公开(公告)号：JPH09319722A

公开(公告)日：1997-12-12

申请号：JP12806596

申请日：1996-05-23

Applicant: IBM

Inventor： SUGANUMA TOSHIO ,  KOMATSU HIDEAKI

IPC: G06F15/16 ,  G06F9/45

Abstract: PROBLEM TO BE SOLVED: To fast run a compiled object program by extracting correctly the reduction, i.e., a loop pattern that frequently appears in a source program out of this source program. SOLUTION: The reduction is detected out of a loop (Step 21) and then converted into a loop that can be carried out effectively and in parallel (Step 22). Finally, the reduction communication is generated based on the converted loop and optimized (Step 23). The detection of reduction consists of three sub- steps. That is, a model is first produced in regard to the mask expression masking every substitute sentence in the loop. In other words, an expression tree concerning the execution condition of every substitute sentence is generated. Then it's decided whether the reduction consists of an expression tree of merged substitute sentences based on the expression tree of a prescribed conditional expression and also the data dependence relation.

公开(公告)号：JP2007179161A

公开(公告)日：2007-07-12

申请号：JP2005374821

申请日：2005-12-27

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： KOSEKI SATOSHI ,  SUGANUMA TOSHIO ,  KOMATSU HIDEAKI

IPC: G06F12/00

CPC classification number: G06F17/3041 ,  G06F17/30471 ,  G06F17/30513 ,  Y10S707/99932 ,  Y10S707/99933

Abstract: PROBLEM TO BE SOLVED: To streamline the execution of a program by pre-fetching data necessary to execute it. SOLUTION: A system is provided for causing a computer to execute a recursive access instruction prior to a program subject to pre-fetching. In the system, query instructions to select records satisfying selection conditions from target tables and generate result tables are detected from iterative processing in the program. An initial query instruction to generate an initial table including variable values to be set prior to the start of the iterative processing in the program is generated. A recursive query instruction to generate, by each recursive query corresponding to the sequentially executed iterative processings, the next intermediate table to be referred to in the subsequent iterative processing from the target tables and an intermediate table including the result tables generated by the preceding target query instructions of the iterative processing is next generated. A final query instruction to generate a final table from the intermediate tables sequentially generated by the recursive queries is then generated. COPYRIGHT: (C)2007,JPO&INPIT

Abstract translation: 要解决的问题：通过预取执行它所需的数据来简化程序的执行。 解决方案：提供一种系统，用于使计算机在执行预取的程序之前执行递归访问指令。 在系统中，从程序中的迭代处理中检测到从目标表选择满足选择条件的记录和生成结果表的查询指令。 产生包括在程序中的迭代处理开始之前要设置的变量值的初始表的初始查询指令。 一种递归查询指令，用于通过与顺序执行的迭代处理相对应的每个递归查询来生成在目标表的后续迭代处理中要参考的下一个中间表，以及包括由先前目标查询生成的结果表的中间表 接下来生成迭代处理的指令。 然后生成从递归查询顺序生成的中间表生成最终表的最终查询指令。 版权所有（C）2007，JPO＆INPIT

公开(公告)号：JP2004272826A

公开(公告)日：2004-09-30

申请号：JP2003066010

申请日：2003-03-12

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： YASUE TOSHIAKI ,  SUGANUMA TOSHIO ,  KOMATSU HIDEAKI

IPC: G06F9/45

CPC classification number: G06F8/443

Abstract: PROBLEM TO BE SOLVED: To provide a compiler system which eliminates the trouble of requiring a long collection processing time even though the number of times for performing an execution path is appropriately collected, a compiler program, a recording medium, a compiling method, an execution time information generating device and an execution time information generation program. SOLUTION: This compiler system 10 for optimizing a program on the basis of the frequency at which each processing in a program of an optimization object is carried out is provided with a loop processing detecting part 110 for detecting the repeatedly executed loop processing in the program, a loop processing frequency collecting part 140 for collecting a loop processing frequency at which the loop processing is carried out in the program, an intra-loop processings frequency collecting part 150 for collecting an intra-loop processing frequency being a frequency at which a plurality of intra-loop processings included in the loop processing are respectively carried out with respect to the number of times when the loop processing is carried out, an intra-loop execution information generating part 160 for generating intra-loop execution information showing a frequency at which the plurality of intra-loop processings are respectively carried out when the program is performed on the basis of the loop processing frequency and the intra-loop processing frequency, and an optimizing part for optimizing the program on the basis of the intra-loop exectuion information. A frequency in which each processing of the program is carried out can be collected at a high speed. COPYRIGHT: (C)2004,JPO&NCIPI

公开(公告)号：JP2013140530A

公开(公告)日：2013-07-18

申请号：JP2012000972

申请日：2012-01-06

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Machines Corporation

Inventor： SUGANUMA TOSHIO ,  RAJARAM B KRISHNAMURTHY ,  GARY CHAPMAN

IPC: G06F9/50

CPC classification number: H04L47/25 ,  H04L67/02 ,  H04L67/10

Abstract: PROBLEM TO BE SOLVED: To maximize performance in distributed parallel computing using a plurality of accelerator devices.SOLUTION: A host computer comprises a profiler unit which performs performance analysis for data transmission speed and calculation speed for a range of a data size, a size optimization unit which determines a chunk size optimal for data transmission through a network on the basis of the data transmission speed and the calculation speed to which the performance analysis has been performed, an encapsulation unit which encapsulates data into a chunk with the optimal chunk size by dividing or aggregating the data of a certain size stored in memory of the host computer, and a dispatcher unit which distributes the encapsulated data to accelerator devices and gives the accelerator devices an instruction to perform pipeline calculation for the encapsulated data received.

Abstract translation: 要解决的问题：使用多个加速器设备最大化分布式并行计算中的性能。解决方案：主计算机包括对数据传输速度和数据大小范围的计算速度执行性能分析的分析器单元，尺寸优化 单元，其基于数据传输速度和执行性能分析的计算速度确定通过网络进行数据传输最佳的块大小;封装单元，其通过分割将数据封装成具有最佳块大小的块; 或聚合存储在主计算机的存储器中的一定大小的数据，以及调度单元，其将封装数据分发到加速器装置，并向加速器装置提供对接收到的封装数据执行流水线计算的指令。

公开(公告)号：JP2010061522A

公开(公告)日：2010-03-18

申请号：JP2008228286

申请日：2008-09-05

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： NAKAIKE TAKUYA ,  INAGAKI TATSUSHI ,  OGASAWARA TAKESHI ,  SUGANUMA TOSHIO

IPC: G06F9/46 ,  G06F9/52 ,  G06F12/00

CPC classification number: G06F9/526 ,  G06F9/3004 ,  G06F9/30087

Abstract: PROBLEM TO BE SOLVED: To provide a mechanism for stopping only a thread in which a fault is caused or a fault may be caused from another thread without stopping the whole system having a locked thread. SOLUTION: A computer system stores a program including a code for a critical section, wherein the critical section includes an instruction for writing or reading a value to/from a shared data area in a memory. The instruction acquires lock for the critical section before the start of an initial instruction in the critical section, writes a value in a thread local area in the memory instead of writing the value in the shared data area in response to a writing instruction to the shared data area, writes the value written in the thread local area in the shared data area in response to the end of a final instruction in the critical section, and unlocks the critical section. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：提供一种仅停止引起故障的线程或者可能由另一线程引起故障而不停止具有锁定线程的整个系统的机构。 解决方案：计算机系统存储包括关键部分的代码的程序，其中临界部分包括用于向/从存储器中的共享数据区域写入或读取值的指令。 该指令在关键部分中的初始指令开始之前获取关键部分的锁定，将值写入存储器中的线程局部区域中，而不是响应于对共享的写入指令将数据写入共享数据区域中 数据区域，响应于关键部分中的最终指令的结束，写入共享数据区域中的线程局部区域中写入的值，并解锁关键部分。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2009265840A

公开(公告)日：2009-11-12

申请号：JP2008112945

申请日：2008-04-23

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： SUGANUMA TOSHIO

IPC: G06F12/00

Abstract: PROBLEM TO BE SOLVED: To provide a cache system effective even in an insert only type application wherein a record of data imparted with the same identification information is added to a database every time creation, update, or deletion of the data is performed.
SOLUTION: A data holding part for caching the data read from the database includes: a hash table 410 registered with the identification information of the data; and a map 420 for storing a map entry showing the record individually associated to the identification information registered in the hash table 410, having the identification information in the database. The cached data are associated to the map entry by the identification information and timestamp information that is information for specifying the record, and are held.
COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：即使在仅插入型应用中也提供有效的缓存系统，其中，每当执行数据的创建，更新或删除时，向数据库中添加赋予相同标识信息的数据的记录 。 解决方案：用于缓存从数据库读取的数据的数据保持部分包括：与数据的识别信息一起登记的散列表410; 以及映射表420，用于存储表示与登记在散列表410中的识别信息单独关联的记录的映射条目，其具有数据库中的标识信息。 缓存的数据通过作为用于指定记录的信息的标识信息和时间戳信息与地图条目相关联，并被保持。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2004062830A

公开(公告)日：2004-02-26

申请号：JP2002224207

申请日：2002-07-31

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： SUGANUMA TOSHIO ,  OGASAWARA TAKESHI

IPC: G06F9/45 ,  G06F7/00

CPC classification number: G06F8/4441

Abstract: PROBLEM TO BE SOLVED: To avoid the waste of overhead and analytic processing by efficiently performing the precision assurance of a single precision operation and a double precision operation in a floating-point operation. SOLUTION: In a compiler 100 mounted on a computer, an object code of a method is generated in a double precision mode in the floating-point operation and a second code of the method is generated in a single precision mode when the method is frequently called in the single precision mode and processing costs can be reduced by implementing the method in the single precision mode. Otherwise, the object code is generated by setting operation precision according to a kind of an operation in a target method to be compiled irrespective of which of the double precision mode and the single precision mode in the floating-point operation is called, a code for matching the precision modes is generated and added based on a relation between the target code and other code for calling the target code or other code called by the target code. COPYRIGHT: (C)2004,JPO

公开(公告)号：JP2010086218A

公开(公告)日：2010-04-15

申请号：JP2008253637

申请日：2008-09-30

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： SUGANUMA TOSHIO ,  ARAI HIROSHI ,  KONO NORIAKI

IPC: G06F9/44 ,  G06F9/45 ,  G06F9/54

Abstract: PROBLEM TO BE SOLVED: To provide a data structure for satisfying specifications of JAVA code, executing the JAVA code with high speed and efficiency; and to provide an execution method of the JAVA code using the data structure, since there is a great difference in language specifications between a COBOL language and a JAVA (R) language, and when converting a COBOL code into a JAVA code, how the JAVA code efficiently achieves processing described on the COBOL code greatly depends on the design of the data structure. SOLUTION: A computer system executing a program including a JAVA code converted from a COBOL code includes a storage unit storing a program including a JAVA code having at least one declaration of an object of which the data type is a wrapper class that is declaration converted from each declaration of data items included in the COBOL code; and a CPU reading out the program from the storage unit to execute the program. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：为了提供满足JAVA代码规范的数据结构，以高速和高效地执行JAVA代码; 并且使用数据结构来提供JAVA代码的执行方法，因为COBOL语言和JAVA（R）语言之间的语言规范有很大差异，并且当将COBOL代码转换为JAVA代码时，JAVA如何 代码有效地实现了对COBOL代码描述的处理，大大取决于数据结构的设计。 解决方案：执行包括从COBOL代码转换的JAVA代码的程序的计算机系统包括存储单元，该存储单元存储包括JAVA代码的程序，JAVA代码具有数据类型是包装类的对象的至少一个声明， 从COBOL代码中包含的数据项的每个声明转换的声明; 以及从存储单元读出程序以执行程序的CPU。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2007066017A

公开(公告)日：2007-03-15

申请号：JP2005251594

申请日：2005-08-31

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： SUGANUMA TOSHIO ,  KOSEKI SATOSHI ,  KOMATSU HIDEAKI

IPC: G06F12/00

CPC classification number: G06F17/30395 ,  G06F17/30454 ,  Y10S707/99933

Abstract: PROBLEM TO BE SOLVED: To furthermore improve the efficiency of access to a database. SOLUTION: A system for controlling access to the database is provided with: an acquisition part for acquiring an access program including a plurality of access instructions to the database; a detection part for detecting a plurality of mutually different tables to be accessed by the plurality of access instructions; a combination control part for executing a combination instruction for acquiring a result table obtained by combining the plurality of detected tables from the database prior to the execution of the plurality of access instructions; and a conversion control part for executing a conversion instruction for extracting records accessed by the plurality of access instructions from the result table and converting the extracted records into structure allowed to be accessed by the plurality of access instructions prior to the execution of the plurality of access instructions. COPYRIGHT: (C)2007,JPO&INPIT

Abstract translation: 要解决的问题：进一步提高访问数据库的效率。 解决方案：用于控制对数据库的访问的系统提供有：获取部分，用于获取包括对数据库的多个访问指令的访问程序; 检测部，用于检测由所述多个访问指令访问的多个相互不同的表; 组合控制部分，用于执行用于获取通过在执行所述多个访问指令之前从所述数据库组合所述多个检测表而获得的结果表的组合指令; 以及转换控制部分，用于执行用于从结果表中提取由多个访问指令访问的记录的转换指令，并且将所提取的记录转换为在执行多个访问之前由多个访问指令允许访问的结构 说明。 版权所有（C）2007，JPO＆INPIT

公开(公告)号：JP2006163596A

公开(公告)日：2006-06-22

申请号：JP2004351492

申请日：2004-12-03

Applicant: Internatl Business Mach Corp ,  インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Maschines Corporation

Inventor： SUGANUMA TOSHIO ,  KOMATSU HIDEAKI

IPC: G06F12/00

CPC classification number: G06F17/30377 ,  Y10S707/99938

Abstract: PROBLEM TO BE SOLVED: To accelerate transaction processing to a data base from before.
SOLUTION: The information processing system which performs the transaction processing to the data base includes a processing program acquiring section for acquiring a processing program which is program transaction processing to the data base being described to be, an object data selecting section which selects at least one object data used as an object accessed by transaction processing of the transaction program of the data base, and a rewriting processing insertion section which inserts object data rewriting processing which rejects the object data updated by the transaction processing among the object data to the data base into the section performed after the accessing object data finally before committing processing result in the processing program.
COPYRIGHT: (C)2006,JPO&NCIPI

Abstract translation: 要解决的问题：从之前加速到数据库的事务处理。 解决方案：对数据库执行事务处理的信息处理系统包括：处理程序获取部分，用于获取对被描述为数据库的程序事务处理的处理程序，对象数据选择部分选择 用作通过数据库的交易程序的事务处理访问的对象的至少一个对象数据，以及重写处理插入部，其将对象数据重写处理插入到对象数据中，所述对象数据重写处理将对象数据中的由事务处理更新的对象数据拒绝到 数据库进入到处理程序结束之后访问对象数据之后执行的部分。 版权所有（C）2006，JPO＆NCIPI