[dependabot]

Updates the requirements on and KernelAbstractions to permit the latest version.
Updates KernelAbstractions to 0.10.0

Release notes

Sourced from KernelAbstractions's releases.

v0.10.0

KernelAbstractions v0.10.0

Diff since v0.9.45

Breaking changes

See the 0.10 section of the changelog: JuliaGPU/KernelAbstractions.jl#0 In short: the CPU backend is now an OpenCL backend running on PoCL, CPU(; static=true), isgpu, @ka_code_llvm and @ka_code_typed were removed, GPU is deprecated, @private arrays are a PrivateArray, @Const arguments are constified at launch, and backends implement KernelInterface 0.4.

Merged pull requests:

• Add @groupreduce and @groupscan (#830) (@​vchuravy)
• KernelInterface: sub-group shuffles, votes, reductions and scans (#831) (@​vchuravy)
• Complete the 0.10 release notes and update the NUMA example (#842) (@​vchuravy)
• POCL: recommend work-group sizes with PoCL's own heuristic (#861) (@​vchuravy)
• Compute indices in Int on PoCL, and mark the index arithmetic nuw/nsw (#879) (@​vchuravy)
• Finalize 0.10: release notes, precompile warnings, compat and CI (#885) (@​maleadt)

Closed issues:

• CPU backend: the bounds mask makes kernel bodies up to 2.3x slower, even when the workgroups tile the ndrange (#845)
• CPU backend: the default workgroup size is the device maximum (4096), so launches of up to 4096 work-items run on one thread (#859)
• CPU backend: work-item loops with fewer than 16 iterations are not vectorized (runtime SCEV checks + tiny trip count) (#868)

Commits

• 61bd51c Finalize 0.10: release notes, precompile warnings, compat and CI (#885)
• daccb07 Add @groupreduce and @groupscan (#830)
• d6392b6 KernelInterface: sub-group shuffles, votes, reductions and scans (#831)
• 2c67ccb Bump version from 0.10.0-dev to 0.10.0
• a0a2e68 Complete the 0.10 release notes and update the NUMA example (#842)
• 4794aae Compute indices in Int on PoCL, and mark the index arithmetic nuw/nsw (#879)
• ad1a8b9 POCL: recommend work-group sizes with PoCL's own heuristic (#861)
• bc4014d DGSEM benchmarks: a local-memory layout per backend (#875)
• 457a1ed DGSEM benchmarks: local memory with the variable index last (#871)
• a587ec6 DGSEM benchmarks: double-buffered slab kernel (#870)
• Additional commits viewable in compare view

Updates KernelAbstractions to 0.10.0

Release notes

Sourced from KernelAbstractions's releases.

v0.10.0

KernelAbstractions v0.10.0

Diff since v0.9.45

Breaking changes

See the 0.10 section of the changelog: JuliaGPU/KernelAbstractions.jl#0 In short: the CPU backend is now an OpenCL backend running on PoCL, CPU(; static=true), isgpu, @ka_code_llvm and @ka_code_typed were removed, GPU is deprecated, @private arrays are a PrivateArray, @Const arguments are constified at launch, and backends implement KernelInterface 0.4.

Merged pull requests:

• Add @groupreduce and @groupscan (#830) (@​vchuravy)
• KernelInterface: sub-group shuffles, votes, reductions and scans (#831) (@​vchuravy)
• Complete the 0.10 release notes and update the NUMA example (#842) (@​vchuravy)
• POCL: recommend work-group sizes with PoCL's own heuristic (#861) (@​vchuravy)
• Compute indices in Int on PoCL, and mark the index arithmetic nuw/nsw (#879) (@​vchuravy)
• Finalize 0.10: release notes, precompile warnings, compat and CI (#885) (@​maleadt)

Closed issues:

• CPU backend: the bounds mask makes kernel bodies up to 2.3x slower, even when the workgroups tile the ndrange (#845)
• CPU backend: the default workgroup size is the device maximum (4096), so launches of up to 4096 work-items run on one thread (#859)
• CPU backend: work-item loops with fewer than 16 iterations are not vectorized (runtime SCEV checks + tiny trip count) (#868)

Commits

• 61bd51c Finalize 0.10: release notes, precompile warnings, compat and CI (#885)
• daccb07 Add @groupreduce and @groupscan (#830)
• d6392b6 KernelInterface: sub-group shuffles, votes, reductions and scans (#831)
• 2c67ccb Bump version from 0.10.0-dev to 0.10.0
• a0a2e68 Complete the 0.10 release notes and update the NUMA example (#842)
• 4794aae Compute indices in Int on PoCL, and mark the index arithmetic nuw/nsw (#879)
• ad1a8b9 POCL: recommend work-group sizes with PoCL's own heuristic (#861)
• bc4014d DGSEM benchmarks: a local-memory layout per backend (#875)
• 457a1ed DGSEM benchmarks: local memory with the variable index last (#871)
• a587ec6 DGSEM benchmarks: double-buffered slab kernel (#870)
• Additional commits viewable in compare view

[dependabot]

Looks like these dependencies are updatable in another way, so this is no longer needed.