Call for participation in the 14th Meetings for application code tuning on A64FX computer systems
HPC WORLD MLの皆様
See below for the English version.
高度情報科学技術研究機構 (RIST)の太田幸宏と申します。MLを
お借りして、オンラインセミナー「第14回A64FX向けチューニング
技術検討会」の案内をさせていただきます。重複して受け取られた
場合はご容赦ください。
理化学研究所計算科学研究センター (RIKEN R-CCS）と高度情報科学
技術研究機構（RIST）は、共同でA64FX向けチューニング技術検討会
を開催しています。本検討会の目的は、A64FX向けのチューニング
技法や、開発環境などの情報を共有し、A64FXアーキテクチャシステム
向けのアプリケーション開発を促進することです。
皆様の参加をお待ちしています。
=======================================================================
# 第14回A64FX向けチューニング技術検討会
[日時] 2024年6月4日(火) 3:00pm – 5:00pm
[話題]
1. SIMD acceleration of the particle-based simulation and Vlasov simulation on A64FX processor
Speakers: Kohji Yoshikawa (University of Tsukuba)
Abstract: SIMD operations are an essential feature to extract
high performance from modern processors, but their effective use
in high-level languages such as C/C++ is not straightforward.
In this talk, I present examples of optimisation using SIMD
operations in particle-based simulations and Vlasov simulations
on 6D phase space, which our research group has performed
at Fugaku and other supercomputer facilities.
2. Acceleration of protein 3D structure prediction AI (AF2) powered by “Fugaku” for Genome2MD project
Speakers: Atsushi Tokuhisa (RIKEN R-CCS)
Abstract: We have been conducting research on comprehensive
structural analysis of the effects of human genome variants
using “Fugaku”. In this research, the throughput of AI
to estimate 3D structures of biomolecules from amino acid
sequences is important. In this seminar, we report that
our developed “OpenFold-for-Fugaku” dramatically improves
the throughput of protein 3D structure estimation by using
multiple nodes of “Fugaku”. We believe that this achievement
will make genome-scale simulation feasible. OpenFold-for-Fugaku
can be downloaded from:
https://github.com/RIKEN-RCCS/OpenFold-for-Fugaku.
[参加申込]
https://www.hpci-office.jp/events/seminars/meeting_A64FX_240604
締め切り: 2024年5月30日（木）
[開催方法]
* Zoomを用いたオンラインでの開催。
* 接続URLは参加申込をされた方に、開催30分前までにメールで案内します。
[言語] 資料: 英語、講演: 英語、質疑応答: 英語/日本語
[お願い]
* 質問等は、チャット機能により、お寄せください。
* 本ミーティングは録画されます。録画された動画はWebサイト上で公開されます。予めご了承ください。
[アーカイブ]
過去の資料及び音声を含む録画は、以下のウェブサイトで公開されています。
https://www.hpci-office.jp/en/events/seminars#A64FX
[問い合わせ]
E-mail: hpc-support-admin[-at-]hpci-office.jp
[-at-]を@に置き換えてください。
—
Dear all,
We are inviting you to attend our upcoming webinar, the 14th
meetings for application code tuning on A64FX computer systems.
RIKEN Center for Computational Science (RIKEN R-CCS) and Research
Organization for Information Science and Technology (RIST)
collaboratively organize a series of meetings for application
code tuning on A64FX computer systems. The aim is to share the
techniques for getting application codes more efficient on A64FX
computer systems so that the number of application programs
that work well on the computer systems increases.
We look forward to your participation.
========================================================================
# The 14th Meeting for application code tuning on A64FX computer systems
[Date] 3:00pm ? 5:00pm (JST), Tuesday, June 4th, 2024
[Topics]
1. SIMD acceleration of the particle-based simulation and Vlasov simulation on A64FX processor
Speakers: Kohji Yoshikawa (University of Tsukuba)
Abstract: SIMD operations are an essential feature to extract
high performance from modern processors, but their effective use
in high-level languages such as C/C++ is not straightforward.
In this talk, I present examples of optimisation using SIMD
operations in particle-based simulations and Vlasov simulations
on 6D phase space, which our research group has performed
at Fugaku and other supercomputer facilities.
2. Acceleration of protein 3D structure prediction AI (AF2) powered by “Fugaku” for Genome2MD project
Speakers: Atsushi Tokuhisa (RIKEN R-CCS)
Abstract: We have been conducting research on comprehensive
structural analysis of the effects of human genome variants
using “Fugaku”. In this research, the throughput of AI
to estimate 3D structures of biomolecules from amino acid
sequences is important. In this seminar, we report that
our developed “OpenFold-for-Fugaku” dramatically improves
the throughput of protein 3D structure estimation by using
multiple nodes of “Fugaku”. We believe that this achievement
will make genome-scale simulation feasible. OpenFold-for-Fugaku
can be downloaded from:
https://github.com/RIKEN-RCCS/OpenFold-for-Fugaku.
[Registration]
https://www.hpci-office.jp/en/events/seminars/meeting_A64FX_240604
The registration will be closed on Thursday, May 30th, 2024 in JST.
[On the meeting]
* We will use Zoom for the meeting.
* URL for the connection is emailed by 30 minutes before to the registrants.
[Language] Material: English, Talk: English, Q/A: English/Japanese
[Remarks]
* Please use the chat feature to ask your questions in the meeting.
* The meeting will be recorded. Please note that the recorded movie will be opened in the website.
[Archive]
The presentation materials and the video recordings including voices
in the meeting would be opened on the following website later.
https://www.hpci-office.jp/en/events/seminars#A64FX
[Contact]
E-mail: hpc-support-admin[-at-]hpci-office.jp
(Please replace [-at-] with @.)
