chad-fast/src/CHAD, branch multihot-cotangentsAn implementation of CHAD, towards performance
https://git.tomsmeding.com/chad-fast/atom?h=multihot-cotangents2026-02-14T17:03:28ZWIP2026-02-14T17:03:28ZTom Smedingtom@tomsmeding.com2026-02-14T17:03:28Zurn:sha1:4e9957bda4d983d667ad8a86013d766b2e013623WIP2026-02-12T19:44:49ZTom Smedingtom@tomsmeding.com2026-02-12T19:37:46Zurn:sha1:095cc16c62fc6b1f039a10a43bf3bd3a79694f4dWIP (continue in UnMonoid)2026-02-12T19:44:47ZTom Smedingtom@tomsmeding.com2026-02-08T14:43:02Zurn:sha1:62796be35e6e768147aab70ba0beeb94c058c714Multihot cotangents WIP (doesn't work)2026-02-12T19:44:06ZTom Smedingtom@tomsmeding.com2026-01-26T22:37:55Zurn:sha1:c2831ef0f8be71f2a72ee4eee446e2ac473fb638
The idea is sound but for a smaller source language. Notes also in
Obsidian, but the theory so far is that dropping support for nested
arrays makes this possible, although making the result type-safe (i.e.
not have partial functions in a bunch of places) would require making
the lack of nested array support explicit in the embedded type system,
i.e. have Accelerate-like stratification.
The point is that multihots can be added heterogeneously using
plusSparseS but not homogeneously with EPlus or plusSparse, because the
indices might differ between the summands. Thus as long as we never need
to homogeneously sum multihot cotangents, we're golden.
Now the crucial observation is that we only need plus to be homogeneous
on array elements. So if array elements cannot themselves be arrays,
i.e. we drop support for nested arrays, no homogeneous plus of multihot
array cotangents is needed, and we can have static multihots.
UnMonoid: EAccum is an interesting case2026-02-12T19:41:39ZTom Smedingtom@tomsmeding.com2026-02-12T19:41:39Zurn:sha1:f5b1b405fa4ba63bdffe0f2998f655f0b06534bf
Just moved the code, nothing else
Remove spurious work duplication in unMonoid EAccum2026-02-12T19:40:49ZTom Smedingtom@tomsmeding.com2026-02-12T19:40:49Zurn:sha1:5c756bc9eefcd98ed8448f491ddbc3197eaa4127Use helper methods in UnMonoid(plus)2026-02-08T11:28:02ZTom Smedingtom@tomsmeding.com2026-02-08T11:28:02Zurn:sha1:4728c3ac5e577f411b4579ee78648f04f9269a30Remove sparsePlus, it's just EPlus on applySparse2026-02-08T11:27:50ZTom Smedingtom@tomsmeding.com2026-02-08T11:27:50Zurn:sha1:7de36c50001f2be63d8260cd52d0b49872590b9aOptimise eflatten2026-01-26T22:36:56ZTom Smedingtom@tomsmeding.com2026-01-26T22:36:56Zurn:sha1:a9e6c72eff3bee8d45e0d906e8cd027066e04793Import CHAD.Language also qualified as L in Example2025-12-10T09:58:46ZTom Smedingtom@tomsmeding.com2025-12-10T09:58:46Zurn:sha1:06bb9c1a82e6680cffa9f34683c48e0885e495b5