From lydia at cgd.ucar.edu Mon May 23 14:20:22 2005 From: lydia at cgd.ucar.edu (Lydia Shiver) Date: Mon May 23 14:20:23 2005 Subject: [CCSM-SEWG] Information on CCSM SEWG Meeting at Workshop Message-ID: <42923B06.8010908@cgd.ucar.edu> Hello everyone, The SEWG is scheduled to meet at the CCSM3 workshop from 8-9:30 on Thursday, June 23. Our initial proposal is to utilize this time with CSEG, SciDAC, and ESMF project updates. However, we would like to give the community the opportunity to offer suggestions for discussion topics. If you would like to suggest that specific topics be covered as part of the project updates, or would like to propose alternative discussion topics, please let us know by May 27th. Mariana Vertenstein (mvertens@ucar.edu) Cecelia DeLuca (cdeluca@ucar.edu) Patrick Worley (worleyph@ornl.gov) SEWG co-chairs From jacob at mcs.anl.gov Wed May 25 10:09:09 2005 From: jacob at mcs.anl.gov (Robert Jacob) Date: Wed May 25 10:12:50 2005 Subject: [CCSM-SEWG] Comments on ESMF Stage 1 Evaluation Plan. In-Reply-To: <42923B06.8010908@cgd.ucar.edu> References: <42923B06.8010908@cgd.ucar.edu> Message-ID: <6.0.0.22.2.20050525105349.02131100@localhost> Hello SEWG members, As a follow up to our March meeting, comments on the CCSM Stage 1 Evaluation plan for ESMF are below. Overall the evaluation plan is reasonable but there are some items which need clarification. Some suggestions to improve the plan are also provided. We understand that time pressures related to ESMF development milestones meant the stage 1 plan had to be approved quickly. But a possible switch to ESMF is a large scale coding change to CCSM and therefore it is both appropriate and necessary that the SEWG membership participate in the rest of the process: both the review of stage 1 results and planning for stage 2. Regards, Rob Jacob Jay Larson Requests for clarification: 1. In the second paragraph of the Introduction, there is a statement that if ESMF passes stage 1 and is used internally in CAM, it will nonetheless be removed if stage 2 fails. Is this still true? Discussions during the March 2005 SEWG meeting questioned that provision. 2. The 3rd paragraph of the introduction outlines a "pre stage 2" test that ESMF must pass. It says ESMF must be no more than 5% slower than MCT for regridding but the number of fields regridded is not specified. How many fields will be regridded? We recommend at least 10, a typical number for the CCSM3 or CAM/CLM coupler. 3. Also in the 3rd paragraph, ESMF must support concurrent applications where each component runs with a different number of threads. Does this refer to different numbers of OpenMP threads per MPI process? 4. Section 2 states that "CCSM currently uses an early version of ESMF, which we shall refer to as the ESMF prototype, for time management". What does this refer to? The coupler, ocean and sea ice models in CCSM2 and 3 do not make any calls to the ESMF prototype. Only CAM and CLM do. Since the coupler controls overall time integration in CCSM, it is not really accurate to say that *CCSM* uses the ESMF prototype for time management. 5. Section 3.2.1. What does "verified to the same level" mean? 6. Section 4, 2nd paragraph states that the evaluation report will be prepared and presented to the SSC through the CCSM SEWG. What will be done by SEWG membership and what will be done by the SEWG co-chairs? We suggest the SEWG membership should be able to comment on drafts and approve the final version. Suggestions: 1. We suggest that the pre stage-2 test in the Introdcution add a preliminary test of inter-component data transfer between models on two different processor pools. We suggest timing the transfer of at least 10 fields between two components on different sets of processors and that it be no more than 5% slower than MCT's equivalent transfer routines. 2. 4th paragraph of the Introduction, first sentence says. "utilizing the ESMF framework" but this first group of bullets are general reasons for implementing a single executable. 3. Also in the Introduction: we suggest adding a statement indicating that a single executable can be implemented without ESMF. Perhaps change the sentence before the second group of bullets to read: "While it is possible to implement a single-executable CCSM without ESMF, reasons for using ESMF include:" 4. The last bullet in the Introduction would be more accurate if it said: "...can reduce code duplication **between CCSM components** and add..." 5. Section 3.1, 4th bullet: You may want to mention that a call to cpl6 mapping routines could also be placed between CAM and CLM and allow CLM to be run on a different grid. There is no scientific incentive for using ESMF. 6. Section 4, 2nd paragraph states that results will be presented to CSEG and CCSM scientists. Results should also be presented to SEWG members. 7. In order to avoid a possible conflict of interest, SEWG co-chair Cecelia Deluca, who is also the ESMF project manager, should recuse herself from co-chair duties related to the final report but could still participate as a SEWG member