chad-fast/src/Interpreter, branch masterAn implementation of CHAD, towards performance
https://git.tomsmeding.com/chad-fast/atom?h=master2025-11-10T20:50:25ZMove module hierarchy under CHAD.2025-11-10T20:50:25ZTom Smedingtom@tomsmeding.com2025-11-10T20:49:45Zurn:sha1:174af2ba568de66e0d890825b8bda930b8e7bb96Reorder TLEither to after TEither2025-04-29T18:38:05ZTom Smedingtom@tomsmeding.com2025-04-29T18:37:06Zurn:sha1:d0eb9a1edfb4233d557d954f46685f25382234d8Complete monoidal accumulator rewrite2025-04-29T13:54:12ZTom Smedingtom@tomsmeding.com2025-04-29T13:54:12Zurn:sha1:3fd8d35cca2a23c137934a170c67e8ce310edf13WIP interpreter support for new monoidal accumulators2025-04-28T10:05:06ZTom Smedingtom@tomsmeding.com2025-04-28T10:05:06Zurn:sha1:919a36f8eed21501357185a90e2b7a4d9eaf7f08WIP revamp accumulators again: explicit monoid types2025-04-27T21:34:59ZTom Smedingtom@tomsmeding.com2025-04-27T21:34:59Zurn:sha1:b1664532eaebdf0409ab6d93fc0ba2ef8dfbf372
No more D2 in accumulators! Paving the way for configurable sparsity of
products and arrays. The idea is to make separate monoid types for a
"product cotangent" and an "array cotangent" that can be lowered to
either a sparse monoid or a non-sparse monoid. Downsides of this
approach: lots of API duplication.
D2[Array] now has a Maybe instead of zero-size for zero2025-03-26T14:11:48ZTom Smedingtom@tomsmeding.com2025-03-26T14:11:48Zurn:sha1:a00234388d1b4e14481067d030bf90031258b756
Remaining problem: 'add' in Compile doesn't use the D2 stuff
Much process with accumulator revamp2025-03-14T22:29:51ZTom Smedingtom@tomsmeding.com2025-03-14T22:29:51Zurn:sha1:fff6beda3523abce3d27037ea2fb020fce31f502Compile: Emit structs in proper order2025-02-25T22:56:16ZTom Smedingtom@tomsmeding.com2025-02-25T22:56:16Zurn:sha1:7fa10a9a07c7160531baf595d1111277c17a38b2Working argument accum mode (...)2024-11-26T22:05:30ZTom Smedingtom@tomsmeding.com2024-11-26T22:05:30Zurn:sha1:ade38c607a8d0dc8dc1d701084ed88df2fa89df9
The derivative of 'neural' in full accum mode is pretty atrocious now; I
think this is because when you have code like this:
\(a :: Arr 1 R) ->
let b = a
in let c = b
in sum d
then because the argument, as well as both let bindings, bind a value of
array type, each will introduce an accumulator, hence resulting in three
(!) nested `with` clauses that each just contribute their result back to
their parent. This is pointless, and we should fix this.
Test gmm2024-11-10T21:40:54ZTom Smedingtom@tomsmeding.com2024-11-10T21:40:54Zurn:sha1:a46f53695d1dfab8834c7cc52707c0c0bb9b8ba0