2010年9月11日

What’s Not Parallel

摘要: Algorithms with State Recurrences Induction Variables Reduction Loop-Carried Dependence Don't use something that is too far afield of the original, though. Less efficient serial algorithms will tend ... 阅读全文

posted @ 2010-09-11 17:00 胡是 阅读(140) 评论(0) 推荐(0) 编辑

并行编程的数据分解模型

摘要: How you divide data structures into contiguous subregions, or "chunks," of data? an array. You can divide arrays along one or more of their dimensions. Other structures that use an array as a compone... 阅读全文

posted @ 2010-09-11 10:44 胡是 阅读(773) 评论(0) 推荐(0) 编辑

Step 4. Tune for Performance: Removing Performance Bottlenecks

摘要: After making sure that you have removed all the threading (and new logic) errors from your code, the final step is to make sure the code is running at its best level of performance. (这是第四步的目的。) Before... 阅读全文

posted @ 2010-09-11 09:55 胡是 阅读(269) 评论(0) 推荐(0) 编辑

Step 3. Test for Correctness: Detecting and Fixing Threading Errors

摘要: Whenever you make code changes to an application, you open the door to the possibility of introducing bugs. Adding code to a serial application in order to generate and control multiple threads is no ... 阅读全文

posted @ 2010-09-11 09:28 胡是 阅读(250) 评论(0) 推荐(0) 编辑

2010年9月10日

gnulib components used by Octave

摘要: libgnu编译后的模块图 阅读全文

posted @ 2010-09-10 08:59 胡是 阅读(153) 评论(0) 推荐(0) 编辑

2010年9月8日

C++ namespace magic for gnulib (zZ from gnu.org)

摘要: The function definitions provided by Gnulib (.c code) are meant to be compiled by a C compiler. The header files (.h files), on the other hand, can be used in either C or C++. By default, when used in... 阅读全文

posted @ 2010-09-08 22:00 胡是 阅读(392) 评论(0) 推荐(0) 编辑

Writeup - A GPGPU Compiler for Memory Optimization and Parallelism Management

摘要: 在《The Art of Concurrency》这本书中,作者介绍了在多核处理器上进行多线程程序开发的方法,可分为四个步骤,即分析(Analysis)、设计和实现(Design and Implementation)、正确性检查(Test for correctness)和性能调谐(Tune for performance)。在已有的串行程序基础上,程序员反复进行以上四个步骤,以达到满意的性能提... 阅读全文

posted @ 2010-09-08 13:45 胡是 阅读(239) 评论(0) 推荐(0) 编辑

改进论文标题的6个技巧

摘要: 把贡献置于题名的前部 新信息出现在开头,人们已知的不太特指的信息则位于句子的结尾 增加动词形式 分词和不定式的动词形式可给题名增加活力 用形容词和数字表述贡献的亮点 fast, highly efficient, robust but not new, novel 清楚且特指的关键词 论文内容的具体程度与题名中具有特指性关键词的数量成正比。不能把关键词埋没在一长串的修饰性作用的名词中。 精... 阅读全文

posted @ 2010-09-08 09:15 胡是 阅读(276) 评论(0) 推荐(0) 编辑

2010年9月7日

如何把任务分配给线程

摘要: We can allocate tasks to threads in two different ways: static scheduling or dynamic scheduling. NOTEUnder worksharing constructs in OpenMP and the parallel algorithms of Intel Threading Building Blo... 阅读全文

posted @ 2010-09-07 15:05 胡是 阅读(601) 评论(0) 推荐(0) 编辑

任务间相关性约束

摘要: Two types of dependencies can occur between tasks. The first is order dependency, where some task relies on the completed results of the computations from another task. This reliance can be a direct ... 阅读全文

posted @ 2010-09-07 14:38 胡是 阅读(246) 评论(0) 推荐(0) 编辑

导航