[Diff since v0.3.7](v0.3.7...v0.3.8)
**Merged pull requests:**
- Allow Zygote v0.7 (#188) (@mcabbott)
**Closed issues:**
- Use EllipsisNotation ? (#186)

## Tullio v0.3.7
[Diff since v0.3.6](v0.3.6...v0.3.7)
**Merged pull requests:**
- FillArrays v1 (#182) (@chriselrod)
**Closed issues:**
- Please update dep of FillArrays to v1. (#181)

## Tullio v0.3.6
[Diff since v0.3.5](v0.3.5...v0.3.6)
**Merged pull requests:**
- Fix deprecated Vararg syntax (#163) (@jishnub)
- use an extension instead of Requires on 1.9+ (#170) (@KristofferC)
- Updates for CUDA v4, KernelAbstractions v0.9 (#177) (@vpuri3)
- properly load chainrulescore ext (#178) (@vpuri3)
- Dummy PR (#179) (@vpuri3)
**Closed issues:**
- HybridArrays (#155)
- [Question] Is it possible to create a vector of SVectors from a Matrix using Tullio? (#161)
- Use package extensions (#164)
- Upgrade to CUDA.CUDAKernels (#168)
- Bug when using Tullio + LoopVectorization (#169)
- CUDA v4 support (#172)
- Using threads, vs setting threads=false gives different result (#173)
- Issue with vectorized functions on GPU (#174)
- Scalar indexing with CUDA (#176)

## Tullio v0.3.5
[Diff since v0.3.4](v0.3.4...v0.3.5)
**Closed issues:**
- Using Tullio within generated functions (#149)
- poor performance for simple GPU loop (#152)
- Symbolic gradient producing surprising results (#153)
**Merged pull requests:**
- Don't call `parent` before `similar` (#159) (@mcabbott)

## Tullio v0.3.4
[Diff since v0.3.3](v0.3.3...v0.3.4)
**Closed issues:**
- Unnecessary recompilations (#138)
- Scalar operation on GPU on a simple operation (#145)
**Merged pull requests:**
- Skip code coverage on CI (#144) (@mcabbott)
- Allow KernelAbstractions 0.8, etc. (#147) (@mcabbott)

## Tullio v0.3.3
[Diff since v0.3.2](v0.3.2...v0.3.3)
**Closed issues:**
- UndefVarError: CUDADevice not defined (#128)
- Wrong result when LoopVectorization is loaded (#131)

## Tullio v0.3.2
[Diff since v0.3.1](v0.3.1...v0.3.2)
**Merged pull requests:**
- Use LoopVectorization.NativeTypes directly (#123) (@chriselrod)

## Tullio v0.3.1
[Diff since v0.3.0](v0.3.0...v0.3.1)
**Closed issues:**
- scalar indexing when handling CUDAArray inside struct (#96)
- StructArrays timing  (#119)
**Merged pull requests:**
- Allow ChainRulesCore v1 (#121) (@mcabbott)
- Pull field access outside of loops (#122) (@mcabbott)

## Tullio v0.3.0
[Diff since v0.2.14](v0.2.14...v0.3.0)
**Closed issues:**
- VSCode debug with @Einsum stops in macro.jl  (#33)
- besselj0 error (#54)
- TODO: reenable ALL tests on Julia 1.6 and Julia nightly (#73)
- Bugs in range inference with `div` (#76)
- Unable to understand `end` in indexing (#97)
- Convolution example doesn't work with LoopVectorization (#98)
- Support for KernelAbstractions v0.6 / CUDA v3? (#102)
- Weird failure with LoopVectorization (#104)
- Tullio Seems to Cause (Minor) Type Instability (#112)
**Merged pull requests:**
- Debug a CI crash? (#79) (@mcabbott)
- Remove automatic use of TensorOperations (#92) (@mcabbott)
- Improve handling of `end` in indexing (#99) (@mcabbott)
- Test with newer gpu packages, incl. CUDAKernels (#103) (@mcabbott)
- Disallow `div` in indexing (#105) (@mcabbott)
- Scatter isn't always unsafe (#108) (@mcabbott)
- Use ChainRules.jl (#111) (@mcabbott)

## Tullio v0.2.14
[Diff since v0.2.13](v0.2.13...v0.2.14)
**Closed issues:**
-  error CUDADevice not defined (#93)
**Merged pull requests:**
- [patch] `cat_shape` -> `Base.cat_shape` (#100) (@johnnychen94)
- Fix linear indexing of offset vectors (#101) (@mcabbott)