Work Efficient Higher-Order Vectorisation
             ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      Ben Lippmeier,    Manuel Chakravarty,   Gabriele Keller
               Roman Leschinskiy,  Simon Peyton Jones

This is a sequential reference implementation of the the nested array 
representation defined in the paper. A parallel production implementation
is available in the dph-lifted-vseg and dph-prim-par packages of DPH 0.6.1.

Primitive, flat unboxed arrays are provided by the 'vector' package (Data.Vector).
A parallel implementation of our array representation would need a parallel
version of the Data.Vector primitives. All the primitives we use are re-exported
from the Data.Array.Unboxed and Data.Array.Vector modules in this distribution.

Tested with GHC 7.0.4 and 7.4.1

  You need the Leijen pretty printer package installed:
    $ cabal update
    $ cabal install wl-pprint

  Then load the examples:
    $ ghci Examples.hs
   
  Once in GHCi you can display the logical and physical views of an array 
  like so:

    > logical arr5
    ["AB", "CDE", "H"]

    > physical arr5
    PArray 3
     PNested
      VSegd  segmap: [0,1,2]
      SSegd sources: [0,0,1]
             starts: [1,3,4]
       Segd lengths: [2,3,1]
            indices: [0,2,5]
      PChars 0: "XABCDE"
             1: "FGXXHXXX"

    > physical (concat arr5)
      PArray 6
       PChar ("ABCDEH")

  The Examples.hs file contains all the example arrays from the paper.

  The Vectorised/ directory contains derivations of the vectorised
  code for the 'retsum', 'retrieve' and 'furthest' examples.
  You can also run this from the main Examples.hs module.

    > logical $ retsum_v sec5_xss sec5_iss
    [[5,4,5], [20,21], [16]]