Subject: Re: C502: (ACES Special Issue) Parallel Iterative Solvers for Unstructured Grids using Directive/MPI Hybrid Programming Model for GeoFEM Platform on SMP Cluster Architectures From: Kengo NAKAJIMA Date: Sun, 21 Oct 2001 12:20:58 +0900 To: Geoffrey Fox CC: Hiroshi Okuda X-UIDL: 9d04697fd6930200 X-Mozilla-Status: 0011 X-Mozilla-Status2: 00000000 Received: by mailer.csit.fsu.edu (mbox gcfpc) (with Cubic Circle's cucipop (v1.31 1998/05/13) Mon Oct 22 08:11:41 2001) X-From_: fox@mailer.csit.fsu.edu Mon Oct 22 08:11:37 2001 Return-Path: Delivered-To: gcfpc@csit.fsu.edu Received: from dirac.csit.fsu.edu (dirac.csit.fsu.edu [144.174.128.44]) by mailer.csit.fsu.edu (Postfix) with ESMTP id CA5DA23A02 for ; Mon, 22 Oct 2001 08:11:36 -0400 (EDT) Received: from localhost by dirac.csit.fsu.edu (AIX4.2/UCB 8.7) id IAA72982; Mon, 22 Oct 2001 08:11:35 -0400 (EDT) Resent-Message-Id: <200110221211.IAA72982@dirac.csit.fsu.edu> Delivered-To: fox@csit.fsu.edu Received: from mask (mask.uits.indiana.edu [129.79.6.184]) by mailer.csit.fsu.edu (Postfix) with ESMTP id 921FE23A11 for ; Sat, 20 Oct 2001 23:21:02 -0400 (EDT) Received: from tokyo.rist.or.jp (rist01.tokyo.rist.or.jp [202.241.71.24]) by mask (8.10.1/8.10.1/IUPO) with ESMTP id f9L3IDS18525 for ; Sat, 20 Oct 2001 22:18:13 -0500 (EST) Received: from nakajima.tokyo.rist.or.jp (rt01.tokyo.rist.or.jp [202.241.71.34]) by tokyo.rist.or.jp (8.9.3/3.7W) with SMTP id MAA18567; Sun, 21 Oct 2001 12:20:40 +0900 (JST) Message-Id: <200110210320.AA01142@nakajima.tokyo.rist.or.jp> In-Reply-To: <3B6DBB1C.3030606@csit.fsu.edu> References: <3B6DBB1C.3030606@csit.fsu.edu> MIME-Version: 1.0 X-Mailer: AL-Mail32 Version 1.12 Content-Type: multipart/mixed; boundary="--------------------5280707899380633" Resent-To: Geoffrey Fox Resent-Date: Mon, 22 Oct 2001 08:11:31 -0400 Resent-From: Geoffrey Fox Dear Geoffrey, This is Kengo Nakajima. I would like to show you how I corrected my paper (C502) according to the comments of the referee : >Referee 1 ************************************************************************************* >* It is stated that to increase performance with a node or PE, loops should be sufficiently >long so that data or instructions can be efficiently pipelined through each PE. I agree >with this statement but I think a comment should be added on the effect of the length or >complexity of the body of the loops. The length or the complexity of the code within a >loop can be an important factor. If the body of a loop is too large the PEs may run out of >registers and/or may not be able to pipeline the instructions efficiently. I put comments on loop length in p.9-p.10. I have already included the comments on the compiler of SR8000 in the 4th paragraph of p.12 describing the results in figure.11. >* An additional comment for Figure 9 is required, where apparently the speed-up is starting >to roll over when using more than 8 SMP-nodes, which I think is due to the small work >load per node when using small problem sizes. I am not sure whether my statement is clear or not for this point. I put further commnent in the last part of the 2nd paragraph of p.12. I attached further revised version of the paper to this e-mail. >* In Figure 11 showing the effect of the coefficient matrix storage method and re-ordering, >a decrease in performance is observed using the PDJDS/CM-RCM method when the >number of DOF exceed 106 >. A comment on that should be added. I have already included the comments on the compiler of SR8000 in the 4th paragraph of p.12 describing the results in figure.11. >Referee 2 ************************************************************************************* > >My main concern about this paper is its use of the English language. The paper could be >greatly improved if someone who reads and writes English well could go over this paper and the >abstract with the authors and correct it. There are too many mistakes (with regard to grammar, >spelling, style, etc.) to write them all down in this report. It would take rewriting the >paper. >One reoccurring pattern is the omission of articles, 'a' or 'the', (see also above title of >paper, which I corrected by adding articles). Also, it is important to check the spelling of >the names of the methods, such as Schwarz (not Schwartz), Cuthill-McKee (not Cuthil Mckee) , >etc. I believe I fixed these mistakes. Thank you very much for your comments. Best Regards, --------------------------------------------- Kengo NAKAJIMA, Engineer. nakajima@tokyo.rist.or.jp