70 likes | 224 Views
Trip to BNL. Goal Installation of newest MAD-X at BNL Setting up space charge for RHIC in frozen mode They do not really want a renormalization of sigma values per turn Adding multipoles Adding “real” BB kicks MPI. 1. I nstallation of newest MAD-X at BNL.
E N D
Trip to BNL • Goal • Installation of newest MAD-X at BNL • Setting up space charge for RHIC in frozen mode • They do not really want a renormalization of sigma values per turn • Adding multipoles • Adding “real” BB kicks • MPI Space Charge Meeting 23 - FS 1
Installation of newest MAD-X at BNL • Setting up of my access to BNL system took 2 days! Despite full advance preparation due to new security measures. • The installation of MAD-X went fine when using IFORT • I have to announce that the present GNU compiler on lxplus “gfortran” is defect! We could identify that the problem is due to optimization so a real problem! Newer versions are again okay. This is a late justification of using commercial compilers like IFORT! • Global SVN access to MAD-X works nicely: • svn co http://svn.cern.ch/guest/madx/trunk/madX mymadX • The newest MAD-X works fine so the effort by Laurent and myself have been successful! Space Charge Meeting 23 - FS 2
Setting-up for RHIC • MAD-X Code fixes and new features: • Needed another flag to suppress the recalculation of the emittances • A misleading naming of variable has been fixed • TRRUN unnecessarily printed tracking data before and after each turn – fixed • Maximum Sliming of Lattice via MAD-X “SEQEDIT” command ➔ next page • MACROS create about 45 SC kicks per Betatron period maybe a bit excessive ➔ further reduce for PS • Various scripts written for data handling Space Charge Meeting 23 - FS 3
Flat and Minimal Lattice ! all set-up ready for clean-up use, period=MACHINE; ! step 1: super flat real sequence even when started with lines ! "SAVE" and "CALL" needed save,sequence=MACHINE,file=blah; call,file=blah; ! Now seqedit seqedit, sequence=MACHINE; ! All classes that can go select, flag=seqedit, class=marker ; select, flag=seqedit, class=monitor ; select, flag=seqedit, class=instrument ; select, flag=seqedit, class=solenoid ; select, flag=seqedit, class=kicker ; select, flag=seqedit, class=multipole,pattern="^eld$" ; remove, element=selected; endedit; ! now use the super clean lattice use, period=MACHINE; oPAC Topical - F. Schmidt
Results • Multipoles inclusion trivial • Extra BB kicks not really foreseen in the SC MACROS but operational • Tuneshift from SC and BB as theoretically and experimentally expected • The sussix code successfully installed including data scripts • Detuning with amplitude successfully determined for RHIC • FMA analysis started • 50’000 turns per day for 1’000 particles (sufficient for frozen space charge!) • Speed-up via MPI ➔ next slide • As a consequence Valery and myself will modify the MACROs Space Charge Meeting 23 - FS 5
MPI • They have very competent staff to handle MPI! • Timing analyse of code functioning and an easy gain of a factor of 4-5 is possible. • In the RHIC case where a renormalization of the sigma values is not needed this should be about a factor of 10. • Several MPI strategies have been discussed • It was agreed to modify the code as little as possible but organize it by running a separate set of data on each node and bringing the data together for the emittance renormalization. • The issue of lost particles and IO have been addressed • A handy script will be provided for easy usage • ➔ Basically they will do the job for us!!! Space Charge Meeting 23 - FS 6
Round Table • The crash issue must be resolved asap! This will need till after the summer • Nevertheless tracking can proceed since data seem alright till the possible crash • Please start using the MPICH2 version which needs a small set-up effort all explained in recent Email • Do we need more resources in terms of the 48 core machines? • MAD-X with frozen space charge will be fully applied to the PS in view of the experiments of June 2012 ➔ Short demonstration any time! • The nonlinear coupled resonances will be studied together with Giuliano ➔ trip next 2 weeks • The benchmark effort and the mitigation effort in full swing ➔ some update during Fermilab visit • PTC-ORBIT Manual framework has been set-up ➔ start filling! Space Charge Meeting 23 - FS 7