JuliaGPU/CUDA.jl issues and pull requests

#1910 - Stream synchronization is slow when waiting on the event from CUDA

Issue - State: closed - Opened by utkinis over 1 year ago - 8 comments
Labels: bug

#1909 - CompatHelper: bump compat for "GPUCompiler" to "0.20"

Pull Request - State: closed - Opened by github-actions[bot] over 1 year ago - 1 comment

#1908 - Bump GPUCompiler.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 2 comments

#1907 - Update manifest

Pull Request - State: closed - Opened by github-actions[bot] over 1 year ago - 1 comment

#1906 - Avoid unneeded macros to cut down on generated code

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1905 - Avoid unneeded macros to cut down on generated code

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1904 - Reduce load time by shifting `mul!` definition

Pull Request - State: open - Opened by dkarrasch over 1 year ago - 6 comments

#1903 - Added support for more transform directions

Pull Request - State: closed - Opened by RainerHeintzmann over 1 year ago - 30 comments

#1902 - Array testsuite compiles kernel with large types

Issue - State: open - Opened by maleadt over 1 year ago - 1 comment

#1901 - Insufficient coverage of CuArrays in the documentation

Issue - State: closed - Opened by emojjon over 1 year ago - 3 comments
Labels: documentation

#1900 - Restrict number of test jobs based on available memory.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment
Labels: tests

#1899 - Reverse with multiple dimensions

Pull Request - State: closed - Opened by RainerHeintzmann over 1 year ago - 5 comments

#1898 - BSOD during package tests

Issue - State: closed - Opened by RainerHeintzmann over 1 year ago - 14 comments
Labels: bug, needs information

#1897 - Update manifest

Pull Request - State: closed - Opened by github-actions[bot] over 1 year ago - 1 comment

#1896 - Odd behavior when loading cuDNN in Flux.

Issue - State: open - Opened by mkschleg over 1 year ago - 5 comments
Labels: bug, needs information

#1895 - WIP: Add an index typevar to CuDeviceArray.

Pull Request - State: open - Opened by maleadt over 1 year ago - 11 comments
Labels: enhancement, help wanted, performance

#1894 - Lovelace fixes

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1893 - QR decomposition requires scalar indexing

Issue - State: closed - Opened by huntekye over 1 year ago - 3 comments
Labels: bug

#1892 - Erratic behaviour of `CUDA.jl` if used in the REPL of VSCode.

Issue - State: closed - Opened by RainerHeintzmann over 1 year ago - 1 comment

#1891 - irfft of real valued CuArray returns a non-CuArray result

Issue - State: open - Opened by RainerHeintzmann over 1 year ago - 1 comment
Labels: bug, good first issue, cuda libraries

#1890 - CUSOLVER: Fix workspace size passing.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment
Labels: bugfix

#1889 - Update manifest

Pull Request - State: closed - Opened by github-actions[bot] over 1 year ago - 1 comment

#1888 - Incorrect colVal array when using CuSparseMatrixCSR command on sparse matrix

Issue - State: closed - Opened by HamMoh94 over 1 year ago - 8 comments
Labels: bug, upstream

#1887 - Cannot construct CuDeviceArray

Issue - State: closed - Opened by 90259025 over 1 year ago - 2 comments

#1886 - Error thrown when launching Julia with Nsight systems or compute.

Issue - State: closed - Opened by jlk9 over 1 year ago - 1 comment
Labels: bug, upstream

#1885 - Fix atomic increment of alloc stats.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1884 - Use atomics for allocation statistics.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1883 - Update CUDA 12.1 to 12.1.1.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1882 - Update manifest

Pull Request - State: closed - Opened by github-actions[bot] over 1 year ago - 1 comment

#1881 - Guard the initialization-time CUDA driver check in a try/catch.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment
Labels: installation

#1880 - Integer overflow error with svd for large matrix

Issue - State: closed - Opened by hsseung over 1 year ago - 2 comments
Labels: bug

#1879 - Fake CUDA library

Issue - State: closed - Opened by michel2323 over 1 year ago - 2 comments
Labels: bug, needs information

#1878 - Remove more unneeded ndims methods.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1877 - Runtime not re-selected after driver upgrade

Issue - State: closed - Opened by maleadt over 1 year ago - 4 comments
Labels: bug, installation

#1876 - remove some type-piracy from cusparse

Pull Request - State: closed - Opened by vtjnash over 1 year ago - 4 comments

#1875 - Incorrect kernel execution with bounds checking using Julia 1.9.0-rc2

Issue - State: closed - Opened by lcw over 1 year ago - 5 comments
Labels: bug

#1874 - Sort fails on Lovelace (sm8.9) GPUs

Issue - State: closed - Opened by xaellison over 1 year ago - 4 comments
Labels: bug

#1873 - Add compatibility entry for Lovelace

Pull Request - State: closed - Opened by xaellison over 1 year ago - 1 comment

#1872 - Don't collect `AbstractQ` objects in tests

Pull Request - State: closed - Opened by dkarrasch over 1 year ago - 8 comments
Labels: cuda array

#1871 - Update manifest

Pull Request - State: closed - Opened by github-actions[bot] over 1 year ago

#1870 - Cannot precompile GPU code with SnoopPrecompile

Issue - State: closed - Opened by marius311 over 1 year ago - 2 comments
Labels: bug, upstream

#1869 - Add Enzyme Forward mode custom rule

Pull Request - State: closed - Opened by wsmoses over 1 year ago - 33 comments
Labels: upstream

#1868 - Assert CUDA.jl is functional when creating the TLS.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1867 - Passing tests using Github CI workflow errors with `libcuda not defined`

Issue - State: closed - Opened by b-fg over 1 year ago - 4 comments

#1866 - Test errors on master

Issue - State: closed - Opened by maleadt over 1 year ago
Labels: bug

#1865 - Allow precompiling CUDA code.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 3 comments

#1864 - PCG produces different output on GPU (via Krylov.jl)

Issue - State: closed - Opened by zjwegert over 1 year ago - 2 comments
Labels: bug

#1863 - Add KernelAbstractions.jl unsafe_free!

Pull Request - State: closed - Opened by pxl-th over 1 year ago

#1861 - Update manifest

Pull Request - State: closed - Opened by github-actions[bot] over 1 year ago - 1 comment

#1860 - The memory copy speed seems to exceed the hardware limit

Issue - State: closed - Opened by Yangyang-Tan over 1 year ago

#1859 - Compat driver: preload all libraries

Issue - State: closed - Opened by maleadt over 1 year ago
Labels: bug

#1858 - Support runtime discovery for non-toolkit libraries (CUTENSOR, CUDNN, CUQUANTUM)

Pull Request - State: closed - Opened by mloubout over 1 year ago - 1 comment
Labels: installation

#1857 - Creating a sparse diagonal matrix of CuArray(u)

Issue - State: closed - Opened by HamMoh94 over 1 year ago - 3 comments
Labels: enhancement, good first issue

#1856 - Unexpexted memory allocation when using `randn!`

Issue - State: closed - Opened by Yangyang-Tan over 1 year ago - 2 comments

#1855 - Support Cholesky factorization of CuSparseMatrixCSR

Issue - State: closed - Opened by shakedregev over 1 year ago - 12 comments
Labels: enhancement, help wanted, cuda libraries

#1854 - Fix system_driver_version on platforms not supported by CUDA_Driver_jll.

Pull Request - State: closed - Opened by maleadt over 1 year ago - 1 comment

#1853 - Add an experimental opaque closure type.

Pull Request - State: open - Opened by maleadt over 1 year ago - 4 comments
Labels: enhancement, cuda kernels, speculative

#1852 - CUBLAS fails on new CUDA.jl v4

Issue - State: open - Opened by marmarelis over 1 year ago - 4 comments
Labels: bug, needs information

#1851 - Can't SnoopPrecompile with CUDA 4

Issue - State: open - Opened by marius311 over 1 year ago - 2 comments
Labels: bug

#1850 - Local runtime discovery does not work for external libraries (CUDNN, CUTENSOR)

Issue - State: closed - Opened by peter-fields over 1 year ago - 7 comments
Labels: bug, installation

#1848 - dlclose'ing the compatibility driver can fail

Issue - State: closed - Opened by maleadt over 1 year ago - 1 comment
Labels: bug, installation

#1846 - Support for Hopper (H100, GH200) GPUs

Issue - State: open - Opened by smartalecH over 1 year ago - 15 comments
Labels: enhancement

#1839 - Error during cuDeviceGetMemPool causes stackoverflow

Issue - State: open - Opened by axsk over 1 year ago - 4 comments
Labels: bug

#1838 - KernelAbstractions.jl-related issues

Issue - State: closed - Opened by maleadt over 1 year ago - 2 comments
Labels: enhancement

#1836 - Windows Defender mis-labels artifacts as threat

Issue - State: closed - Opened by rtdietrich over 1 year ago - 5 comments
Labels: upstream

#1833 - Support for JLD2

Issue - State: closed - Opened by denizyuret over 1 year ago - 15 comments
Labels: bug, help wanted

#1831 - Calling CUDA.set_runtime_version!() with float parameter makes CUDA.jl unusable.

Issue - State: closed - Opened by jlk9 over 1 year ago - 2 comments
Labels: bug

#1830 - Attempted to add method definitions for supporting the SubArray type …

Pull Request - State: open - Opened by JamieMair over 1 year ago

#1825 - Scalar indexing in CUSPARSE by mutipliying Complex Sparse Matrix by Float64 CuArray

Issue - State: open - Opened by walexaindre over 1 year ago - 2 comments
Labels: bug

#1823 - Add wrappers for NVPERF

Pull Request - State: open - Opened by vchuravy over 1 year ago

#1821 - Feature request: Integrating the latest CUDA library "cuLitho" into CUDA.jl

Issue - State: closed - Opened by algorithmx over 1 year ago - 2 comments
Labels: enhancement

#1820 - `cusparseSetStream_v2` not defined

Issue - State: closed - Opened by maleadt over 1 year ago - 1 comment
Labels: bug

#1819 - Use NVTX.jl to give streams a name

Issue - State: open - Opened by simonbyrne almost 2 years ago - 13 comments
Labels: enhancement, good first issue

#1819 - Use NVTX.jl to give streams a name

Issue - State: open - Opened by simonbyrne almost 2 years ago - 13 comments
Labels: enhancement, good first issue

#1818 - Allow but warn when testing on not fully-supported devices.

Pull Request - State: closed - Opened by maleadt over 1 year ago
Labels: tests

#1815 - Testing fails on unsupported devices.

Issue - State: closed - Opened by stmorgenstern over 1 year ago
Labels: tests

#1814 - dlopen("libcudart") results in duplicate libraries

Issue - State: closed - Opened by maleadt over 1 year ago - 7 comments
Labels: bug, installation

#1811 - Add EnzymeRules support for CUDA.jl (for forward mode here)

Issue - State: closed - Opened by wsmoses over 1 year ago - 2 comments
Labels: enhancement

#1810 - accumulate(+) gives different results for CuArray compared to Array.

Issue - State: closed - Opened by CamMkI over 1 year ago
Labels: bug, good first issue

#1810 - accumulate(+) gives different results for CuArray compared to Array.

Issue - State: closed - Opened by CamMkI over 1 year ago
Labels: bug, good first issue

#1810 - accumulate(+) gives different results for CuArray compared to Array.

Issue - State: closed - Opened by CamMkI over 1 year ago
Labels: bug, good first issue

#1809 - Initialize globals to nothing

Pull Request - State: closed - Opened by simonbyrne over 1 year ago - 11 comments

#1808 - "Error: No CUDA Runtime library found" for ≥v4.0.0

Issue - State: closed - Opened by wsshin over 1 year ago - 13 comments

#1807 - Use try catch instead of isdefined

Pull Request - State: closed - Opened by simonbyrne over 1 year ago - 1 comment

#1798 - Buggy precompilation of init-defined symbols can break CUDA_Driver_jll initialization

Issue - State: closed - Opened by simonbyrne over 1 year ago - 26 comments
Labels: bug, upstream

#1790 - Use Atomix

Pull Request - State: open - Opened by vchuravy over 1 year ago - 4 comments

#1782 - `LinearAlgebra.norm(x)` falls back to generic implementation for `x::Transpose` and `x::Adjoint`

Issue - State: closed - Opened by jondeuce over 1 year ago - 6 comments
Labels: bug

#1782 - `LinearAlgebra.norm(x)` falls back to generic implementation for `x::Transpose` and `x::Adjoint`

Issue - State: closed - Opened by jondeuce over 1 year ago - 6 comments
Labels: bug

#1779 - Trouble using nsight systems for profiling CUDA in Julia

Issue - State: closed - Opened by JamieMair over 1 year ago - 4 comments
Labels: bug, needs information, upstream

#1779 - Trouble using nsight systems for profiling CUDA in Julia

Issue - State: closed - Opened by JamieMair over 1 year ago - 4 comments
Labels: bug, needs information, upstream

#1779 - Trouble using nsight systems for profiling CUDA in Julia

Issue - State: closed - Opened by JamieMair over 1 year ago - 4 comments
Labels: bug, needs information, upstream

#1778 - Adding copyto for non-contiguous matrices and vectors

Pull Request - State: open - Opened by evelyne-ringoot over 1 year ago - 3 comments
Labels: cuda array

#1769 - Use `adapt` symmetrically in `CuIterator`

Pull Request - State: closed - Opened by mcabbott almost 2 years ago - 4 comments
Labels: enhancement

#1768 - Limitations of `CuIterator`

Issue - State: closed - Opened by mcabbott almost 2 years ago

#1767 - [CUSOLVER] Add a with_workspaces function to allocate two buffers (Device / Host)

Issue - State: closed - Opened by amontoison almost 2 years ago
Labels: enhancement, good first issue, cuda libraries

#1767 - [CUSOLVER] Add a with_workspaces function to allocate two buffers (Device / Host)

Issue - State: closed - Opened by amontoison almost 2 years ago
Labels: enhancement, good first issue, cuda libraries

#1767 - [CUSOLVER] Add a with_workspaces function to allocate two buffers (Device / Host)

Issue - State: closed - Opened by amontoison almost 2 years ago
Labels: enhancement, good first issue, cuda libraries

#1764 - Support for qr of strided inputs (non-contiguous views)

Pull Request - State: open - Opened by evelyne-ringoot almost 2 years ago - 7 comments
Labels: enhancement, cuda libraries

#1761 - Type constructor in broadcast doesn't compile

Issue - State: closed - Opened by bicycle1885 almost 2 years ago - 6 comments
Labels: bug, help wanted

#1746 - Unreachable control flow leads to illegal divergent barriers

Issue - State: closed - Opened by parfenyev almost 2 years ago - 23 comments
Labels: bug, needs information

GitHub / JuliaGPU/CUDA.jl issues and pull requests