Bug 2262752

Spec URL: https://fdedden.fedorapeople.org/ghc-bitwise.spec
SRPM URL: https://fdedden.fedorapeople.org/ghc-bitwise-1.0.0.1-1.fc40.src.rpm

Description:
Unboxed multidimensional bit packed Bool arrays with fast aggregate operations
based on lifting Bool operations to bitwise operations.

There are many other bit packed structures out there, but none met all of these
requirements:

(1) unboxed bit packed Bool array,

(2) multi-dimensional indexing,

(3) fast (de)serialization, or interoperable with foreign code,

(4) fast aggregate operations (fold, map, zip).

Quick tour of the bitwise library:

[Data.Bits.Bitwise] Lift boolean operations on 'Bool' to bitwise operations on
'Data.Bits.Bits'.

[Data.Array.BitArray] Immutable bit arrays.

[Data.Array.BitArray.ST] Mutable bit arrays in 'Control.Monad.ST.ST'.

[Data.Array.BitArray.IO] Mutable bit arrays in 'IO'.

[Data.Array.BitArray.ByteString] (De)serialization.

[Codec.Image.PBM] Portable bitmap monochrome 2D image format.

Very rough performance benchmarks:

* immutable random access single bit reads: 'BitArray ix' is about 40% slower
than 'UArray ix Bool',

* 'Control.Monad.ST.ST' mutable random access single bit reads: 'STBitArray s
ix' is about the same as 'STUArray s ix Bool',

* immutable map 'Bool -> Bool': 'BitArray ix' is about 85x faster than 'UArray
ix Bool',

* immutable zipWith 'Bool -> Bool -> Bool': 'BitArray ix' is about 1300x faster
than 'UArray ix Bool'.

Fedora Account System Username: fdedden