ox-arrays/src/Data/Array/Nested/Internal, branch exper-hs-numopsNested, compositional struct-of-arrays orthotope arrays
https://git.tomsmeding.com/ox-arrays/atom?h=exper-hs-numops2024-05-26T12:59:22ZBenchmark with haskell num ops via mutable vectors2024-05-26T12:59:22ZTom Smedingtom@tomsmeding.com2024-05-26T12:59:06Zurn:sha1:5d3e5c2d001e5724629b7d4e2d88f7bb8ebc6c59
Conclusions:
- orthotope zipWith (+) -> zipWithT: 26.9 ms ± 1.6 ms
- Data.Vector.Storable.zipWith (+): 10.1 ms ± 945 μs
- hsaddDoubleVV (this commit): 1.52 ms ± 49 μs
- cbits/arith.c: 1.29 ms ± 116 μs
- hmatrix: 1.25 ms ± 49 μs
Fast Fractional ops via C code2024-05-26T12:57:34ZTom Smedingtom@tomsmeding.com2024-05-26T12:57:34Zurn:sha1:e80b2593edc3d216905279ebcfa797593a1efbfcRefactor C interface to pass operation as enum2024-05-25T22:11:00ZTom Smedingtom@tomsmeding.com2024-05-25T22:11:00Zurn:sha1:34a9ac8e4497e776c3ca499c41ef749f4edf8383
This is hmatrix style, less proliferation of functions as the number of
ops increases
Bugfix in wrapper for C reduce op2024-05-23T21:09:54ZTom Smedingtom@tomsmeding.com2024-05-23T21:09:54Zurn:sha1:2fad94b1549bcf2ca3ac5a0571e37e3d5b08ca96Fast sum2024-05-23T11:47:18ZTom Smedingtom@tomsmeding.com2024-05-23T11:47:18Zurn:sha1:4c86a3a4231cecc5b7c31491398f43b4ba667eea
Also fast product, but that's currently unused
Fast numeric operations for Num2024-05-22T22:18:17ZTom Smedingtom@tomsmeding.com2024-05-22T22:18:17Zurn:sha1:a0010622885dcb55a916bf3514c0e9040f6871e9