it seems unlikely to me that this would cause a degradation in
performance with respect to lists...
that might depend on the number of operations per transposition, i guess.
lists and explicit transpositions make it very obvious what is going on in
terms of
iteration order, so i would be tempted to group operations by whether they need
a plain or transposed structure. arrays hide this kind of thing, so i might pay
the
cost for every transposed operation without the code warning me.
the original poster was doing (transpose . op . transpose) for each out-of-order
op in the list version, so that might be on par with what the array version does, or
not. optimising array computations is a fun business.
but i misread the part i quoted to talk about slowdown on switching between
row-column and column-row, whereas it was generally about slowdown on
switching variable-order operations from lists to arrays. so i assume that posters
here were already aware of the transposition of layout affecting arrays as well.
claus
_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe
