Skip to content

v1.10.0

Latest

Choose a tag to compare

@github-actions github-actions released this 03 Jul 10:39
b8d46ce

Metal v1.10.0

Diff since v1.9.3

Closed issues:

  • Switch to XCode's Metal compiler? (#7)
  • Support for exceptions (#69)
  • Kernel hang related to noreturn function attributes (#113)
  • MPS - Support for Convolutional Neural Network kernels (#210)
  • FFT support (#270)
  • Metal.jl produces incorrect (incomplete) results with DiffEqGPU on Julia v1.10 (#275)
  • Add Support for BFloat16 (#298)
  • Support for Paravirtualized Graphics for Github Actions CI (#309)
  • Metal 3.1 and 3.2 (#373)
  • Add a metal_version function (#393)
  • Implement MPSMatrixSolve (#402)
  • Display summary for Metal.@profile (#431)
  • Simple throwing kernel hangs (#433)
  • Base._cpow requires Metal.jl-specific implementation to avoid use of Float64 (#441)
  • multiplication fails with discontinuous view (#460)
  • Improve memory pressure detection (#524)
  • Support task switches from command buffer callbacks (#532)
  • Add CI for earliest supported macOS version (#540)
  • Dynamic dispatch when defining @device_override Base.max(x::Int64, y::Int64) (#547)
  • += yields incorrect results sometimes (#549)
  • Allow kwargs for @profile (#562)
  • Writing tuple to array doesn't work with bounds check (#589)
  • [Meta] OS support policy (#596)
  • Segfault in MPS Copy test (#600)
  • implementation of ldiv? (#653)
  • 1.13 support (#664)
  • Turn on MTL_DEBUG_LAYER_WARNING_MODE for Api Validation tests (#683)
  • Support ComplexF32 division (#736)
  • Occasional hang/segfault in mtl/metal tests (#742)
  • Replace device overrides with an intrinsic raising pass (#746)
  • ComplexF32 broadcasted division fails (#763)
  • Bug during copyto! (#766)
  • Back-end crash due to large aggregate loads (#792)
  • randn!(::MtlArray{Float32}) still produces NaNs on Metal.jl 1.9.3 (Apple M4 Pro) (#795)
  • Julia 1.13 regression. KernelAbstractions printing fails to compile (#799)
  • code_llvm is confusing (#800)
  • [Julia 1.13] BFloat16 compilation failures (#817)
  • Tensor matmul errors on Julia 1.10 & 1.11 (#825)
  • Error output errors on Julia 1.14 (#826)
  • Native gemm algorithm selection (#829)
  • Tensor gemm shader validation failures (#834)
  • Fix flashattention example (#838)
  • Unexpected recompilation when REPL is loaded (#840)