Differences between revisions 43 and 83 (spanning 40 versions)
Revision 43 as of 2018-11-15 02:00:24
Size: 2611
Editor: kko
Comment:
Revision 83 as of 2018-12-19 09:08:14
Size: 1234
Editor: sato
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
= Temp = = Introduction to Our Research Topics (研究紹介) =
Line 3: Line 3:
Please place your research introductions temporarily under this section. == High-Level, High-Performance General-Purpose Programming Frameworks ==
Line 5: Line 5:
 * [[並列パーサ]]  * [[MassiveThreads|MassiveThreads: Super Lightweight Thread Library]]
 * [[高性能なタスク並列スケジューラ]]
 * [[Research/分散共有メモリ|分散共有メモリ]]
 * [[PerformanceAnalysisTools]]
 * [[MPI+ULT]]
 * [[Research/Lightweight Threading Library and OpenMP Runtime System|Lightweight Threading Library and OpenMP Runtime System]]
 * [[Research/GXP|GXP Parallel Shell/Scripting Tool]]
Line 7: Line 13:
= Research = == Very High-Level, High-Performance Domain-Specific Programming Frameworks ==
Line 9: Line 15:
([[新4年生向紹介文]]もどうぞ)  * [[大規模な文字列データ向けのアドホックな並列処理系]]
 * [[Research/CFG構文解析|CFG構文解析]]
 * [[ParaLite|ParaLite: Parallel Database]]
 * [[Research/ParallelPattern|並列パターン]]
Line 11: Line 20:
Some of the ongoing efforts are described below. == Deep Learning (深層学習) ==
Line 13: Line 22:
 MassiveThreads: Super Lightweight Thread Library::
 :: A thread library that can spawn threads two orders of magnitude faster than native operating system threads. Based on this technology, we are working on high-level parallel programming languages that target a range of machines from multi-core computers to massively parallel processors
 (https://github.com/massivethreads/massivethreads).
 * [[Analysis of Large Mini-Batch Training of Neural Networks]]
 * [[Accelerating Neural Networks Having Dynamic Computation Graphs]]
 * [[深層学習フレームワークの性能解析]]
Line 17: Line 26:
 ParaLite: Parallel Database::
 :: A parallel database that can easily parallelize workflow applications consisting of existing commands, using SQL syntax. We are building a large scale data processing system based on ParaLite.
== Storage and Operating System(ストレージとOS) ==
Line 20: Line 28:
 GXP Parallel Shell/Scripting Tool::
 :: A parallel processing tool that enables parallel processing built on existing commands, no matter where you are and where your resources are (http://www.logos.t.u-tokyo.ac.jp/gxp/, https://github.com/qnu/gxp).

 * [[http://www.eidos.ic.i.u-tokyo.ac.jp/labdocs/intros/2012/m_intro/taura_lab_2012e.pdf|Introduction to the Laboratory (pdf)]]

= 研究 =

 MassiveThreads超軽量スレッドライブラリ::
 :: 通常のスレッドよりも二桁高速にスレッドを作れるスレッドライブラリで
 す. それを土台にして, マルチコア計算機から, それらを多数結合した超並列
 機までを,高水準にプログラミングできる言語を作る計画です
 (https://github.com/massivethreads/massivethreads).

 ParaLite並列データベース::
 :: SQLが持つ高水準なデータ処理と, 既存のプログラムを容易に統合でき,か
 つそれを並列に実行できる並列データベースです. 大規模データ処理のための
 基盤システムです.

 GXP並列シェル・スクリプティング環境::
 :: 既存のプログラムを組み合わせた並列処理を,クラスタ, クラウドなど, 所
 構わずどこでも行えるようにした,並列処理ツールです
 (http://www.logos.t.u-tokyo.ac.jp/gxp/, https://github.com/qnu/gxp).

 * [[http://www.eidos.ic.i.u-tokyo.ac.jp/labdocs/intros/2012/m_intro/taura_lab_2012.pdf|研究室紹介(pdf)]]

 * [[http://www.eidos.ic.i.u-tokyo.ac.jp/labdocs/intros/2012/b3_intro/b3_intro_poster_2012.pdf|卒論配属説明会ポスター(pdf)]],[[http://www.eidos.ic.i.u-tokyo.ac.jp/labdocs/intros/2012/b3_intro/b3_intro_poster_2012.pptx|(pptx)]],
 * [[次世代SSDによるVMMを用いたメモリ空間の拡張]]

Taura Laboratory: Research (last edited 2018-12-19 09:08:14 by sato)