v2.0.2

CUDA v2.0.2

Diff since v2.0.1

Closed issues:

• cu() behavior for complex floating point numbers (#91)
• Error when following example on using multiple GPUs on multiple processes (#468)
• MacOS without nvidia GPU is trying to download CUDA111 on julia nightly (#469)
• Drop BinaryProvider? (#474)
• Latest version of master doesn't work on Windows (#477)
• sum(CUDA.rand(3,3)) broken (#480)
• copyto!() between cpu and gpu with subarrays (#491)

Merged pull requests:

• Adapt to GPUCompiler changes. (#458) (@maleadt)
• Fix initialization of global state (#471) (@maleadt)
• Remove 'view' implementation. (#472) (@maleadt)
• Workaround new artifact"" eagerness that prevents loading on unsupported platforms (#473) (@ianshmean)
• Remove BinaryProvider dep. (#475) (@maleadt)
• typo: libcuda.dll -> libcuda.so on Linux (#476) (@Alexander-Barth)
• NFC array simplifications. (#481) (@maleadt)
• Update manifest (#485) (@github-actions[bot])
• Convert AbstractArray{ComplexF64} to CuArray{ComplexF32} by default (#489) (@pabloferz)

v2.0.1

CUDA v2.0.1

Diff since v2.0.0

Closed issues:

• Can't update (#462)

Merged pull requests:

• Remove duplicate comment (#464) (@blegat)
• Add functionality to precompile the runtime library. (#465) (@maleadt)
• Update manifest (#470) (@github-actions[bot])

v2.0.0

CUDA v2.0.0

Diff since v1.3.3

Closed issues:

• Test failure during threading tests (#15)
• Bad allocations in memory pool after device_reset! (#16)
• CuArrays can lose Blas on reshaped views (#78)
• allowscalar performance (#87)
• Indexing with a CuArrays causes a 'scalar indexing disallowed' error from checkbounds (#90)
• 5-arg mul! for CUSPARSE (#98)
• copyto!(Device, Host) uses scalar iteration in case of type mismatch (#105)
• Array primitives broken for CUSPARSE arrays (#113)
• SplittingPool: CPU allocations (#117)
• error while concatenating to an empty CuArray (#139)
• Showing sparse arrays goes wrong (#146)
• Improve test coverage (#147)
• CuArrays allocates a lot of memory on the default GPU (#153)
• [Feature Request] Indexing CuArray with CuArray (#155)
• Reshaping CuArray throws error during backpropagation (#162)
• Match syntax and APIs against Julia 1.0 standard libraries (#163)
• CURAND_STATUS_PREEXISTING_FAILURE when setting seed multiple times. (#212)
• RFC: converts SparseMatrixCSC to CuSparseMatrixCSR via cu by default (#216)
• Add a CuSparseMatrixCOO type (#220)
• Test runner stumbles over path separators (#236)
• Error: Invalid bitcode signature when loading CUDA.jl after precompilation (#293)
• Atomic operations only work on global memory (#311)
• Performance: cudnn algorithm selection (#318)
• CUSPARSE is broken in CUDA.jl 1.2 (#322)
• Device-side broadcast regression on 1.5 (#350)
• API for fast math-like mode (#354)
• CUDA 11.0 Update 1: cublasSetWorkspace (#365)
• Can't precompile CUDA.jl on Kubuntu 20.04 (#396)
• CuPtr should be Ptr in cudnnGetDropoutDescriptor (#397)
• CUDA throws OOM error when initializing API on multiple devices (#398)
• Cannot launch kernel with > 5 args using Dynamic Parallelism (#401)
• Reverse performance regression (#410)
• Tag for LLVM 3? (#412)
• CUDA not working (#415)
• StatsBase.transform fails on CuArray (#426)
• Further unification of CUBLAS.axpy! and LinearAlgebra.BLAS.axpy! (#432)
• size(range), length(range) and range[end] fail inside CUDA kernels (#434)
• InitError: Cannot use memory pool 'binned' when CUDA.jl was precompiled for memory pool 'split'. (#446)
• Missing dispatch for matrix multiplication with views? (#448)
• New version not available yet? (#452)
• using CUDA or CUArray, output: UndefVarError: AddrSpacePtr not defined (#457)
• Unable to upgrade to the latest version (#459)

Merged pull requests:

• Performance improvements by calling cuDNN API (#321) (@gartangh)
• Use ccall wrapper for correct pointer type conversions (#392) (@maleadt)
• Simplify Statistics.var and fix dims=tuple. (#393) (@maleadt)
• Adapt to GPUArrays test change. (#394) (@maleadt)
• Default to per-thread stream semantics (#395) (@maleadt)
• Add a missing context argument for stateless codegen. (#399) (@maleadt)
• Keep track of package latency timings. (#400) (@maleadt)
• Update manifest (#402) (@github-actions[bot])
• Latency improvements (#403) (@maleadt)
• Fix bounds checking with GPU views. (#404) (@maleadt)
• Force specialization for dynamic_cudacall to support more arguments. (#407) (@maleadt)
• Fix some wrong pointer types in the CUDNN headers. (#408) (@maleadt)
• Refactor CUSPARSE (#409) (@maleadt)
• Fix typo (#411) (@yixingfu)
• Update manifest (#413) (@github-actions[bot])
• Simplify library wrappers by introducing a CUDA Ref (#414) (@maleadt)
• Simplify and update wrappers (#416) (@maleadt)
• GEMM improvements (#417) (@maleadt)
• CompatHelper: add new compat entry for "BFloat16s" at version "0.1" (#418) (@github-actions[bot])
• add CuSparseMatrixCOO (#421) (@marius311)
• Update manifest (#423) (@github-actions[bot])
• Global math mode for easy use of lower-precision functionality (#424) (@maleadt)
• Improve init error message (#425) (@maleadt)
• CUBLAS: wrap rot! to implement rotate! and reflect! (#427) (@maleadt)
• CUFFT-related optimizations (#428) (@maleadt)
• Fix reverse/view regression (#429) (@maleadt)
• Update packages (#433) (@maleadt)
• Introduce StridedCuArray (#435) (@maleadt)
• Retry curandGenerateSeeds when OOM. (#436) (@maleadt)
• Introduce DenseCuArray union (#437) (@maleadt)
• Array simplifications (#438) (@maleadt)
• Fix and test reverse on wrapped array. (#439) (@maleadt)
• Fixes after recent array wrapper changes (#441) (@maleadt)
• Adapt to GPUArrays changes. (#442) (@maleadt)
• Provide CUBLAS with a pool-backed workspace. (#443) (@maleadt)
• Fix finalization of copied arrays. (#444) (@maleadt)
• Support for/Add CUDA 11.1 (#445) (@maleadt)
• Update manifest (#449) (@github-actions[bot])
• Allow use of strided vectors with mul! (gemv! and gemm!) (#450) (@maleadt)
• Have convert call CuSparseArray's constructors. (#451) (@maleadt)

v1.3.3

CUDA v1.3.3

Diff since v1.3.2

Closed issues:

• Type changing Array conversions give error when allowscalar(false) (#344)
• getindex(::CuArray, ::Adjoint, ::Colon) fails (#345)
• View with array indices causes memory copy before broadcast (#384)
• Regression with Julia 1.5 (#390)

Merged pull requests:

• Replace DevicePtr with Core.LLVMPtr. (#199) (@maleadt)
• Make sure view indices reside on the GPU too. (#388) (@maleadt)
• CompatHelper: Update DataStructures to v0.18 (#389) (@ChrisRackauckas)

v1.3.2

CUDA v1.3.2

Diff since v1.3.1

Closed issues:

• LLVM WMMA errors (#380)

Merged pull requests:

• Fix handling of tests to skip. (#386) (@maleadt)
• Update manifest (#387) (@github-actions[bot])

v1.3.1

CUDA v1.3.1

Diff since v1.3.0

Closed issues:

• Element-wise conversion fails (#378)
• atomic_min fails for Int32 in global CuDeviceArrays (#379)
• Segmentation fault from @cuprint on char (#381)
• error in versioninfo(), name not defined (#385)

Merged pull requests:

• Fix docs (#330) (@maleadt)
• Wrap cusparseSpMV (#351) (@marius311)
• specify Cchar rather than char in the doc for @cuprint (#382) (@MasonProtter)
• Adapt to LLVM.jl changes for stateless codegen. (#383) (@maleadt)

v1.3.0

CUDA v1.3.0

Diff since v1.2.1

Closed issues:

• Trouble with the @. macro (#346)
• NVMLError: Not Supported (code 3) (#348)
• Nvidia Xavier devices: exception thrown during kernel execution on device Xavier (#349)
• Could not load CUTENSOR artifact dll on Windows 10 (#355)
• CuTextureArray for 3D array (#357)
• Bug in julia 1.5.0 I have CUDA 11.0 installed in Ubuntu 18.04 (#360)
• Callback-based logging (#366)
• Artifact download timeout (#369)
• sum! accumulates when called multiple times (#370)
• nvprof does not detect kernel launches (#371)
• KernelError: passing and using non-bitstype argument (#372)
• CUDA.jl fails to find libcudadevrt.a due on a cluster install with multi-arch target (#376)

Merged pull requests:

• Make the memory allocator context-aware (#253) (@maleadt)
• Update manifest (#347) (@github-actions[bot])
• Guard against unsupported NVML usage in the test runner. (#352) (@maleadt)
• Bump CUDNN to v8.0.2 (#353) (@maleadt)
• Rework thread state management (#356) (@maleadt)
• Update manifest (#358) (@github-actions[bot])
• Memory allocator simplifications (#361) (@maleadt)
• Deduplicate code from memory pools (#362) (@maleadt)
• Fix show of ArrayBuffer. (#363) (@maleadt)
• Clean-up the Buffer interface. (#364) (@maleadt)
• Use callback APIs to get library debug logs. (#367) (@maleadt)
• Allow selecting the memcheck tool. (#368) (@maleadt)
• Update GPUArrays. (#373) (@maleadt)
• Update to CUDA 11.0 update 1 (#374) (@maleadt)
• Number and iterate devices in versioninfo() following CUDA. (#375) (@maleadt)
• Reinstate support for Julia 1.3 (#377) (@maleadt)

v1.2.1

CUDA v1.2.1

Diff since v1.2.0

Closed issues:

• CuArrays.zeros(T, 0) fails (#81)
• CUDAnative.cos calls the base cos function in nested broadcast (#102)
• CuSparseMatrixHYB * CuMatrix = nothing (#256)
• Strange reordering of struct fields with dynamic parallelism (#263)
• Performance: bias add (#298)
• CUDA 11 libraries incorrectly looked up in artifact (#300)
• CUTENSOR for windows (#301)
• Performance: sum (#302)
• Performance: getindex(a, i::Array{Int}) (#303)
• Display for CuArray within Tuples does not respect :limit=>true (#305)
• Performance: elementwise operations (#307)
• Performance: perceptron (#312)
• windows install error: isfile(__libcupti[]) (#324)
• std with dims is not type stable (#336)

Merged pull requests:

• Re-enable threading tests. (#25) (@maleadt)
• Reorganize and simplify some includes (#296) (@maleadt)
• Only run benchmarks on the master branch. (#297) (@maleadt)
• Optimizations for broadcast (#299) (@maleadt)
• Update manifest (#304) (@github-actions[bot])
• Test runner improvements for multigpu mode (#309) (@maleadt)
• Artifact improvements for CUDA 11 on Windows (#310) (@maleadt)
• Optimize element-wise operations (#313) (@maleadt)
• Check if reported GPU memory use is available. (#314) (@maleadt)
• Update artifacts: include cusolverMg, and use Yggdrasil binaries. (#315) (@maleadt)
• Specialization fixes for mapreducedim. (#316) (@maleadt)
• Fix invalid conversion of pointer to signed integer. (#317) (@maleadt)
• Work around (presumed) Windows driver bug in exception test. (#319) (@maleadt)
• Update manifest (#323) (@github-actions[bot])
• Bump CUDNN and CUTENSOR (#325) (@maleadt)
• Simplify NVML discovery. (#326) (@maleadt)
• Separate CURAND wrappers from Random impl. (#327) (@maleadt)
• Simplify discovering binaries by using Sys.which. (#328) (@maleadt)
• Add wrapper for NVML utilization rates. (#329) (@maleadt)
• Attach CUSPARSE docstrings to bare methods, not empty functions. (#331) (@maleadt)
• Eagerly reduce the amount of worker threads. (#332) (@maleadt)
• Bump dependencies. (#333) (@maleadt)
• Clean-up library wrappers [NFC] (#334) (@maleadt)
• Fix CUDNN v8 discovery and loading on Windows (#335) (@maleadt)
• Fix type stability of Statistics.var with dims. (#337) (@maleadt)
• Fix parameter alignment for dynamic parallelism. (#338) (@maleadt)
• Micro-optimize Base.fill. (#339) (@maleadt)

v1.2.0

CUDA v1.2.0

Diff since v1.1.0

Closed issues:

• Segmentation fault when creating CuArray of CuArray (#133)
• CUDNN tests fail with CUDNN 6.0.20 (#134)
• CURAND fail to initialize, code 203 (#255)
• Deprecation warnings (#277)
• Can we pleeeeeeeease make cu(x) eltype preserving? (#278)
• On the use of @sync during benchmarking in the documentation (#279)
• Example in Multiple GPUs doc fails (#282)
• LLVM error: Cannot cast between two non-generic address spaces (#286)

Merged pull requests:

• Host-side CUTENSOR (#243) (@kshyatt)
• Add and document a non-blocking version of at-sync. (#280) (@maleadt)
• Use a custom adaptor for cu so that adapt(CuArray) preserves element types. (#281) (@maleadt)
• Check and warn for library versions. (#284) (@maleadt)
• Add note about nvml dll missing (#288) (@kshyatt)
• Update your PR to have tests pass (#289) (@kshyatt)
• Update manifest (#290) (@github-actions[bot])
• Support CUDA 11 (#291) (@maleadt)
• do not open the file twice when reading the libdevice bitcode (#294) (@jakebolewski)

v1.1.0

CUDA v1.1.0

Diff since v1.0.2

Closed issues:

• Fix NSight detection (#29)
• versioninfo() (#34)
• throw_... messages: invalid call to jl_alloc_string (#54)
• INTERNAL_ERROR during CUDNN handle creation (#183)
• Improve benchmarking suite (#222)
• How to load CUDA.jl conditional on the computer having a CUDA-compatible GPU? (#237)
• CUSOLVER.heevd! returning Float and not Complex (#238)
• Broadcasting fails with Float64 -> Int conversion (#240)
• Running ] test CUDA with OhMyREPL in startup.jl causes some tests to fail (#246)
• ERROR: Your LLVM does not support the NVPTX back-end. in local project environment (#249)
• CUDAnative: UndefVarError: AddrSpacePtr not defined on julia master (#250)
• Error while freeing CUDA.CuPtr (#254)
• Non-artifact initialization of CUDA.jl using CUDA 11 fails on Windows (#262)
• Library handle creation close to OOM fails with ERROR_NOT_INITIALIZED (#264)
• has(::TargetIterator, name::String) deprecation warning (#271)

Merged pull requests:

• Add texture support from CuTextures.jl (#209) (@maleadt)
• Memory pinning with interval trees (#233) (@maleadt)
• Better nsys detection. (#234) (@maleadt)
• CompatHelper: add new compat entry for "IntervalTrees" at version "1.0" (#235) (@github-actions[bot])
• Update manifest (#239) (@github-actions[bot])
• Replace slash by path separator to properly skip tests on Windows. (#241) (@maleadt)
• Retry cudnnCreate on CUDNN_STATUS_INTERNAL_ERROR and CUDNN_STATUS_NOT_INITIALIZED (#244) (@maleadt)
• Add issue templates (#245) (@maleadt)
• Import wrapper tooling, wrap NVML (#248) (@maleadt)
• Ignore some potentially unsupported NVML features. (#251) (@maleadt)
• Assert NVPTX availability by just calling the initializer. (#252) (@maleadt)
• Update manifest (#257) (@github-actions[bot])
• Adapt to AddrSpacePtr rename. (#258) (@maleadt)
• Typo in installation overview docs (#260) (@clintonTE)
• Update GPUCompiler.jl (#266) (@maleadt)
• Retry library initialization failure due to (badly reported) OOM. (#268) (@maleadt)
• Upgrade CUTENSOR to v1.1.0. (#269) (@maleadt)
• Use CUDNN from Yggdrasil. (#272) (@maleadt)
• Update manifest (#273) (@github-actions[bot])
• Improve local CUDA discovery for CUDA 11 (#274) (@maleadt)
• Compatibility with latest LLVM and GPUCompiler (#275) (@maleadt)