Tags: vtjnash/CUDA.jl
Tags
## CUDA v4.1.4 [Diff since v4.1.3](JuliaGPU/CUDA.jl@v4.1.3...v4.1.4) **Closed issues:** - Buggy precompilation of init-defined symbols can break CUDA_Driver_jll initialization (JuliaGPU#1798) - Calling CUDA.set_runtime_version!() with float parameter makes CUDA.jl unusable. (JuliaGPU#1831) - Unexpexted memory allocation when using `randn!` (JuliaGPU#1856) - The memory copy speed seems to exceed the hardware limit (JuliaGPU#1860) - PCG produces different output on GPU (via Krylov.jl) (JuliaGPU#1864) **Merged pull requests:** - Fix system_driver_version on platforms not supported by CUDA_Driver_jll. (JuliaGPU#1854) (@maleadt) - Update manifest (JuliaGPU#1861) (@github-actions[bot])
## CUDA v4.1.3 [Diff since v4.1.2](JuliaGPU/CUDA.jl@v4.1.2...v4.1.3) **Closed issues:** - CUDA.versioninfo() triggers download of lazy artifacts (JuliaGPU#1844) **Merged pull requests:** - Choose parallel tests based on CPUs, not threads. (JuliaGPU#1842) (@maleadt) - Adapt to LLVM.jl 5 and GPUCompiler.jl 0.19. (JuliaGPU#1847) (@maleadt)
## CUDA v4.1.2 [Diff since v4.1.1](JuliaGPU/CUDA.jl@v4.1.1...v4.1.2) **Closed issues:** - Flux's `gradient` differentiating`rfft` leads to non-bit error (JuliaGPU#1835) **Merged pull requests:** - switch to using defined globals (JuliaGPU#1832) (@simonbyrne) - Update manifest (JuliaGPU#1837) (@github-actions[bot])
## CUDA v4.1.1 [Diff since v4.1.0](JuliaGPU/CUDA.jl@v4.1.0...v4.1.1) **Merged pull requests:** - Fix export of CUDABackend (JuliaGPU#1834) (@vchuravy)
## CUDA v4.1.0 [Diff since v4.0.1](JuliaGPU/CUDA.jl@v4.0.1...v4.1.0) **Closed issues:** - ERROR: LoadError: bin\cublas64_11.dll when installing CUDA (JuliaGPU#1750) - System-wide CUDA in LD_LIBRARY_PATH breaks CUBLAS (JuliaGPU#1755) - CuDeviceTexture getindex breaks when executed on the CPU (JuliaGPU#1757) - cuDNN.version can cause Julia to crash, missing `cudnn_ops_infer64_8.dll` (JuliaGPU#1777) - cuDNN compile error "ERROR: LoadError: ArgumentError: invalid version string: local" (JuliaGPU#1783) - "Error: No CUDA Runtime library found" for ≥v4.0.0 (JuliaGPU#1808) - sqrt broken in kernels 'Format of __nvvm__reflect function not recognized' (JuliaGPU#1817) **Merged pull requests:** - Add support for CUDA 12.0. (JuliaGPU#1742) (@maleadt) - Add more fixes and tests for CUDA toolkit 12.0 (JuliaGPU#1756) (@amontoison) - Update manifest (JuliaGPU#1758) (@github-actions[bot]) - Fix test/cusparse/interfaces.jl (JuliaGPU#1762) (@amontoison) - Simplify the function sig. (JuliaGPU#1763) (@N5N3) - Update manifest (JuliaGPU#1770) (@github-actions[bot]) - Make versioninfo() resilient against NVML EPERM. (JuliaGPU#1771) (@maleadt) - Move CUDAKernels to CUDA.jl (JuliaGPU#1772) (@vchuravy) - [CUSPARSE] Improve conversion and tests between sparse matrices (JuliaGPU#1774) (@amontoison) - Use geam for + and - operations with CuMatrix{<:CublasFloat} (JuliaGPU#1775) (@amontoison) - Update manifest (JuliaGPU#1776) (@github-actions[bot]) - Update manifest (JuliaGPU#1781) (@github-actions[bot]) - Update manifest (JuliaGPU#1784) (@github-actions[bot]) - [CUSPARSE] Update preconditioners.jl (JuliaGPU#1785) (@amontoison) - [CUSOLVER] Avoid the conversion to CSR format for reordering routines (JuliaGPU#1786) (@amontoison) - Bump GPUCompiler. (JuliaGPU#1787) (@maleadt) - Remove unneeded variable. (JuliaGPU#1788) (@maleadt) - [CUSPARSE] Update conversions.jl (JuliaGPU#1791) (@amontoison) - Update to CUDNN 8.8.1 for CUDA 12 compatibility. (JuliaGPU#1792) (@maleadt) - Add support for CUDA 12.1 (JuliaGPU#1793) (@maleadt) - [CUSPARSE] Interface color reordering (JuliaGPU#1794) (@amontoison) - [CUSPARSE] Interface gtsv2 (JuliaGPU#1795) (@amontoison) - Update manifest (JuliaGPU#1796) (@github-actions[bot]) - Adapt to GPUCompiler 0.18 (JuliaGPU#1799) (@maleadt) - Follow `Array`'s behavior when initializing (JuliaGPU#1800) (@lcw) - [CUSOLVER] Support A \ b for rectangular matrices (JuliaGPU#1802) (@amontoison) - Use symbols instead of values when emitting code, when possible. (JuliaGPU#1804) (@maleadt) - Refactor CI pipeline a little. (JuliaGPU#1805) (@maleadt) - [CUSOLVER] Improve the dispatch for LAPACK routines (JuliaGPU#1806) (@amontoison) - Diagonal for lower triangular of LU decomposition set incorrectly (JuliaGPU#1813) (@tgymnich) - CompatHelper: add new compat entry for "KernelAbstractions" at version "0.9" (JuliaGPU#1824) (@github-actions[bot]) - Rebuild CUPTI API with support for STRUCT_SIZE (JuliaGPU#1827) (@vchuravy) - Release CUDA 4.1 (JuliaGPU#1828) (@vchuravy)
## CUDA v4.0.0 [Diff since v3.13.1](JuliaGPU/CUDA.jl@v3.13.1...v4.0.0) **Closed issues:** - Missing implementation of right multiply for QR decomposition (JuliaGPU#1738) - [CUSPARSE] Type error with mm! (JuliaGPU#1743) **Merged pull requests:** - Implement rmul for qr. (JuliaGPU#1739) (@maleadt) - Update manifest (JuliaGPU#1741) (@github-actions[bot]) - Update CUSPARSE for CUDA v12.0 (JuliaGPU#1744) (@amontoison) - Fix nvprof command (JuliaGPU#1745) (@lucifer1004) - Update manifest (JuliaGPU#1747) (@github-actions[bot]) - Fix grammar (JuliaGPU#1748) (@lucifer1004)
## CUDA v3.13.1 [Diff since v3.13.0](JuliaGPU/CUDA.jl@v3.13.0...v3.13.1) **Closed issues:** - CUDA.jl cuFFT underperforming against CuPy cuFFT (JuliaGPU#1682) - Is block-spmm supported？ (JuliaGPU#1736) **Merged pull requests:** - Introduce cuFFT plan cache; switch to auto-managed memory. (JuliaGPU#1734) (@maleadt) - Stop pirating GPUArrays' RNG methods. (JuliaGPU#1735) (@maleadt)
## CUDA v3.12.2 [Diff since v3.12.1](JuliaGPU/CUDA.jl@v3.12.1...v3.12.2) **Closed issues:** - CUDA.jl cuFFT underperforming against CuPy cuFFT (JuliaGPU#1682) - Error during CUDA test (JuliaGPU#1718) - Kernel error from bad broadcast (should be regular error?) (JuliaGPU#1720) - Freeze into StackOverflow when `JULIA_DEBUG=CUDA` set (JuliaGPU#1721) - Use of linear operators in CUDA.jl (JuliaGPU#1727) - Is block-spmm supported？ (JuliaGPU#1736) **Merged pull requests:** - Allow `copy(::RNG)` (JuliaGPU#1719) (@mcabbott) - Update manifest (JuliaGPU#1722) (@github-actions[bot]) - Simplify CuError rendering before library initialization. (JuliaGPU#1723) (@maleadt) - Simplify CuError rendering before library initialization (master branch version) (JuliaGPU#1724) (@maleadt) - Make device RNG test more robust. (JuliaGPU#1725) (@maleadt) - Rely on LLVM.jl's typed_ccall for more intrinsics. (JuliaGPU#1728) (@maleadt) - Backports for 3.13 (JuliaGPU#1729) (@maleadt) - Simplify CUBLAS and CUSPARSE wrappers, reducing code generated. (JuliaGPU#1730) (@maleadt) - Add Julia 1.9 CI. (JuliaGPU#1731) (@maleadt) - Use released dependencies. (JuliaGPU#1732) (@maleadt) - Remove NVTX. (JuliaGPU#1733) (@maleadt) - Introduce cuFFT plan cache; switch to auto-managed memory. (JuliaGPU#1734) (@maleadt) - Stop pirating GPUArrays' RNG methods. (JuliaGPU#1735) (@maleadt)
## CUDA v3.13.0 [Diff since v3.12.1](JuliaGPU/CUDA.jl@v3.12.1...v3.13.0) **Closed issues:** - Error during CUDA test (JuliaGPU#1718) - Kernel error from bad broadcast (should be regular error?) (JuliaGPU#1720) - Freeze into StackOverflow when `JULIA_DEBUG=CUDA` set (JuliaGPU#1721) - Use of linear operators in CUDA.jl (JuliaGPU#1727) **Merged pull requests:** - Allow `copy(::RNG)` (JuliaGPU#1719) (@mcabbott) - Update manifest (JuliaGPU#1722) (@github-actions[bot]) - Simplify CuError rendering before library initialization. (JuliaGPU#1723) (@maleadt) - Simplify CuError rendering before library initialization (master branch version) (JuliaGPU#1724) (@maleadt) - Make device RNG test more robust. (JuliaGPU#1725) (@maleadt) - Rely on LLVM.jl's typed_ccall for more intrinsics. (JuliaGPU#1728) (@maleadt) - Backports for 3.13 (JuliaGPU#1729) (@maleadt) - Simplify CUBLAS and CUSPARSE wrappers, reducing code generated. (JuliaGPU#1730) (@maleadt) - Add Julia 1.9 CI. (JuliaGPU#1731) (@maleadt) - Use released dependencies. (JuliaGPU#1732) (@maleadt) - Remove NVTX. (JuliaGPU#1733) (@maleadt)
PreviousNext