February 2017 - Allocations-admins - mailman3.cels.anl.gov

[LCRC Accounts] Project Allocation Request
by accounts＠lcrc.anl.gov 28 Feb '17

28 Feb '17

Hello, A change in allocation has been requested: Requester: vama (Alvaro VazquezMayagoitia) Project: TDSE-APS Title: Time dependend quantum approaches to support APS X-ray experiments Description: We plan to run ~80 calculations using 1 node each for ~1-10 hrs. We recently added OpenMP to our code and applied vectorization techniques which reduced 10x times the walltime of our calculations. we plan to extend our parallelization with MPI. Most of the calculations will be carried out by two summer interns. Current: undetermined amount Justification: We currently run a binary that computes transitions states between core electrons to continuum states and internal transitions. We recently modified our serial code to use multiple threads. Our code scales to 32 cores per node. We plan to apply MPI parallelization. Requested: 50000 A specific reason has been given: We would like to continue using LCRC resources for our research in x-ray pulses in atoms and molecules. Blues has been extremely useful, and we recently published our results in a peer-reviewed PRA paper: Normal Auger processes with ultrashort x-ray pulses in neon Raymond Sullivan, Junteng Jia, Álvaro Vázquez-Mayagoitia, and Antonio Picón Phys. Rev. A 94, 043421 – Published 26 October 2016 https://journals.aps.org/pra/abstract/10.1103/PhysRevA.94.043421#fulltext Thanks Alvaro Vazquez-Mayagoitia This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System

1 0

Re: [allocations-admins] Allocation Change Request for ApsRenewalLattice
by Bair, Raymond A. 24 Feb '17

24 Feb '17

FYI, The throughput on the Haswell nodes has been so good that Michael does not need the reservation at this time. Ray ----------------------------------------- Ray Bair Argonne National Laboratory and the University of Chicago On 2/23/17, 12:05 PM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of Bair, Raymond A." <allocations-admins-bounces(a)lcrc.anl.gov on behalf of rbair(a)anl.gov> wrote: Thank you for your quick responses and helpful feedback. We have a majority in favor of this plan for ApsRenewalLattice: (1) He will start some runs on the Blues Haswell nodes when they are available, e.g., today. (2) He will have a 20-node reservation on the Sandy Bridge nodes, starting March 1. (3) The weekends have been slow on Blues, so the balance of jobs may be able to be run on weekends. Regards, Ray ----------------------------------------- Ray Bair Argonne National Laboratory On 2/22/17, 7:40 PM, "alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of Merzari, Elia" <alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of emerzari(a)anl.gov> wrote: Given the strategy proposed, I am OK with this. > On Feb 22, 2017, at 7:39 PM, Babnigg, Gyorgy <gbabnigg(a)anl.gov> wrote: > > OK with me as well, > Gyorgy > > -----Original Message----- > From: alloc-com-bounces(a)lists.lcrc.anl.gov [mailto:[email protected]] On Behalf Of Steven Pieper > Sent: Wednesday, February 22, 2017 5:50 PM > To: Wagner, Albert F. <wagner(a)anl.gov> > Cc: 'alloc-com(a)lists.lcrc.anl.gov' <alloc-com(a)lists.lcrc.anl.gov>; LCRC Core <lcrc-core(a)lcrc.anl.gov> > Subject: Re: [LCRC alloc-com] Allocation Change Request for ApsRenewalLattice > > OK with me. > > Steve > > Steven C. Pieper, Senior Physicist Emeritus: spieper(a)anl.gov > Argonne National Laboratory, Physics Division, Bldg. 203, Argonne, IL 60439 > Phone: 630-252-4232 Fax -6008 Secretary, Debra Beres, -4100 > http://www.phy.anl.gov/theory/staff/scp.html > >> On 02/22/2017 05:05 PM, Wagner, Albert F. wrote: >> This project has to get done and the plan below sounds like a >> reasonable solution. >> >> Al >> >> On 2/22/17, 4:33 PM, "alloc-com-bounces(a)lists.lcrc.anl.gov on behalf >> of Bair, Raymond A."<alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of >> rbair(a)anl.gov> wrote: >> >>> We considered some options, looked at Borland’s job sizes, and then I >>> called Michael to talk about the runs needed. Here is a multipronged >>> approach that should meet the APS-U needs at minimal impact to other >>> LCRC >>> users: >>> >>> (1) He will start some runs on the Blues Haswell nodes when they are >>> available, e.g., today. >>> (2) He will have a 20-node reservation on the Sandy Bridge nodes, >>> starting March 1. >>> (3) The weekends have been slow on Blues, so the balance of jobs may >>> be able to be run on weekends. >>> >>> They plan to be done in May, and will back off to a low level then. >>> >>> How does that sound to all? >>> >>> Regards, >>> >>> Ray >>> >>> >>> ----------------------------------------- >>> Ray Bair >>> Argonne National Laboratory >>> >>> >>> On 2/22/17, 2:57 PM, "alloc-com-bounces(a)lists.lcrc.anl.gov on behalf >>> of Steven Pieper"<alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of >>> spieper(a)anl.gov> wrote: >>> >>> How about Rick giving them a special allocation on Theta? >>> >>> Steve >>> >>> Steven C. Pieper, Senior Physicist Emeritus: >>> spieper(a)anl.gov >>> Argonne National Laboratory, Physics Division, Bldg. 203, Argonne, IL >>> 60439 >>> Phone: 630-252-4232 Fax -6008 Secretary, Debra Beres, >>> -4100 >>> http://www.phy.anl.gov/theory/staff/scp.html >>> >>>> On 02/22/2017 02:33 PM, Steven Pieper wrote: >>>> If I have done the arithmetic right, 600K hrs from now to March 31 >>>> is 41 blues nodes continuously. Without fusion and bebop, this is a >>>> substantial impact. But I guess we have no choice. I'd urge him to >>>> delay as much as possible until bebop is available. >>>> >>>> Steve >>>> Steven C. Pieper, Senior Physicist Emeritus: spieper(a)anl.gov Argonne >>>> National Laboratory, Physics Division, Bldg. 203, Argonne, IL >>>> 60439 >>>> Phone: 630-252-4232 Fax -6008 Secretary, Debra Beres, -4100 >>>> http://www.phy.anl.gov/theory/staff/scp.html >>>> >>>>> On 02/22/2017 02:07 PM, Bair, Raymond A. wrote: >>>>> Clarification: He needs 600K between now and March 31, and the >>>>> other 400K in Q3. >>>>> >>>>> >>>>> ----------------------------------------- >>>>> Ray Bair >>>>> Argonne National Laboratory >>>>> >>>>> >>>>> On 2/22/17, 12:08 PM, "Bair, Raymond A."<rbair(a)anl.gov> wrote: >>>>> >>>>> Dear LCRC Allocations Committee, >>>>> >>>>> Michael Borland tells me â€œWe are in the midst of another big >>>>> round of simulations with what is probably the final configuration >>>>> for the upgrade. We need to finish these for the Preliminary Design >>>>> Report, which is a "notable goal" for the lab in FY17.â€ >>>>> >>>>> He is nearly out of time, and requests another 1M core-hours. >>>>> >>>>> Please respond by noon Friday, 2/24, with your approval/disapproval >>>>> and comments. >>>>> >>>>> Thanks. >>>>> >>>>> --Michael >>>>> >>>>> >>>>> ----------------------------------------- >>>>> Ray Bair >>>>> Argonne National Laboratory >>>>> >>>>> >>>>> On 2/16/17, 10:20 AM, "allocations-admins-bounces(a)lcrc.anl.gov on >>>>> behalf of >>>>> accounts(a)lcrc.anl.gov"<allocations-admins-bounces(a)lcrc.anl.gov on >>>>> behalf of accounts(a)lcrc.anl.gov> wrote: >>>>> >>>>> Hello, >>>>> >>>>> A change in allocation has been requested: >>>>> >>>>> Requester: borland (Michael Borland) >>>>> Project: ApsRenewalLattice >>>>> Title: Lattice Development and Beam Dynamics for APS Upgrade >>>>> Description: APS has developed a method of using multi-objective >>>>> genetic algorithms (MOGA) for optimization of accelerator lattices. >>>>> We have also developed a systematic method of evaluating optimized >>>>> lattices for robustness, using many ensembles that represent likely >>>>> outcomes of commissioning. In the coming year, we will refine two >>>>> multi-bend achromat (MBA) lattice designs that promise to push APS >>>>> x-ray brightness to world-leading levels. Significant progress has >>>>> already been made developing a design that is consistent with >>>>> engineering constraints. The design has been presented at external >>>>> reviews related to the on-going upgrade project. >>>>> >>>>> In FY17, there will be an emphasis on faster iteration with >>>>> engineering changes, as well as refinement of physics models. A >>>>> down-selection to a nominal design is expected in early FY17. It is >>>>> likely that two lattices will be carried forward in the physics >>>>> analysis to provide a point of comparison and backup in case issues >>>>> arise with the nominal design (which is expected to be considerably >>>>> more aggressive than the other candidate). >>>>> >>>>> The codes being used for MOGA and ensemble evaluation have all been >>>>> used in past years on fusion. These include Pelegant, a parallel >>>>> accelerator simulation code used primarily for particle tracking >>>>> and acceptance determination, as well as a Tcl script used for >>>>> genetic optimization. The simulations are close to embarrassingly >>>>> parallel and thus scale very well. >>>>> >>>>> In addition to lattice development, we model the effects of the >>>>> higher harmonic cavity (HHC), which plays a vital role in >>>>> increasing the beam lifetime and reducing emittance blow-up. The >>>>> simulations involve tracking beams consisting of 48 or more bunches >>>>> of 100,000 or more particles each, and computing the collective >>>>> electromagnetic interaction of these particles with the vacuum chamber and rf cavities. >>>>> The methods are well advanced and will be employed with new >>>>> lattices as they are generated. Detailed modeling has at present >>>>> been performed mostly for the less aggressive of the two candidate >>>>> lattices, and will need to be extended to the more aggressive >>>>> lattice. Alternatives, such as a three-frequency system and >>>>> low-frequency cavities, will continue to be explored. >>>>> >>>>> Beam loss analysis began will continue in FY17. This involves >>>>> detailed and time-intensive simulation of the injection process and >>>>> the Touschek scattering process. It also involves evaluation of >>>>> beam loss collimation schemes, which are needed in order to protect >>>>> accelerator components and reduce radiation outside the shielding wall. >>>>> >>>>> Current: undetermined amount >>>>> Justification: The primary code being used, ELEGANT, has run >>>>> successfully on fusion, blues, intrepid, mira, cory, franklin, and >>>>> other systems. Most simulations are close to embarrassingly >>>>> parallel, since they are of single-particle dynamics only. >>>>> Simulations with collective effects show greater than 70% >>>>> efficiency for 256 or more cores, depending on problem details. >>>>> >>>>> >>>>> Requested: 1000000 >>>>> >>>>> A specific reason has been given: >>>>> The APS Upgrade project is targeting completion of the preliminary >>>>> design report (PDR) draft by May of this year in time for an >>>>> external review. Completing the PDR is an FY17 lab goal. We need >>>>> additional allocation this quarter to complete required >>>>> calculations for the newly-adopted lattice. >>>>> >>>>> This needs to be approved and the final allocation amount decided upon. >>>>> >>>>> Thank You, >>>>> The LCRC Accounts System >>>>> _______________________________________________ >>>>> allocations-admins mailing list >>>>> allocations-admins(a)lcrc.anl.gov >>>>> https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins >>>>> >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> alloc-com mailing list >>>>> alloc-com(a)lists.lcrc.anl.gov >>>>> https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com >>> >>> >>> _______________________________________________ >>> alloc-com mailing list >>> alloc-com(a)lists.lcrc.anl.gov >>> https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com >> > _______________________________________________ > alloc-com mailing list > alloc-com(a)lists.lcrc.anl.gov > https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com > _______________________________________________ > alloc-com mailing list > alloc-com(a)lists.lcrc.anl.gov > https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins

1 0

Re: [allocations-admins] [LCRC Accounts] Project Allocation Request
by Roberts, John E. 24 Feb '17

24 Feb '17

Time granted. -- John Roberts Argonne National Laboratory CELS Systems jeroberts(a)anl.gov On 2/23/17, 9:51 AM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of Bair, Raymond A." <allocations-admins-bounces(a)lcrc.anl.gov on behalf of rbair(a)anl.gov> wrote: I suggest we approve this request. ----------------------------------------- Ray Bair Argonne National Laboratory On 2/22/17, 6:56 PM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov" <allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov> wrote: Hello, A change in allocation has been requested: Requester: neveu (Nicole Neveu) Project: AWA-beam-dynamics Title: High Accuracy Beam Dynamic Simulations Description: The primary goal of this project is to simulate the two linear electron accelerators at the Argonne Wakefield Accelerator facility. To date, a full 3D models of the beam lines has not been simulated. Using OPAL-T and the parallel capabilities of Blues, we seek better agreement between simulations and experiment by accounting for 3D effects, space charge, coherent synchrotron radiation, and wakefields. Current: undetermined amount Justification: Requested: 90000 A specific reason has been given: The current optimization project is larger than expected. In addition to the need for large random sample runs (to probe/define the parameter space), single optimization runs are taking a day or more to converge. Also, 3D RF field maps were recently incorporated, which adds time to any run, including those mentioned above. This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins

1 0

Re: [allocations-admins] [LCRC Accounts] Project Allocation Request
by Roberts, John E. 24 Feb '17

24 Feb '17

Time granted. -- John Roberts Argonne National Laboratory CELS Systems jeroberts(a)anl.gov On 2/23/17, 9:51 AM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of Bair, Raymond A." <allocations-admins-bounces(a)lcrc.anl.gov on behalf of rbair(a)anl.gov> wrote: I recommend we approve this request. ----------------------------------------- Ray Bair Argonne National Laboratory On 2/22/17, 9:10 PM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov" <allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov> wrote: Hello, A change in allocation has been requested: Requester: jacob (Robert Jacob) Project: ACME Title: Accelerated Climate Modeling for Energy Description: ACME started with a branch of the Community Earth System Model (CESM). CESM is a fully-coupled, global climate model that provides state-of-the-art computer simulations of the Earth's past, present, and future climate states. ACME's atmosphere component is the Community Atmosphere Model. The CAM model has flexible formulations for atmospheric dynamics and has recently transitioned to the spectral finite element method coupled to an extensive suite of sub-grid physical parameterizations in its standard configuration. It runs on unstructured quadrilateral grids. The land model, CLM, contains a suite of column process parameterizations running at each grid point with no communication between grid points. CLM represents surface and subsurface water, energy, carbon, and nitrogen dynamics, with a nested hierarchical sub-grid parameterization that allows glaciers, lakes, urban areas, agricultural fields, managed forest, and natural (unmanaged) vegetation types to share space on each grid-cell. Incident radiation is intercepted in a two-layer canopy, with vegetation, soil, snow aging, and black carbon impacts on albedo. Subsurface processes include vertically resolved biogeochemistry, options for carbon and nutrient cycle parameterization, and recently improved treatment of wetlands and permafrost dynamics. In the fully-coupled configuration, the sea-ice and ocean (MPAS-CICE and MPAS-O) component models run with a nominal grid spacing of 1/10° (approximately 15km at the equator and 5km in polar regions) and, for MPAS-O, 100 levels in the vertical. CAM grids have both uniform and variable resolution configurations. The uniform grid with an average grid point spacing at the equator of 1/4° contains 86,400 elements, each with 30 vertical levels. Doubling the resolution to an average grid spacing of 1/8° requires 345,600 elements. The variable resolution configuration allows a small region of 1/8° grid spacing transitioning to an average global resolution of 1°. The CLM is run on the same grid as CAM. The output of the ACME consists of monthly means of several hundred quantities, plus daily averages of a subset of these quantities and hourly output of some key variabl es. ACME requires Fortran90 and C compilers and the MPI library to run. We will be using time on Blues to perform regular testing of ACME. ACME contains a large set of system tests that perform runs of the full model to test various functions. The runs sizes range from a few nodes to a few dozen nodes. The output sizes are relatively small and we will have modest storage needs. In FY17 we will continue regular testing with the gnu compiler and the large "acme_integration" test suite. In Current: undetermined amount Justification: Requested: 80000 A specific reason has been given: We should only need about 11K hours/month to run our regular testing. But a few ACME folks are accidentally running jobs in the batch instead of ACME queue. I'll remind them. This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins

1 0

Re: [allocations-admins] SHARP-ARC - Project Allocation Request
by Roberts, John E. 23 Feb '17

23 Feb '17

Added 150K to this project plus 500 to get them out of the negative balance. -- John Roberts Argonne National Laboratory CELS Systems jeroberts(a)anl.gov On 2/23/17, 12:01 PM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of Bair, Raymond A." <allocations-admins-bounces(a)lcrc.anl.gov on behalf of rbair(a)anl.gov> wrote: We have converged on adding 150K to SHARP-ARC ----------------------------------------- Ray Bair Argonne National Laboratory On 2/23/17, 11:13 AM, "Yu, Yiqi" <yyu(a)anl.gov> wrote: Dear Ray, I appreciate that very much. I understand that heavy load on blues. I will borrow sometime from other project if we are short. Yiqi -----Original Message----- From: Bair, Raymond A. Sent: Thursday, February 23, 2017 11:11 AM To: Aithal, Shashikant M.; Yu, Yiqi Subject: Re: SHARP-ARC - Project Allocation Request Dear Yiqi, Note that your new second-half allocation will start April 1, providing more time for your project. Could you make do with 150,000 more hours now, to bridge the gap until April 1? Regards, Ray ----------------------------------------- Ray Bair Argonne National Laboratory On 2/23/17, 10:45 AM, "Aithal, Shashikant M." <aithal(a)cels.anl.gov> wrote: Thanks, Yiqi. We will try to make a decision soon. Shashi ________________________________________ From: Yu, Yiqi Sent: Thursday, February 23, 2017 10:42 AM To: Aithal, Shashikant M. Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Sure, As you know 1M elements for nek5000 case is common. So we assume we need to run the case with 1M element. It normal takes 100K core hours to converge. We need to run at least 5 of these case. We still need some time for debugging and for another code. So that's how 1M comes from. Of course, If the resource don't permit, we can try to simplify the case and use URANS instead of LES to save some time. Let me know if you need any other information. -----Original Message----- From: Aithal, Shashikant M. Sent: Thursday, February 23, 2017 10:34 AM To: Yu, Yiqi Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Thanks, Yiqi. It will be helpful if you could let us know the typical time it takes for a run and how many runs you need to finish - say if a run converges in 10 hours on 256 cores, that would be 2560 core-hours. If you have to run 10 such cases, you would need some 25-30K core-hours (assuming some runs take longer or diverge). We (the LCRC-core) can get an idea of how the allotted time is going to be used and approximately how much work can get done. Please provide us with this information (based on the runs you have already done) and we can make a quick decision as to how much time we can grant your project. Thanks, Shashi ________________________________________ From: Yu, Yiqi Sent: Thursday, February 23, 2017 10:21 AM To: Aithal, Shashikant M. Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Shashi, We are going to run some multi-physics simulation(Nek5000 and some neutronic code coupled ) with SAHRP. As you may know, we don't have choice to run these calculation on Mira and Cetus since we are not able to compile the SHARP package on these cluster (very complicated situation). The time for this project is negative now. We need to run some calculations till we can get the next round time till April 1st. Please grant us as much as you can and we will appreciate it very much. Best Yiqi -----Original Message----- From: Aithal, Shashikant M. Sent: Thursday, February 23, 2017 10:14 AM To: Yu, Yiqi Cc: Bair, Raymond A. Subject: SHARP-ARC - Project Allocation Request Yiqi: Can you give us some more information about the jobs you plan to run between now and end of March and how you computed the 1M core-hour request? As you know, our old 2500 core machine Fusion, has been "retired" to make room for the new machine. So there is almost a 30% extra load on Blues. Hence we (LCRC and the allocation committee) are unable to grant large core-hour requests at this time. We might be able to grant you about 80-100K core-hours from now to March end. You might also want to consider running some of the simulations on Mira or Theta since these machines are better suited for jobs requiring large core-hours. Please let me know if you have any other questions or concerns. Thanks, Shashi ________________________________________ From: allocations-admins-bounces(a)lcrc.anl.gov [allocations-admins-bounces(a)lcrc.anl.gov] on behalf of accounts(a)lcrc.anl.gov [accounts(a)lcrc.anl.gov] Sent: Wednesday, February 22, 2017 1:56 PM To: allocations-admins(a)lcrc.anl.gov Subject: [allocations-admins] [LCRC Accounts] Project Allocation Request Hello, A change in allocation has been requested: Requester: yyu (Yiqi Yu) Project: SHARP_ARC Title: Multi-physics Simulation of SFR Core Radial Expansion with SHARP Description: As the SHARP modeling and simulation toolkit is maturing, the focus is on customer-oriented applications and further validation exercises. This allocation will help with those applications, for performing small to moderate scale simulations and for scaling up to ALCF. The SHARP toolkit is regularly compiled on Blues using Intel or GNU compilers. The components of SHARP—the Nek5000 computational fluid dynamics code and the PROTEUS neutron transport code—have already been demonstrated to scale well not only on Blues but on Blue Gene/P. This project will be used by 4-5 analysts in the NE and MCS Divisions. Current: undetermined amount Justification: FY17 efforts for SHARP are highly focused on solving challenge problems for customers and code validation. FY16 efforts were focused on code development, as there was a major release, and so computational requirements were lower than usual. If the project is successful, we anticipate that FY17 and beyond will be highly resource intensive. The components of SHARP—the Nek5000 computational fluid dynamics code and the PROTEUS neutron transport code—have already been demonstrated to scale well not only on Fusion but on Blue Gene/P. Nek5000 has demonstrated > 80% parallel efficiency on BG/P (strong scaling) out to P=131,072 processors, using only 7300 points/processor at the largest scale. PROTEUS has demonstrated greater than 90% strong space-angle scaling at the 100,000 processer level and was nominated for the Gordon Bell Prize for its parallel performance. Requested: 1000000 A specific reason has been given: Used up all the time for now. have urgent job to finish before March This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins

1 0

Re: [allocations-admins] [LCRC alloc-com] Allocation Change Request for ApsRenewalLattice
by Bair, Raymond A. 23 Feb '17

23 Feb '17

Thank you for your quick responses and helpful feedback. We have a majority in favor of this plan for ApsRenewalLattice: (1) He will start some runs on the Blues Haswell nodes when they are available, e.g., today. (2) He will have a 20-node reservation on the Sandy Bridge nodes, starting March 1. (3) The weekends have been slow on Blues, so the balance of jobs may be able to be run on weekends. Regards, Ray ----------------------------------------- Ray Bair Argonne National Laboratory On 2/22/17, 7:40 PM, "alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of Merzari, Elia" <alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of emerzari(a)anl.gov> wrote: Given the strategy proposed, I am OK with this. > On Feb 22, 2017, at 7:39 PM, Babnigg, Gyorgy <gbabnigg(a)anl.gov> wrote: > > OK with me as well, > Gyorgy > > -----Original Message----- > From: alloc-com-bounces(a)lists.lcrc.anl.gov [mailto:[email protected]] On Behalf Of Steven Pieper > Sent: Wednesday, February 22, 2017 5:50 PM > To: Wagner, Albert F. <wagner(a)anl.gov> > Cc: 'alloc-com(a)lists.lcrc.anl.gov' <alloc-com(a)lists.lcrc.anl.gov>; LCRC Core <lcrc-core(a)lcrc.anl.gov> > Subject: Re: [LCRC alloc-com] Allocation Change Request for ApsRenewalLattice > > OK with me. > > Steve > > Steven C. Pieper, Senior Physicist Emeritus: spieper(a)anl.gov > Argonne National Laboratory, Physics Division, Bldg. 203, Argonne, IL 60439 > Phone: 630-252-4232 Fax -6008 Secretary, Debra Beres, -4100 > http://www.phy.anl.gov/theory/staff/scp.html > >> On 02/22/2017 05:05 PM, Wagner, Albert F. wrote: >> This project has to get done and the plan below sounds like a >> reasonable solution. >> >> Al >> >> On 2/22/17, 4:33 PM, "alloc-com-bounces(a)lists.lcrc.anl.gov on behalf >> of Bair, Raymond A."<alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of >> rbair(a)anl.gov> wrote: >> >>> We considered some options, looked at Borland’s job sizes, and then I >>> called Michael to talk about the runs needed. Here is a multipronged >>> approach that should meet the APS-U needs at minimal impact to other >>> LCRC >>> users: >>> >>> (1) He will start some runs on the Blues Haswell nodes when they are >>> available, e.g., today. >>> (2) He will have a 20-node reservation on the Sandy Bridge nodes, >>> starting March 1. >>> (3) The weekends have been slow on Blues, so the balance of jobs may >>> be able to be run on weekends. >>> >>> They plan to be done in May, and will back off to a low level then. >>> >>> How does that sound to all? >>> >>> Regards, >>> >>> Ray >>> >>> >>> ----------------------------------------- >>> Ray Bair >>> Argonne National Laboratory >>> >>> >>> On 2/22/17, 2:57 PM, "alloc-com-bounces(a)lists.lcrc.anl.gov on behalf >>> of Steven Pieper"<alloc-com-bounces(a)lists.lcrc.anl.gov on behalf of >>> spieper(a)anl.gov> wrote: >>> >>> How about Rick giving them a special allocation on Theta? >>> >>> Steve >>> >>> Steven C. Pieper, Senior Physicist Emeritus: >>> spieper(a)anl.gov >>> Argonne National Laboratory, Physics Division, Bldg. 203, Argonne, IL >>> 60439 >>> Phone: 630-252-4232 Fax -6008 Secretary, Debra Beres, >>> -4100 >>> http://www.phy.anl.gov/theory/staff/scp.html >>> >>>> On 02/22/2017 02:33 PM, Steven Pieper wrote: >>>> If I have done the arithmetic right, 600K hrs from now to March 31 >>>> is 41 blues nodes continuously. Without fusion and bebop, this is a >>>> substantial impact. But I guess we have no choice. I'd urge him to >>>> delay as much as possible until bebop is available. >>>> >>>> Steve >>>> Steven C. Pieper, Senior Physicist Emeritus: spieper(a)anl.gov Argonne >>>> National Laboratory, Physics Division, Bldg. 203, Argonne, IL >>>> 60439 >>>> Phone: 630-252-4232 Fax -6008 Secretary, Debra Beres, -4100 >>>> http://www.phy.anl.gov/theory/staff/scp.html >>>> >>>>> On 02/22/2017 02:07 PM, Bair, Raymond A. wrote: >>>>> Clarification: He needs 600K between now and March 31, and the >>>>> other 400K in Q3. >>>>> >>>>> >>>>> ----------------------------------------- >>>>> Ray Bair >>>>> Argonne National Laboratory >>>>> >>>>> >>>>> On 2/22/17, 12:08 PM, "Bair, Raymond A."<rbair(a)anl.gov> wrote: >>>>> >>>>> Dear LCRC Allocations Committee, >>>>> >>>>> Michael Borland tells me â€œWe are in the midst of another big >>>>> round of simulations with what is probably the final configuration >>>>> for the upgrade. We need to finish these for the Preliminary Design >>>>> Report, which is a "notable goal" for the lab in FY17.â€ >>>>> >>>>> He is nearly out of time, and requests another 1M core-hours. >>>>> >>>>> Please respond by noon Friday, 2/24, with your approval/disapproval >>>>> and comments. >>>>> >>>>> Thanks. >>>>> >>>>> --Michael >>>>> >>>>> >>>>> ----------------------------------------- >>>>> Ray Bair >>>>> Argonne National Laboratory >>>>> >>>>> >>>>> On 2/16/17, 10:20 AM, "allocations-admins-bounces(a)lcrc.anl.gov on >>>>> behalf of >>>>> accounts(a)lcrc.anl.gov"<allocations-admins-bounces(a)lcrc.anl.gov on >>>>> behalf of accounts(a)lcrc.anl.gov> wrote: >>>>> >>>>> Hello, >>>>> >>>>> A change in allocation has been requested: >>>>> >>>>> Requester: borland (Michael Borland) >>>>> Project: ApsRenewalLattice >>>>> Title: Lattice Development and Beam Dynamics for APS Upgrade >>>>> Description: APS has developed a method of using multi-objective >>>>> genetic algorithms (MOGA) for optimization of accelerator lattices. >>>>> We have also developed a systematic method of evaluating optimized >>>>> lattices for robustness, using many ensembles that represent likely >>>>> outcomes of commissioning. In the coming year, we will refine two >>>>> multi-bend achromat (MBA) lattice designs that promise to push APS >>>>> x-ray brightness to world-leading levels. Significant progress has >>>>> already been made developing a design that is consistent with >>>>> engineering constraints. The design has been presented at external >>>>> reviews related to the on-going upgrade project. >>>>> >>>>> In FY17, there will be an emphasis on faster iteration with >>>>> engineering changes, as well as refinement of physics models. A >>>>> down-selection to a nominal design is expected in early FY17. It is >>>>> likely that two lattices will be carried forward in the physics >>>>> analysis to provide a point of comparison and backup in case issues >>>>> arise with the nominal design (which is expected to be considerably >>>>> more aggressive than the other candidate). >>>>> >>>>> The codes being used for MOGA and ensemble evaluation have all been >>>>> used in past years on fusion. These include Pelegant, a parallel >>>>> accelerator simulation code used primarily for particle tracking >>>>> and acceptance determination, as well as a Tcl script used for >>>>> genetic optimization. The simulations are close to embarrassingly >>>>> parallel and thus scale very well. >>>>> >>>>> In addition to lattice development, we model the effects of the >>>>> higher harmonic cavity (HHC), which plays a vital role in >>>>> increasing the beam lifetime and reducing emittance blow-up. The >>>>> simulations involve tracking beams consisting of 48 or more bunches >>>>> of 100,000 or more particles each, and computing the collective >>>>> electromagnetic interaction of these particles with the vacuum chamber and rf cavities. >>>>> The methods are well advanced and will be employed with new >>>>> lattices as they are generated. Detailed modeling has at present >>>>> been performed mostly for the less aggressive of the two candidate >>>>> lattices, and will need to be extended to the more aggressive >>>>> lattice. Alternatives, such as a three-frequency system and >>>>> low-frequency cavities, will continue to be explored. >>>>> >>>>> Beam loss analysis began will continue in FY17. This involves >>>>> detailed and time-intensive simulation of the injection process and >>>>> the Touschek scattering process. It also involves evaluation of >>>>> beam loss collimation schemes, which are needed in order to protect >>>>> accelerator components and reduce radiation outside the shielding wall. >>>>> >>>>> Current: undetermined amount >>>>> Justification: The primary code being used, ELEGANT, has run >>>>> successfully on fusion, blues, intrepid, mira, cory, franklin, and >>>>> other systems. Most simulations are close to embarrassingly >>>>> parallel, since they are of single-particle dynamics only. >>>>> Simulations with collective effects show greater than 70% >>>>> efficiency for 256 or more cores, depending on problem details. >>>>> >>>>> >>>>> Requested: 1000000 >>>>> >>>>> A specific reason has been given: >>>>> The APS Upgrade project is targeting completion of the preliminary >>>>> design report (PDR) draft by May of this year in time for an >>>>> external review. Completing the PDR is an FY17 lab goal. We need >>>>> additional allocation this quarter to complete required >>>>> calculations for the newly-adopted lattice. >>>>> >>>>> This needs to be approved and the final allocation amount decided upon. >>>>> >>>>> Thank You, >>>>> The LCRC Accounts System >>>>> _______________________________________________ >>>>> allocations-admins mailing list >>>>> allocations-admins(a)lcrc.anl.gov >>>>> https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins >>>>> >>>>> >>>>> >>>>> >>>>> _______________________________________________ >>>>> alloc-com mailing list >>>>> alloc-com(a)lists.lcrc.anl.gov >>>>> https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com >>> >>> >>> _______________________________________________ >>> alloc-com mailing list >>> alloc-com(a)lists.lcrc.anl.gov >>> https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com >> > _______________________________________________ > alloc-com mailing list > alloc-com(a)lists.lcrc.anl.gov > https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com > _______________________________________________ > alloc-com mailing list > alloc-com(a)lists.lcrc.anl.gov > https://lists.lcrc.anl.gov/mailman/listinfo/alloc-com

1 0

Re: [allocations-admins] SHARP-ARC - Project Allocation Request
by Bair, Raymond A. 23 Feb '17

23 Feb '17

We have converged on adding 150K to SHARP-ARC ----------------------------------------- Ray Bair Argonne National Laboratory On 2/23/17, 11:13 AM, "Yu, Yiqi" <yyu(a)anl.gov> wrote: Dear Ray, I appreciate that very much. I understand that heavy load on blues. I will borrow sometime from other project if we are short. Yiqi -----Original Message----- From: Bair, Raymond A. Sent: Thursday, February 23, 2017 11:11 AM To: Aithal, Shashikant M.; Yu, Yiqi Subject: Re: SHARP-ARC - Project Allocation Request Dear Yiqi, Note that your new second-half allocation will start April 1, providing more time for your project. Could you make do with 150,000 more hours now, to bridge the gap until April 1? Regards, Ray ----------------------------------------- Ray Bair Argonne National Laboratory On 2/23/17, 10:45 AM, "Aithal, Shashikant M." <aithal(a)cels.anl.gov> wrote: Thanks, Yiqi. We will try to make a decision soon. Shashi ________________________________________ From: Yu, Yiqi Sent: Thursday, February 23, 2017 10:42 AM To: Aithal, Shashikant M. Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Sure, As you know 1M elements for nek5000 case is common. So we assume we need to run the case with 1M element. It normal takes 100K core hours to converge. We need to run at least 5 of these case. We still need some time for debugging and for another code. So that's how 1M comes from. Of course, If the resource don't permit, we can try to simplify the case and use URANS instead of LES to save some time. Let me know if you need any other information. -----Original Message----- From: Aithal, Shashikant M. Sent: Thursday, February 23, 2017 10:34 AM To: Yu, Yiqi Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Thanks, Yiqi. It will be helpful if you could let us know the typical time it takes for a run and how many runs you need to finish - say if a run converges in 10 hours on 256 cores, that would be 2560 core-hours. If you have to run 10 such cases, you would need some 25-30K core-hours (assuming some runs take longer or diverge). We (the LCRC-core) can get an idea of how the allotted time is going to be used and approximately how much work can get done. Please provide us with this information (based on the runs you have already done) and we can make a quick decision as to how much time we can grant your project. Thanks, Shashi ________________________________________ From: Yu, Yiqi Sent: Thursday, February 23, 2017 10:21 AM To: Aithal, Shashikant M. Cc: Bair, Raymond A. Subject: RE: SHARP-ARC - Project Allocation Request Shashi, We are going to run some multi-physics simulation(Nek5000 and some neutronic code coupled ) with SAHRP. As you may know, we don't have choice to run these calculation on Mira and Cetus since we are not able to compile the SHARP package on these cluster (very complicated situation). The time for this project is negative now. We need to run some calculations till we can get the next round time till April 1st. Please grant us as much as you can and we will appreciate it very much. Best Yiqi -----Original Message----- From: Aithal, Shashikant M. Sent: Thursday, February 23, 2017 10:14 AM To: Yu, Yiqi Cc: Bair, Raymond A. Subject: SHARP-ARC - Project Allocation Request Yiqi: Can you give us some more information about the jobs you plan to run between now and end of March and how you computed the 1M core-hour request? As you know, our old 2500 core machine Fusion, has been "retired" to make room for the new machine. So there is almost a 30% extra load on Blues. Hence we (LCRC and the allocation committee) are unable to grant large core-hour requests at this time. We might be able to grant you about 80-100K core-hours from now to March end. You might also want to consider running some of the simulations on Mira or Theta since these machines are better suited for jobs requiring large core-hours. Please let me know if you have any other questions or concerns. Thanks, Shashi ________________________________________ From: allocations-admins-bounces(a)lcrc.anl.gov [allocations-admins-bounces(a)lcrc.anl.gov] on behalf of accounts(a)lcrc.anl.gov [accounts(a)lcrc.anl.gov] Sent: Wednesday, February 22, 2017 1:56 PM To: allocations-admins(a)lcrc.anl.gov Subject: [allocations-admins] [LCRC Accounts] Project Allocation Request Hello, A change in allocation has been requested: Requester: yyu (Yiqi Yu) Project: SHARP_ARC Title: Multi-physics Simulation of SFR Core Radial Expansion with SHARP Description: As the SHARP modeling and simulation toolkit is maturing, the focus is on customer-oriented applications and further validation exercises. This allocation will help with those applications, for performing small to moderate scale simulations and for scaling up to ALCF. The SHARP toolkit is regularly compiled on Blues using Intel or GNU compilers. The components of SHARP—the Nek5000 computational fluid dynamics code and the PROTEUS neutron transport code—have already been demonstrated to scale well not only on Blues but on Blue Gene/P. This project will be used by 4-5 analysts in the NE and MCS Divisions. Current: undetermined amount Justification: FY17 efforts for SHARP are highly focused on solving challenge problems for customers and code validation. FY16 efforts were focused on code development, as there was a major release, and so computational requirements were lower than usual. If the project is successful, we anticipate that FY17 and beyond will be highly resource intensive. The components of SHARP—the Nek5000 computational fluid dynamics code and the PROTEUS neutron transport code—have already been demonstrated to scale well not only on Fusion but on Blue Gene/P. Nek5000 has demonstrated > 80% parallel efficiency on BG/P (strong scaling) out to P=131,072 processors, using only 7300 points/processor at the largest scale. PROTEUS has demonstrated greater than 90% strong space-angle scaling at the 100,000 processer level and was nominated for the Gordon Bell Prize for its parallel performance. Requested: 1000000 A specific reason has been given: Used up all the time for now. have urgent job to finish before March This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System

1 0

Re: [allocations-admins] [LCRC Accounts] Project Request: pozdyn
by Roberts, John E. 23 Feb '17

23 Feb '17

Project created and time granted. -- John Roberts Argonne National Laboratory CELS Systems jeroberts(a)anl.gov On 2/21/17, 11:15 AM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov" <allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov> wrote: Hello, A new project on the LCRC cluster has been requested. Please forward the information on to the LCRC Allocation sub-committee. Applicant's name: Mark Pfeifle Applicant's institution: ANL Applicant's division: CSE Project Name: pozdyn Project title: Dissociation Dynamics of Primary Ozonides in Alkene Ozonolysis Associated funding: DAAD (German Academic Exchange Service) Other Systems: Linus (CSE) (no fixed time allocation, depending on current total load) Science: Ozonolysis is a major tropospheric degradation process for alkenes. In the first reaction step, a cyclic primary ozonide (POZ) is formed which is subsequently decomposed into a Criegee intermediate (CI) and a carbonyl compound. The CI population is initially hot due to the exothermic POZ formation. This nascent energy distribution of the CI is essential to adequately model the yields of promptly decomposing and stabilized CI. These quantities in turn are of considerable interest because the thermalized CI acts as an atmospheric oxidant for various species and enhances aerosol formation. It is commonly assumed that the excess energy is distributed statistically among the fragments in the spirit of phase-space theory. However, this assumption is questionable when the exit channel potential energy surface is strongly repulsive, as is the case with POZ decomposition. In this project, direct dynamics calculations for C2H4 + O3 are carried out. Besides con certed POZ cleavage, a step-wise diradical pathway contributes comparably to the reaction flux. Density functional theory based trajectory calculations, initiated at both exit transition states, will serve as a basis to test the statistical partitioning assumption and to provide appropriate input distributions for CI master equation calculations. Project description: The trajectory calculations will be carried out using a self-made driver program by one of our group members (http://tcg.cse.anl.gov/papr/codes/dint.html) It interfaces either the Gaussian 09 or Molpro 2012/2015 program packages to do the actual electronic structure calculations. One typical trajectory run (ca. 1000 time steps using 4 gradient evaluations per iteration) takes about two days on a single processor with density functional theory methods and a moderate-sized one-electron basis set (Gaussian 09). We expect to carry out around 200-300 trajectory runs to achieve statistically significant results. This has to be done for two different transition states that are kinetically relevant. The DFT calculations can be run in parallel in principle. However, according to initial tests, the parallelization efficiency is not optimal for these kind of jobs. Therefore, the quantum-chemical calculations themselves will be run as single-core jobs. Parallelization will be simply achieved by running N separate trajectory jobs on one node (e.g. N=16 for the blues machines). The computational burden within the driver routine is negligible compared to the cost of the electronic structure calculations themselves. In addition to the DFT runs, CASPT2 calculations using Molpro are planned for at least one of the transition states. These are computationally more expensive than DFT calculations, but otherwise the same considerations about parallization apply in this case, too. Presumably, only one member will work on this project. Industry partnership: Project URL: Requested allocation: 54000 Q1: 0 Q2: 18000 Q3: 18000 Q4: 18000 Justification: Storage requirements: The requester has used undetermined amount hours of their initial startup project. In addition to approving an initial amount, please specify a Category and Subcategory for this project. For a list of the current selection of approved categories, please see: https://wiki.lcrc.anl.gov/wiki/Processes/Categories Once the Allocation committee has approved the project, please go to the Project Management page to create it: https://accounts.lcrc.anl.gov/projects.php Thank You, The LCRC Accounts System _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins

1 0

Re: [allocations-admins] [LCRC Accounts] Project Allocation Request
by Roberts, John E. 23 Feb '17

23 Feb '17

Time granted. -- John Roberts Argonne National Laboratory CELS Systems jeroberts(a)anl.gov On 2/21/17, 4:14 PM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov" <allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov> wrote: Hello, A change in allocation has been requested: Requester: qiaobf (Baofu Qiao) Project: DMDOHEMA_1 Title: Metal Ion Transport through Nano-Structured Lipophilic Systems: Atomistic MD Simulations Description: In the last one year, large-scale atomistic molecular dynamics (MD) simulations were performed using the LCRC cluster. In the simulations, we were aiming to couple computer simulations with synchrotron X-ray scattering technique to understand the aggregation behaviors of lanthanide metal-bearing systems. These systems root their significant role in the refinery of lanthanide metals using the solvent extraction technique. We have shown the large-scale MD simulations are able to reproduce the X-ray scattering data (e.g., SAXS). Consequently, the experimental structural feature could be inferred from the simulation structures. Moreover, the energetics which drive the metal transport from aqueous phase to organic phase have been explained. We have published two papers, and submitted one paper. We are currently preparing some manuscripts, which will be submitted soon. Detailed discussion have been provided in the progress report. In the coming facial year, we are aiming to study two projects. In the first project, we are planning to study the interfacial behavior of alcohol at water-organic interface. The role of water-organic interface for the metal separations in the solvent extraction process is still unknown to us. In the present project, we are studying a simple model system of alcohol/water/organic biphasic solution. We are aiming to study the surface activity of alcohol molecules at water-organic interface. Alcohol molecular was sometimes used as co-extractant to enhance the transport kinetics of metal from aqueous phase to organic phase. In our preliminary simulations using the classical pairwise CGenFF force field, the aggregation of alcohol was observed to form a monolayer at the water-dodecane biphasic solution. More interestingly, a second layer of alcohol is also observed in some systems, even though it is relatively very weak. This suggests that the alcohols are preorganized into an asy mmetric bilayer at the water-organic interface, which might explain the elevated extraction kinetics of metal transport across the interface. Such an asymmetric bilayer has never been reported to the best of our knowledge. In the coming facial year, we are aiming to run MD simulations using polarizable force field. These polarizable force field simulations will be used to double check the influences of the simulations force field. In these project, polarizable molecular dynamics (MD) simulations will be performed using the Drude force field. Drude force field has been well developed for simple systems as in the present project. We will run the simulations using the NAMD simulation package. The NAMD package supports Drude force field, and scales well up to thousands of CPUs. High impact publications are expected based on the novel finding, and its significance in both the fundamental research and technical application. In the second project, we are aiming to continue our existing project to study the surface feature of metallate (platinum group metal). In collaboration with the SFG experiments led by Dr. Ahmet Uysal (Assistant Chemist at CSE), we are aiming to study the surface aggregation feature of different metallate (i.e, PtCl62- and PdCl42-), and understand their roles of in metallate separation. We have obtained some results. And Some more simulations are planned to complete the project. Some papers are expected when appropriate. Current: undetermined amount Justification: Note that this account (DMDOHEMA_1) was very recently granted 200,000 CPU hours. Therefore, I am not asking for the allocation for the first two quarters. Requested: 200000 A specific reason has been given: Hi, Could you please provide me more allocation? I am currently in the middle of finishing some simulations to study the metal-surfactant interactions. But the remaining allocation is apparently not enough. Thanks a lot! regards, Baofu This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins

1 0

Re: [allocations-admins] [LCRC Accounts] Project Allocation Request
by Roberts, John E. 23 Feb '17

23 Feb '17

Time granted. -- John Roberts Argonne National Laboratory CELS Systems jeroberts(a)anl.gov On 2/20/17, 3:00 PM, "allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov" <allocations-admins-bounces(a)lcrc.anl.gov on behalf of accounts(a)lcrc.anl.gov> wrote: Hello, A change in allocation has been requested: Requester: pbalapra (Prasanna Balaprakash) Project: perfopt Title: Optimization Algorithms for Empirical Performance Description: There are different strategies for performance model construction. Analytical approaches for performance modeling are cheaper in the sense that they don't suffer from the profiling and model generation cost as in the case of machine learning techniques. But analytical models are often difficult to construct for complex programs and they require detailed understanding of the underlying hardware and implementation of communication functions. Machine learning techniques come into play when it is difficult to construct an analytical model for an application. Models generated using machine learning techniques may not be as highly precise as an analytical model but still they help to understand the performance of an application that in turn help in optimizations such as auto tuning. In this research, we first try to build analytical models for communication of various application kernels. We use architecture-independent parameters as described in well known communication models such as LogGP family to build the analytical models. If the constructed analytical model is accurate enough in predicting the communication time, it saves the huge overhead in gathering training data for the application of machine learning algorithms. We will test the accuracy of the developed analytical models in predicting the communication time of several application kernels from the CORAL benchmarks. If the analytical model is not accurate enough (the error in prediction goes above acceptable range), we switch to gathering profile data and then generate models using machine learning techniques on that data. This hybrid approach gives us a balance between the complexity of building analytical models and the overhead of running expensive machine learning algorithms on huge amount of collected profiling data. Also the machine learning models are limited in the sense that we can't do extrapolation. Fitted analytical models come into play in that case. On Blues, we will conduct an experimental study of communication models. A major benefit Blues brings to this research is the ability to test the impact of MPI implementations and interconnect techniques on Intel-based clusters. Eventually, we hope to deploy our techniques for improving the performance of scientific applications running on Blues. We estimate 100,000 core hours for our experimental study: evaluation of the strengths and limitations of LogGP models = 15,000 core hours; experimental analysis on multiple problem classes on different architectures = 30,000 core hours; empirical analysis for new approaches = 30,000 core hours; tests on parallel performance tuning approaches = 25,000 core hours; We will strongly involve in research dissemination activities such as writing scientific articles and giving seminars at Argonne and in conferences. In the the first and second quarter, our jobs will be 80% parallel. For the third and fourth quarter, we will have 100%. Current: undetermined amount Justification: Requested: 80000 A specific reason has been given: Q2: 15000 Q3: 40000 Q4: 25000 This needs to be approved and the final allocation amount decided upon. Thank You, The LCRC Accounts System _______________________________________________ allocations-admins mailing list allocations-admins(a)lcrc.anl.gov https://lists.lcrc.anl.gov/mailman/listinfo/allocations-admins

1 0