Open davidbowler opened 4 years ago
Hi,
Not so sure if it would be feasible without much changes to the code, probably using MPI-IO. But how about outputting everything S,H,P, etc. into one matrix each? When having a large number of nodes, I find it a bit inconvenient to have thousands of matrices.
Best, Marius
マリウス ビュクル, Ph.D.
bI技gt合研究所
C能材料コンピュテショナルデザイン研究センタ
305-8568 茨城hつくば市梅@ 1-1-1 つくば中央第2
TEL: 029-862-6947 FAX: 029-861-3171 E-mail: marius.buerkle@aist.go.jp
Marius Buerkle, Ph.D.
Research Center for Computational Design of Advanced Functional Materials, AIST
Central2, 1-1-1, Umezono, Tsukuba, 305-8568, Japan
TEL: 029-862-6947 FAX: 029-861-3171 E-mail: marius.buerkle@aist.go.jp
From: David Bowler notifications@github.com Sent: Tuesday, January 28, 2020 23:25 To: OrderN/CONQUEST-release CONQUEST-release@noreply.github.com Cc: Subscribed subscribed@noreply.github.com Subject: [OrderN/CONQUEST-release] Analysis of electronic structure for large systems (#28)
We will enable the analysis of electronic structure from large-scale linear scaling calculations via the Sakurai-Sugiura method. This is likely to have two stages:
We have the output of matrices largely implemented already.
― You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/OrderN/CONQUEST-release/issues/28?email_source=notifications&email_token=ADINPWS3II3Y5IXVSRLO6X3RAA56LA5CNFSM4KMTKKEKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IJHTYCQ, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ADINPWVEL2XBQMSLFYMODEDRAA56LANCNFSM4KMTKKEA.
I suspect that it would be significantly easier to bring the files together in a post-processing step (which we do with the charge density, for instance). Writing out as one file with large systems (high core counts) would probably incur very long wait times.
We will enable the analysis of electronic structure from large-scale linear scaling calculations via the Sakurai-Sugiura method. This is likely to have two stages:
We have the output of matrices largely implemented already.