lhapdf

view TODO @ 1110:2ba52dbf2312

Add release instructions for new sets
author Andy Buckley <andy@insectnation.org>
date Tue, 17 Oct 2017 17:12:04 +0100
parents c740095b9efc
children 5bd93a8d9386
line source
1 /**
2 @page todolist Project to-do list
5 LHAPDF6 TODO list
6 =================
9 VERSION 6.2.2
10 -------------
12 - **CPU-optimize the grid PDF interpolator code**
14 Cache log(x), log(Q) between samplings -> log() still accounts for 15% of
15 CPU. Caching grid index lookup, caching interpolation weights, encorage
16 vectorization.
18 - **New Fortran API completeness**
20 Extrapolation, interpolation, and alpha_s control, and metadata & uncertainty functions in Fortran.
23 - **Add an 'lhapdf show' command**
25 To print pdfsets.index + cat .info for already installed PDFs.
28 - **Consider extrapolated cubic splines at subgrid edges (cf. Valerio Bertone et al)**
32 AS AND WHEN
33 -----------
35 - **Handle zipped data files (AB)**
37 Prefer zipped single member data files rather than virtual filesystem access
38 to the tarball? Can transparently read zipped files with LD_PRELOAD and
39 zlibc: is that enough? Or embed a zip-stream and link against zlib?
42 - **Make it possible to find all metadata keys -- both locally and cascaded (AB)**
45 - **.LHgrid etc. old-name-tolerance control -- TranslateLHA5Names flag?**
48 - **Speed up interpolation with GPUs**
50 Interpolation of PDFs seems like an potential use case for GPUs, since it's
51 normal to query for all partons in the set at once: if we can load the
52 relevant ipol anchors for all flavours onto the GPU then we can maybe get a
53 substantial speedup. OpenMP did not particuarly help, from quick tests.
56 - **PDF flavor aliasing mechanism**
58 e.g. allow anti-flavours to be identical without duplicating their grids in
59 the data files or memory. How could we implement this?
62 - **Allow use of valence/sea etc. decompositions?**
64 GridPDF may be inherited from to allow the returned values to be built from
65 separate interpolations of component PDFs such as interpolated valence, sea,
66 or difference PDFs that are combined to make the physical ones. The PDG ID
67 code range for "generator specific" applications may be used, but we'll need
68 to bear in mind that this will mean that the flavor ID list has different
69 meanings and contents for internal and external purposes: maybe the
70 "internal" PDG ID list needs to become part of the grid data header, or can
71 the metadata be used?
74 - **Using std::function to generically modify the interpolation measures in x, Q (AB)**
77 - **Separate the x and Q2 inter/extrapolation?**
79 Allow mix & match combinations. Would this simplify the code since the
80 1D interpolation methods are very simple and the 2D is built from them?
83 - **Make GridPDFs not read their info or data blocks until an xf value is requested?!**
85 Super-laziness! But is there a real gain other than < 1 sec initialization speed?
87 */