v0.13.2

GPUCompiler v0.13.2

Diff since v0.13.1

Merged pull requests:

• Fixes for late optimization of kernel state arguments (#249) (@maleadt)
• Update manifest (#250) (@github-actions[bot])

v0.13.1

GPUCompiler v0.13.1

Diff since v0.13.0

Merged pull requests:

• Add Enzyme.jl to the reverse testing (#245) (@vchuravy)
• Update manifest (#247) (@github-actions[bot])
• Perform late optimization, after deferred codegen and byval lowering. (#248) (@maleadt)

v0.13.0

GPUCompiler v0.13.0

Diff since v0.12.9

Closed issues:

• julia_method_table function names (#229)
• Compilation of Optimisers.jl pipeline (#238)

Merged pull requests:

• Rewrite byval lowering using clone utils. (#234) (@maleadt)
• NFCs. (#235) (@maleadt)
• Introduce a kernel state argument. (#236) (@maleadt)
• Update manifest (#237) (@github-actions[bot])
• Delete function bodies instead of replacing them. (#239) (@maleadt)
• Remove unneeded test methods. (#240) (@maleadt)
• Update manifest (#241) (@github-actions[bot])
• Merge duplicate functions from deferred codegen. (#242) (@maleadt)
• Update manifest (#243) (@github-actions[bot])
• 1.8 compatibility. (#244) (@maleadt)
• Use at-inline instead of _inline_meta. (#246) (@maleadt)

v0.12.9

GPUCompiler v0.12.9

Diff since v0.12.8

Merged pull requests:

• Actually use our own opt pipeline. (#233) (@maleadt)

v0.12.8

GPUCompiler v0.12.8

Diff since v0.12.7

Merged pull requests:

• Use CFGSimplifyOptions to get closer to Base's optimization pipeline. (#231) (@maleadt)

v0.12.7

GPUCompiler v0.12.7

Diff since v0.12.6

Merged pull requests:

• Update manifest (#225) (@github-actions[bot])
• Don't require DataStructures.jl. (#226) (@maleadt)
• Use byval with type on LLVM 12. (#228) (@maleadt)
• More fixes for the SPIRV compiler (#230) (@maleadt)

v0.12.6

GPUCompiler v0.12.6

Diff since v0.12.5

Merged pull requests:

• Update manifest (#222) (@github-actions[bot])
• Fix Julia IR emission in codegen() (#223) (@wilwxk)
• Emit CUDA version info in the LLVM module. (#224) (@maleadt)

v0.12.5

GPUCompiler v0.12.5

Diff since v0.12.4

Closed issues:

• parameters from ExprTools and LLVM (#217)
• Stack overflow during invalidation (#218)
• GPUCompiler doesn't work on aarch64-apple-darwin (#219)

Merged pull requests:

• Adapt to LLVM changes. (#212) (@maleadt)
• Update manifest (#216) (@github-actions[bot])
• Prevent against recursion during invalidation. (#220) (@maleadt)
• Adapt LazyCodegen tests for Orcv2 (#221) (@vchuravy)

v0.12.4

GPUCompiler v0.12.4

Diff since v0.12.3

Closed issues:

• Both ExprTools and LLVM export "parameters" (#214)

Merged pull requests:

• Fixes for 1.7 (#209) (@maleadt)
• Update manifest (#210) (@github-actions[bot])
• Only import from ExprTools what we need. (#215) (@maleadt)

v0.12.3

GPUCompiler v0.12.3

Diff since v0.12.2

Merged pull requests:

• Update manifest (#205) (@github-actions[bot])
• Hash structures by fields. (#206) (@maleadt)