v0.0.158

What's Changed

• mlir: Func call reverse diff by @Pangoraw in #2127
• mlir: implement forward mode for func.call by @Pangoraw in #2134
• mlir: cache and reuse reverse funcs by @Pangoraw in #2133
• Fixes warnings and adds missing header guards by @MattBolitho in #2124
• Improve cache index error message by @wsmoses in #2138
• tgamma error improvement by @wsmoses in #2139
• Implement tgamma derivative by @wsmoses in #2140

New Contributors

• @Pangoraw made their first contribution in #2127

Full Changelog: v0.0.157...v0.0.158

v0.0.157

What's Changed

• Adapt to const tblgen by @wsmoses in #2128
• Fix undef value storage by @wsmoses in #2129
• Improve runtime activity err message by @wsmoses in #2132
• Fix combined temp cache for reverse by @wsmoses in #2131

Full Changelog: v0.0.156...v0.0.157

v0.0.156

What's Changed

• Add gcloaded TT by @wsmoses in #2125

Full Changelog: v0.0.155...v0.0.156

v0.0.155

What's Changed

• Add header files to ClangEnzyme target by @jandrej in #2062
• Fix blas decl updater indexing by @wsmoses in #2123

New Contributors

• @jandrej made their first contribution in #2062

Full Changelog: v0.0.154...v0.0.155

v0.0.154

What's Changed

• Main tmp by @wsmoses in #2095
• Adapt to const tblgen main function by @wsmoses in #2099
• Update MustExitScalarEvolution.cpp for upstream LLVM changes by @dklimkin in #2098
• Fix llvm main build by @wsmoses in #2100
• add sinh/cosh/tanh support for llvm19+ by @ZuseZ4 in #2096
• Size returning new handler by @wsmoses in #2101
• MLIR: permit locals in mlir tblgen by @wsmoses in #2102
• MLIR: improved tensor type error messages by @wsmoses in #2103
• MLIR: orig to new map cf by @wsmoses in #2104
• MLIR: fix invert pointer by @wsmoses in #2105
• MLIR: AD Dataflow by @wsmoses in #2106
• MLIR: Recursive readonly by @wsmoses in #2107
• MLIR: Consider readonly in activity analysis by @wsmoses in #2108
• MLIR: improve dataflow interface by @wsmoses in #2110
• MLIR: prioritize dataflow interface by @wsmoses in #2111
• MLIR: fix result positions to shadow by @wsmoses in #2112
• Preserve primal shadow info by @wsmoses in #2113
• [mlir] Fix build for LLVM 20 by @pengmai in #2115
• MLIR: Use terminator location for arg-less reverse function by @wsmoses in #2116
• LLVM integrate for mlir changes by @wsmoses in #2118
• Support more Julia 1.11 functions by @wsmoses in #2121
• Fix handle sync by @wsmoses in #2122
• Improve unknown function error messages by @wsmoses in #2120

New Contributors

• @dklimkin made their first contribution in #2098

Full Changelog: v0.0.153...v0.0.154

v0.0.153

What's Changed

• Disable sparse tests on 18+ by @wsmoses in #2093
• Subnull even with runtime activity by @wsmoses in #2094

Full Changelog: v0.0.152...v0.0.153

v0.0.152

What's Changed

• nofree vprintf by @wsmoses in #2090
• Fix reverse pass ordering by @wsmoses in #2089
• Fix cuda realloc by @wsmoses in #2091
• Fix subnull by @wsmoses in #2092

Full Changelog: v0.0.151...v0.0.152

v0.0.151

What's Changed

• cleanup unused SCEV code by @ZuseZ4 in #2081
• Fix clang compilation (failing since 0.0.150) by @rmoyard in #2082
• BLAS: symv by @wsmoses in #2088

Full Changelog: v0.0.150...v0.0.151

v0.0.150

What's Changed

• Consider true type parsing by @wsmoses in #2079

Full Changelog: v0.0.149...v0.0.150

v0.0.149

What's Changed

• Integrate LLVM at llvm/llvm-project@ede40da1f8c1 by @wsmoses in #2073
• Bump peter-evans/create-pull-request from 6 to 7 by @dependabot in #2072
• Bump DoozyX/clang-format-lint-action from 0.16.2 to 0.18.2 by @dependabot in #2071
• Set minimum LLVM to 15+ by @wsmoses in #2074
• Try to fix memory error by @wsmoses in #2077
• Runtime Activity: move to parameter by @wsmoses in #2076
• Option to prevent caching certain values by @wsmoses in #2078

Full Changelog: v0.0.148...v0.0.149