-
公开(公告)号：KR101613971B1
公开(公告)日：2016-04-21
申请号：KR1020090134367
申请日：2009-12-30
Applicant: 삼성전자주식회사 , 서울대학교산학협력단
Abstract: 복수개의워크아이템들이하나의컴퓨팅유닛에포함된복수개의프로세싱엘리먼트에각각할당되어동시에수행되도록작성된프로그램코드의변환방법이제공된다. 프로그램코드를변환하는변환기는, 프로그램코드에대하여, 복수개의워크아이템이복수개의프로세싱엘리먼트보다적은개수의프로세싱엘리먼트에서직렬적으로수행되도록, 프로그램코드에포함된동기화배리어함수를기준으로워크아이템합체루프로감싸질 2 이상의프로그램코드영역을식별하는동작과, 식별된프로그램코드영역각각에대하여워크아이템합체루프로감싸는동작을수행한다.
-
公开(公告)号：KR1020120082295A
公开(公告)日：2012-07-23
申请号：KR1020110003707
申请日：2011-01-13
Applicant: 삼성전자주식회사 , 서울대학교산학협력단
CPC classification number: G06F8/4441 , G06F9/45529 , G06F8/37 , G06F15/80
Abstract: PURPOSE: A web browsing method through storing/optimizing a java script code and an apparatus thereof are provided to store type information, a stub code and a machine code generated in a runtime and perform gradual optimization through a spare core, thereby reducing a compile overhead of a java script. CONSTITUTION: A compiler(330) compiles a java script of a visited website. A storage(340) stores type information and a code generated by compiling the java script. A information confirmation unit(320) confirms the information about the java script. An optimizing module(350) performs a normal state compile by using the code in order to generate a optimized code. The compiler and the optimizing module are separately or simultaneously operated at separate cores.
Abstract translation: 目的：提供通过存储/优化java脚本代码的网页浏览方法及其装置来存储在运行时间中生成的类型信息，存根代码和机器代码，并通过备用核进行逐步优化，从而减少编译开销 的java脚本。 构成：编译器（330）编译被访网站的java脚本。 存储（340）存储类型信息和通过编译java脚本生成的代码。 信息确认单元（320）确认关于java脚本的信息。 优化模块（350）通过使用代码执行正常状态编译以便生成优化的代码。 编译器和优化模块分开或同时在独立的核心上运行。
-
公开(公告)号：KR1020110077720A
公开(公告)日：2011-07-07
申请号：KR1020090134367
申请日：2009-12-30
Applicant: 삼성전자주식회사 , 서울대학교산학협력단
Abstract: PURPOSE: A conversion method of a program code is provided to convert an open CL kernel code made to be processed in a plurality of processing elements so that the open CL kernel code can be efficiently used in a device with a limited number of processing elements. CONSTITUTION: A kernel code converter identifies two or more code areas to be surrounded by a WCL(Work-item Coalescing Loop) on the basis of a synchronization barrier function included in a program code(310). The kernel code converter extends the private variable of each work item(320). The kernel code converter surrounds a kernel code area identified as a WCR(Work-item Coalescing Region) by the WCL(330).
Abstract translation: 目的：提供程序代码的转换方法来转换在多个处理元件中进行处理的打开的CL内核代码，使得可以在具有有限数量的处理元素的设备中有效地使用打开的CL内核代码。 构成：内核代码转换器基于包括在程序代码（310）中的同步屏障功能来识别要被WCL（工作项合并循环）包围的两个或更多个代码区域。 内核代码转换器扩展每个工作项的私有变量（320）。 内核代码转换器围绕由WCL（330）识别为WCR（工作项目合并区域）的内核代码区域。
-
公开(公告)号：KR101639672B1
公开(公告)日：2016-07-15
申请号：KR1020100000614
申请日：2010-01-05
Applicant: 삼성전자주식회사 , 서울대학교산학협력단
CPC classification number: G06F9/528 , G06F9/3004 , G06F9/30087 , G06F9/3857 , G06F12/0811 , G06F12/0815 , G06F12/0864
Abstract: 오버플로우된데이터의처리가가능한무한트랜잭션메모리시스템이제공된다. 무한트랜잭션메모리시스템은호스트프로세서, 메모리및 메모리프로세서를포함한다. 호스트프로세서는트랜잭션을실행하는실행유닛및 데이터를임시저장하는캐시를포함한다. 메모리프로세서는트랜잭션실행중 캐시에서데이터가오버플로우되는오버플로우이벤트가발생되면, 오버플로우이벤트에응답하여, 오버플로우된데이터를메모리에포함된오버플로우스토리지에저장한다.
-
公开(公告)号：KR101782995B1
公开(公告)日：2017-09-29
申请号：KR1020110003707
申请日：2011-01-13
Applicant: 삼성전자주식회사 , 서울대학교산학협력단
CPC classification number: G06F8/4441 , G06F9/45529
Abstract: 멀티코어프로세서를사용하는모든시스템, 특히스마트폰과같이프로세서의계산능력이상대적으로떨어지고플래시와같은빠른저장장치를갖는시스템에서사용자의웹 브라우징경험을향상시키는웹 브라우징방법및 장치에관한것으로, 최적화한기계어코드를파일에저장하여점진적인최적화를달성하고이를통해웹 어플리케이션의자바스크립트프로그램은작은컴파일오버헤드를가짐과동시에매우빠르게수행할수 있다.
-
Patent Agency Ranking
公开(公告)号：KR1020110080412A
公开(公告)日：2011-07-13
申请号：KR1020100000614
申请日：2010-01-05
Applicant: 삼성전자주식회사 , 서울대학교산학협력단
CPC classification number: G06F9/528 , G06F9/3004 , G06F9/30087 , G06F9/3857 , G06F12/0811 , G06F12/0815 , G06F12/0864
Abstract: PURPOSE: An infinite transaction memory system and an operation method thereof are provided to rapidly access a memory in which overflow data is stored. CONSTITUTION: A host processor(110A,110B,110C,110D) includes an execution unit executing transaction and a cache temporarily storing data. In case an overflow event occurs in the execution of transaction, a memory processor(120) stores the overflow data in an overflow storage included in a memory(130) in response to the overflow event. The overflow event is the generation of data which is overflowed in the cache.
Abstract translation: 目的：提供无限交易存储器系统及其操作方法以快速访问其中存储溢出数据的存储器。 构成：主处理器（110A，110B，110C，110D）包括执行事务的执行单元和临时存储数据的高速缓存。 在执行事务时发生溢出事件的情况下，存储器处理器（120）响应于溢出事件将溢出数据存储在包括在存储器（130）中的溢出存储器中。 溢出事件是在缓存中溢出的数据的生成。
-
公开(公告)号：KR101775029B1
公开(公告)日：2017-09-05
申请号：KR1020160175915
申请日：2016-12-21
Applicant: 삼성전자주식회사
Abstract: 여러어플리케이션(application)이공유하여사용하는멀티코어시스템상의스케쥴링에관한기술이개시된다. 본발명의일 양상에따라, 코어가동작하는동안의존성제거(dependency resolving) 작업및 워크(work)의탐색시간을중첩시킬수 있는기술을제공할수 있다.
-
公开(公告)号：KR1020120051419A
公开(公告)日：2012-05-22
申请号：KR1020100112850
申请日：2010-11-12
Applicant: 삼성전자주식회사
IPC: G06F17/21
CPC classification number: G06F17/227
Abstract: PURPOSE: An apparatus for extracting a CSS(Cascading Style Sheet) rule and a method thereof are provided to shorten times to extract a CSS rule by extracting a CSS rule applied to a Dom node based on identification marks included in an ID(Identification) list. CONSTITUTION: A generating unit(130) generates an ID(Identification) list including identification marks which are matched with a DOM(Document Object Model) node and a chooser. A rule extracting unit(150) selects a target DOM node from DOM nodes and extracts a CSS(Cascading Style Sheet) rule applied to the target DOM node based on identification marks included in the ID list. The generating unit includes an ID List generation unit(131) which generates the ID list matched with the DOM node whenever the DOM node is generated.
Abstract translation: 目的：提供一种用于提取CSS（级联样式表）规则的方法及其方法，以缩短时间以提取CSS规则，方法是根据包含在ID（识别）列表中的识别标记提取应用于Dom节点的CSS规则 。 构成：生成单元（130）生成包括与DOM（文档对象模型）节点和选择器匹配的识别标记的ID（识别）列表。 规则提取单元（150）从DOM节点中选择目标DOM节点，并且基于ID列表中包含的识别标记提取应用于目标DOM节点的CSS（级联样式表）规则。 生成单元包括ID列表生成单元（131），每当生成DOM节点时，生成与DOM节点匹配的ID列表。
-
公开(公告)号：KR101375836B1
公开(公告)日：2014-04-01
申请号：KR1020070063135
申请日：2007-06-26
Applicant: 삼성전자주식회사
Inventor： 조승모
IPC: G06F15/16
CPC classification number: G06F9/5027 , G06F9/5033 , G06F9/5038 , G06F2209/5012
Abstract: 본 발명은 멀티코어 프로세서 상에서 연관된 작업의 수행 방법 및 그 장치에 관한 것으로, 복수 개의 코어들을 포함하는 멀티코어 프로세서 상에서 적어도 하나의 연관된 작업들을 수행하는 방법에 있어서, 작업들의 수행에 요구되는 데이터 및 주소 정보가 멀티코어 프로세서내의 각각의 코어들로 로딩되었는지 여부를 판단하고, 판단 결과에 기초하여 상기 멀티코어 프로세서내의 각각의 코어들이 동시에 작업을 개시하도록 제어함으로써, 연관된 작업들이 수행되는 동안 주 메모리에 불필요하게 접근하는 것을 방지함으로써 시스템의 성능을 향상시킨다.
멀티코어 시스템, 동기화-
公开(公告)号：KR1020120064446A
公开(公告)日：2012-06-19
申请号：KR1020100125688
申请日：2010-12-09
Applicant: 삼성전자주식회사
CPC classification number: G06F9/322 , G06F9/3806 , G06F9/45508 , G06F9/32 , G06F8/40
Abstract: PURPOSE: Apparatus and method for processing branch of bytecode on computing system are provided to aid branch prediction of a host processor treating a virtual imaginary machine like Java or Dalvik, thereby minimizing the failure of branch prediction. CONSTITUTION: Apparatus and method for processing branch of bytecode comprises: an interpreter conducting program to bytecode shape in a virtual machine; a branch information generator(320) acquiring branch information from calculated path of the branch in advance; and a BTB updater(330) updating the branch target buffer(341) on computing system by acquired branch information.
Abstract translation: 目的：提供用于处理计算系统上字节码分支的装置和方法，以帮助处理虚拟虚拟机（如Java或Dalvik）的主处理器的分支预测，从而最小化分支预测的失败。 构成：用于处理字节码分支的装置和方法包括：解释器，在虚拟机中执行字节码形状的程序; 分支信息发生器（320）预先从分支的计算路径获取分支信息; 以及BTB更新器（330），通过获取的分支信息更新计算系统上的分支目标缓冲器（341）。
-
-
-
-
-
-
-
-
-
Search Results
26
records
(took 0.024 seconds)
