Go to the Indico Home Page Oldest event Older event Up to category Newer event Newest event Printable version
  • Compact style
  • Indico style
  • Indico style - inline minutes
  • Indico style - numbered
  • Indico style - numbered + minutes
  • Indico Weeks View
    • Switch to the management area of this event
    Login

    MAD-NG Normal Form and Xsuite

    → Europe/Zurich
    MAD-NG Normal Form and Xsuite
    Zoom Meeting ID
    63144281271
    Host
    Rogelio Tomas Garcia
    Useful links
    Join via phone
    Zoom URL
    Hide

    15/09/2021 LNO meeting on MAD-NG Normal Form and Xsuite

    News
    Toilets can now be used by more than one person respecting a minimum distance of 1.5 m.
    Tobias says the expanded drift around the closed orbit has been implemented in MAD-X (exact attribute in Twiss) with results agreeing with PTC. This has solved a previously observed issue with tilted drifts generating a non-zero orbit, as checked by Leon. A similar implementation could only be performed for dipoles but with heavy expressions. For quadrupoles and solenoids integration as exact_drift-kick should be pursued when this is an issue.
    Concerning Ghislain's tapering proposal : finalized with Ghislain, pull request expected in some weeks, for the next MAD-X release.
    Alex Huschauer says there is good progress in setting up crab cavities without beam in the SPS, in preparation of tests with beam planned in ~ 10 days. A very useful engineering / alignment workshop is taking place. He is in contact with Ghislain to integrate FCC-lattices into the acc-models repository.
    Helmut mentions the LBS meeting organized with Roderik next Monday https://indico.cern.ch/event/1073308/ and contacts with the vacuum group to join forces on synchrotron radiation related software motivated by FCC.

    Normal Form implementation in MAD-NG, Laurent Deniau, slides
    ------------------------------------------------------------
    Laurent illustrates the main formulas for Normal Forms and their implementation in MAD-NG. Code is remarkably easy to follow and compact. This new feature is applied, as example, to a FODO lattice made of bends, quadrupoles and sextupoles, all being non-linear thick elements generating a full non-linear one-turn map.
    He uses GTPSA ( Generalised Truncated Power Series Algebra ) to describe and normalise the resulting transport map. The algorithm normalises the map order by order, allowing to further extract key parameters as eigenvalues (tunes) for the linear part and perturbations and resonance driving terms from the non-linear part. He shows the results obtained in the FODO lattice example and says that MAD-NG runs ~ 10 faster than PTC/FPP. The normalizing maps are shown. 

    Action:
    Rogelio and Riccardo are interested in a comparison of actual Resonance Driving Terms and detuning terms with PTC. A simple input/output implementation could follow MADX-PTC approach. A meeting should follow shortly for Laurent to present this, continue with questions to Laurent (as time was limited today) and possibly discuss more advanced of input/output formats.

     

    Xsuite code, Riccardo De Maria + Giovanni Iadarola, slides
    ----------------------------------------------------------
    Riccardo says we currently have at least five CERN developed multiparticle tracking codes used in production. MAD-X, Sixtrack, Sixtracklib, PyHEADTAIL and COMBI are used for studies of CERN synchrotrons. None of them contains all features needed, while standard features are duplicated in different ways. This is hard to maintain and difficult for new users.
    The Xsuite is designed to combine all features without any code duplication. Design goals are simplicity for the users and sustainability with a minimization of lines of code required. The code is provided in the form of a set of Python packages (Xline, Xtrack, Xpart) and based on timing critical code written in C. Vendor independent support of Graphics Processing Units (GPUs) is a necessary requirement. Giovanni says Xsuite already fully integrates COMBI and Sixtracklib, such that maintenance of these as independent packages is no longer needed.
    Documentation is available on https://xsuite.readthedocs.io and codes in the GitHub repository https://github.com/xsuite/


    Discussion : Tobias says he was happy to see that MAD-X lattice descriptions can be imported, however duplicating lattice building should be avoided. Giovanni says that MAD-X input is not sufficient to implement the physics of beam-beam, space-charge, ecloud and SR. Rogelio proposes do discuss this with the goal either to maintain an extended MAD-X format as main input specification or to have a Python-only input to complement MAD-X description.
    Piotr says we should not forget about transfer lines or linear accelerator codes like PLACET. Giovanni replies that while Xsuite started for synchrotrons it is not restricted to rings and that integration of features from PLACET is planned. Laurent highlighted that keeping the deferred expressions feature was a key ingredient to choose LUA versus Python. Giovanni replies that indeed the deferred expression in under investigation in Xsuite and that Xsuite could not be pursed in other language than python as most of the codes used were available in C or python making it very efficient to put together the environment and even discontinue development of codes like Combi as they are fully embeded in Xsuite now. Laurent replies that connecting MAD-NG to Python is a task that was planned for later.

    Action:

    Continue to discuss in the computing working group the key points:  (1) Extended 'lattice' input in MADX Versus Python, pros and cons -> recommendation.  (2) Deferred expressions in Xsuite, technical solution, timeline. (3) MAD-NG connection to Python/Xsuite, technical solution, timeline.  Come back to LNO with outcome of discussions possibly in few months.

     

     

     


     

    There are minutes attached to this event. Show them.