency
TWI501102B (en) * 2013-08-27 2015-09-21 Inst Information Industry Virtual time control apparatus, method, and computer program product thereof
TWI556601B (en) * 2013-05-06 2016-11-01 英派爾科技開發有限公司 Computing device performance monitor
CN112860529A (en) * 2019-11-28 2021-05-28 瑞昱半导体股份有限公司 Universal analysis device and method

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103186449A (en) * 2011-12-28 2013-07-03 英业达股份有限公司 Method for managing hardware efficiency
TWI556601B (en) * 2013-05-06 2016-11-01 英派爾科技開發有限公司 Computing device performance monitor
TWI501102B (en) * 2013-08-27 2015-09-21 Inst Information Industry Virtual time control apparatus, method, and computer program product thereof
CN112860529A (en) * 2019-11-28 2021-05-28 瑞昱半导体股份有限公司 Universal analysis device and method
CN112860529B (en) * 2019-11-28 2022-11-08 瑞昱半导体股份有限公司 Universal analysis device and method
US11507485B2 (en) 2019-11-28 2022-11-22 Realtek Semiconductor Corporation Universal profiling device and method for simulating performance monitoring unit
TWI791929B (en) * 2019-11-28 2023-02-11 瑞昱半導體股份有限公司 Universal profiling device and method

Similar Documents

Publication Publication Date Title
Austin et al. SimpleScalar: An infrastructure for computer system modeling
Gregg Systems performance: enterprise and the cloud
Patel et al. MARSS: A full system simulator for multicore x86 CPUs
Halpern et al. Mosaic: cross-platform user-interaction record and replay for the fragmented android ecosystem
Zhang et al. Panappticon: Event-based tracing to measure mobile application and platform performance
US20130326202A1 (en) Load test capacity planning
Wong et al. Parallel application signature for performance analysis and prediction
CN105630575B (en) For the performance estimating method of KVM virtualization server
US9418181B2 (en) Simulated input/output devices
Gebai et al. Fine-grained preemption analysis for latency investigation across virtual machines
TW201140313A (en) Virtual performance monitoring apparatus and method for using to monitor system performance
Ryckbosch et al. Fast, accurate, and validated full-system software simulation of x86 hardware
Fournier et al. Analyzing blocking to debug performance problems on multi-core systems
Lai et al. Fast profiling framework and race detection for heterogeneous system
Hung et al. System-wide profiling and optimization with virtual machines
Nambiar et al. Model driven software performance engineering: Current challenges and way ahead
Bartalos et al. Engineering energy-aware web services toward dynamically-green computing
US8160845B2 (en) Method for emulating operating system jitter
Ko et al. Hardware-in-the-loop simulation for CPU/GPU heterogeneous platforms
Reischer et al. Bio-algebras
TW201232408A (en) Cycle-count-accurate (CCA) processor modeling for system-level simulation
Wu et al. Cloud server benchmark suite for evaluating new hardware architectures
Sarma Abstracting AI Evaluation Environments with Virtual Machines
Tu et al. Mcemu: A framework for software development and performance analysis of multicore systems
Zhou et al. All about Sample-Size Calculations for A/B Testing: Novel Extensions & Practical Guide