主催：NII 先端ソフトウェア工学国際研究センター（GRACEセンター）
http://grace-center.jp/
日時：2010年5月21日（金）10:00-12:00
場所：国立情報学研究所(NII) 20階講義室1(2005) （地図）
参加費：無料
お問い合わせ：加藤 弘之 (kato_AT_nii.ac.jp)_AT_を＠に書き換えてください。
参加をご希望の方は、セミナー前日までに下記よりご登録をお願いします。
http://grace-center.jp/regist/seminar
===========================================================
FIRST SPEAKER
Tetsuo Yokoyama (Nanzan University)
=== Title ===
Optimization of Input-Erasing Clean Reversible Simulation for Injective Functions
=== Abstract ===
Bennett showed that a clean reversible simulation of injective programs is possible without returning the input of a program as additional output. His method makes use of two computation and two uncomputation phases. This paper argues that for a certain class of injective programs only half of the computation and uncomputation steps are necessary. A practical consequence is that the reversible simulation runs twice as fast as Bennett’s simulation. The method is demonstrated by developing lossless encoders and decoders for run-length encoding and range coding. The range coding program is further optimized by conserving models during the text generation. This may thus provide a new view on developing efficient reversible simulations for a certain class of injective functions.
SECOND SPEAKER
Holger Bock Axelsen, Department of Computer Science, DIKU, University of Copenhagen.
Currently visiting the Graduate School of Information Science and Technology, IST, University of Tokyo.
=== Title ===
Clean Translation of Reversible Programming Languages
=== Abstract ===
I shall describe the major translation techniques used in a compiler from the high-level reversible imperative language Janus to the low-level reversible assembly language PISA. The translation is
semantics-preserving: A target program computes the same function as the source program (modulo data representation) with no generation of so-called garbage data in the output, so the translation is “clean”.
Furthermore, the translation is clean down to the level of individual Janus
statements: A target program will not accumulate garbage data across neither
(translated) statement sequences nor in nested statements. The translation methods are generic, and can be applied to any (imperative) source language describable with reversible flowcharts and reversible updates.
