v0.5.1

What's Changed

• feat: split tensors into constant and input dependent by @martinjrobins in #48
• bug: turn off atomic add if not threaded by @martinjrobins in #49
• feat: compiler can be created from raw object file with inhouse jit engine by @martinjrobins in #50
• feat: re-add cranelift and llvm jit and keep inhouse jit by @martinjrobins in #51
• feat: remove write to wasm and no default features by @martinjrobins in #52
• bug: prefix minus by @martinjrobins in #55
• chore: update dependency versions by @martinjrobins in #53
• play with ci by @martinjrobins in #57

Full Changelog: v0.4.0...v0.5.1

v0.4.0

What's Changed

• feat: add gradient and adjoints wrt inputs by @martinjrobins in #45
• bug: remove rhs and lhs from data layout by @martinjrobins in #46
• feat: out tensor is optional and provided externally via args by @martinjrobins in #47

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's Changed

• bug: calculate internal tensors in out by @martinjrobins in #26
• bug: stop calculates time and state dep tensors by @martinjrobins in #27
• bug: fix abs intrinsic function for llvm codegen by @martinjrobins in #30
• feat: test on llvm 18 and windows (just cranelift) by @martinjrobins in #31
• feat: multithreaded execution by @martinjrobins in #34
• doc: fix indices syntax on tensors by @martinjrobins in #35
• feat: add reverse-mode autodiff by @martinjrobins in #38
• version 0.3.0 by @martinjrobins in #39

Full Changelog: v0.2.2...v0.3.0

v0.2.2

What's Changed

• feat: if out not defined set as state by @martinjrobins in #25

Full Changelog: v0.2.1...v0.2.2

v0.2.1

What's Changed

• New-eqn by @martinjrobins in #6
• Allow for any llvm version by @martinjrobins in #7
• Add clippy, doc and fmt to CI by @martinjrobins in #8
• I9 docs by @martinjrobins in #10
• #9 more docs by @martinjrobins in #11
• add optimisations and benchmarking by @martinjrobins in #12
• #13 add heaviside function by @martinjrobins in #14
• allow mass to be truly optional by @martinjrobins in #15
• Enzyme-capi by @martinjrobins in #16
• support llvm 10 - 17 by @martinjrobins in #18
• add llvm to include by @martinjrobins in #19
• bug: fix panic if no f (also other special tensors) by @martinjrobins in #20
• add cranelift codegen backend by @martinjrobins in #21
• bug: fix for case if inputs come in different order by @martinjrobins in #23
• feat: add has_mass to get_dims, add compile to llvm backend by @martinjrobins in #24

Full Changelog: https://github.com/martinjrobins/diffsl/commits/v0.2.1