v1.6.1

oneAPI v1.6.1

Diff since v1.6.0

Merged pull requests:

• Implement sincos intrinsic to fix exp. (#443) (@maleadt)
• Bump back-end compiler. (#456) (@maleadt)
• CompatHelper: bump compat for GPUCompiler to 0.27, (keep existing compat) (#457) (@github-actions[bot])
• Don't evict memory before freeing it. (#459) (@maleadt)
• Queue operations on the current device, not the one owning the array. (#460) (@maleadt)
• Bump LLVM compat (#462) (@maleadt)
• Bump oneAPI support package. (#464) (@maleadt)
• Use the version 2024.2.1 for CI (#467) (@amontoison)
• Revert "Use the version 2024.2.1 for CI" (#468) (@maleadt)

Closed issues:

• Conversions from float to integers fail (#48)
• Evicting memory before freeing? (#439)
• Segmentation fault with the release 1.6.0 (#455)
• sind and cosd not working with Float32 oneArrays (#463)

v1.6.0

oneAPI v1.6.0

Diff since v1.5.0

Merged pull requests:

• CompatHelper: bump compat for LLVM to 8, (keep existing compat) (#444) (@github-actions[bot])
• Update oneAPI.jl for the release 2024.2.0 (#446) (@amontoison)
• Avoid ReshapedArray using Int128 in kernel (#449) (@tgymnich)
• Export OutOfGPUMemoryError (#451) (@maleadt)
• Bump minimal Julia requirement to v1.10. (#452) (@maleadt)
• Bump driver. (#453) (@maleadt)

Closed issues:

• Backport Metal.jl ReshapedArray improvement (#447)
• Windows: first call to using oneAPI errors, terminates testing; second call succeeds (#448)
• ERROR: UndefVarError: OutOfGPUMemoryError not defined (#450)

v1.5.0

oneAPI v1.5.0

Diff since v1.4.0

Merged pull requests:

• Update manifest (#366) (@github-actions[bot])
• CompatHelper: bump compat for GPUCompiler to 0.25, (keep existing compat) (#367) (@github-actions[bot])
• Bump actions/checkout from 3 to 4 (#368) (@dependabot[bot])
• Update manifest (#369) (@github-actions[bot])
• CompatHelper: bump compat for CEnum to 0.5, (keep existing compat) (#370) (@github-actions[bot])
• Update manifest (#372) (@github-actions[bot])
• [oneMKL] Interface lapack routines (#376) (@amontoison)
• Add a generator of C interfaces (#378) (@amontoison)
• Support more kwarg syntax with kernel launches (#380) (@maleadt)
• Preserve the buffer type when broadcasting. (#383) (@maleadt)
• Avoid redundant pointer conversions for threadgroup memory. (#385) (@maleadt)
• Enable oneAPI 2024.0.0 support. (#386) (@kballeda)
• CompatHelper: bump compat for GPUCompiler to 0.26, (keep existing compat) (#387) (@github-actions[bot])
• Bump the compiler toolchain. (#388) (@maleadt)
• Update the deps folder for oneAPI 2024 (#389) (@amontoison)
• Bump the support library. (#390) (@maleadt)
• Bump again the support library (#391) (@amontoison)
• Avoid constructing MulAddMuls on Julia v1.12+ (#392) (@dkarrasch)
• Add a C interface for oneAPI -- SPARSE (#393) (@amontoison)
• Initial support for native Windows (#394) (@maleadt)
• Memory improvements (#395) (@maleadt)
• [oneMKL] Interface sparse routines (#400) (@amontoison)
• Update the C interface to avoid conflicts between buffer and usm routines (#402) (@amontoison)
• Upgrade the C interface for oneAPI 2024.1.0 (#403) (@amontoison)
• Bump the toolchain. (#405) (@maleadt)
• Interface more oneMKL routines (#407) (@amontoison)
• Add more methods convert in the C interface (#408) (@amontoison)
• [oneMKL] Interface EVD and SVD decompositions (#409) (@amontoison)
• Improve the dispatch for sparse routines (#410) (@amontoison)
• [oneMKL] Interface batched LAPACK routines (#411) (@amontoison)
• Add a finalizer to release the sparse matrix handle (#412) (@amontoison)
• [oneMKL] Interface sparse_optimize_* routines (#413) (@amontoison)
• Support for Julia 1.11 (#415) (@maleadt)
• Use oneMKL with Float16 *. (#416) (@maleadt)
• Use a generic OpenCL ICD loader. (#417) (@maleadt)
• Another attempt to use mkl_free_buffers (#418) (@pengtu)
• Call mkl_sycl_destructor before deleting SYCL queue (#420) (@pengtu)
• Revert call to mkl_sycl_destructor (#421) (@pengtu)
• Bump support library (#423) (@maleadt)
• Update onemkl_epilogue.cpp and onemkl_epilogue.h (#424) (@amontoison)
• Remove flush workaround. (#429) (@maleadt)
• Roll our own launch configuration (#431) (@maleadt)
• Bump julia-actions/setup-julia from 1 to 2 (#433) (@dependabot[bot])
• Make LinearAlgebra.generic_trimatmul more specific (#434) (@tgymnich)
• CompatHelper: bump compat for LLVM to 7, (keep existing compat) (#435) (@github-actions[bot])
• Define resize! (#436) (@mtfishman)
• Bump toolchain. (#438) (@maleadt)

Closed issues:

• Windows support (#92)
• res/local.jl Inconsistency in commandline options with multiple vendor libraries installed (#170)
• oneMKL: test failures with rotate and reflect (#283)
• Some test failures with A770 (#302)
• Support for Gen9 hardware regressed? (#356)
• MKL upgrade breaks support library (#357)
• Intel Data Center GPU Max series is not supported? (#365)
• Alchemist hardware (#371)
• test results on ARC770 (#373)
• Failed to initialize oneAPI (#375)
• oneapi 2024.0 - make_queue native API prototype update (#379)
• Error on Linux/NixOS (ENOENT on shared libaries) (#381)
• oneMKL tests failing due to missing symbols in oneapi support library (#384)
• oneAPI with nested integrals (#397)
• Fail to import (#399)
• When should we use FORCE_MKL_FLUSH in the C interface? (#401)
• PVC support (#406)
• oneAPI 2024.1 : Fix valid warnings encountered post updating oneAPI to 2024.1 (#422)
• Infinite partial_mapreduce_device recursion (#426)
• Remove oneMKL workaround (#427)
• Launch configuration: use ZE_extension_kernel_max_group_size_properties (#430)
• Impossible to cast a float into an integer (#441)

v1.4.0

What's Changed

• Update manifest by @github-actions in #340
• Update manifest by @github-actions in #341
• Update manifest by @github-actions in #342
• Update manifest by @github-actions in #344
• Update manifest by @github-actions in #345
• 1.10 enablement by @dkarrasch in #330
• CompatHelper: bump compat for "GPUCompiler" to "0.22" by @github-actions in #346
• Update manifest by @github-actions in #347
• enable dependabot for GitHub actions by @ranocha in #349
• Bump peter-evans/create-pull-request from 3 to 5 by @dependabot in #351
• Bump actions/checkout from 2 to 3 by @dependabot in #350
• Update manifest by @github-actions in #353
• Move to Julia 1.8. by @maleadt in #354
• Use released GPUCompiler. by @maleadt in #355
• Switch to GPUArrays buffer management. by @maleadt in #358
• Update manifest by @github-actions in #359
• Downgrade MKL. by @maleadt in #361
• Bump GPUCompiler. by @maleadt in #362
• Fix MKL tests for 1.10 compatibility by @maleadt in #364

New Contributors

• @ranocha made their first contribution in #349

Full Changelog: v1.3.0...v1.4.0

v1.3.0

oneAPI v1.3.0

Diff since v1.2.3

Merged pull requests:

• Update to LLVM.jl 6. (#334) (@maleadt)
• Update manifest (#335) (@github-actions[bot])
• Update manifest (#337) (@github-actions[bot])
• Bump binary dependencies. (#338) (@maleadt)
• Update manifest (#339) (@github-actions[bot])

v1.2.3

oneAPI v1.2.3

Diff since v1.2.2

Closed issues:

• Broken with SPIRV_LLVM_Translator_unified_jll 0.3 (#320)
• oneAPI.rand() creates a CPU array (#324)

Merged pull requests:

• Reduce number of mul! methods (#318) (@dkarrasch)
• Bump GPUCompiler. (#322) (@maleadt)
• Bump SPIRV LLVM translator. (#323) (@maleadt)
• Add GPUArrays out-of-place random methods (#325) (@tgymnich)
• Update manifest (#326) (@github-actions[bot])
• Update manifest (#328) (@github-actions[bot])
• Replace macros with functions to cut down on generated code. (#329) (@maleadt)
• Update manifest (#331) (@github-actions[bot])
• CompatHelper: bump compat for "GPUCompiler" to "0.21" (#332) (@github-actions[bot])

v1.2.2

oneAPI v1.2.2

Diff since v1.2.1

Closed issues:

• Level Zero loader picks system libze_intel_gpu (#315)

Merged pull requests:

• Add JLL debug mode (#312) (@maleadt)
• CompatHelper: add new compat entry for "Preferences" at version "1" (#313) (@github-actions[bot])
• Update manifest (#314) (@github-actions[bot])
• Update manifest (#316) (@github-actions[bot])
• Update manifest (#317) (@github-actions[bot])
• Update manifest (#319) (@github-actions[bot])

v1.2.1

oneAPI v1.2.1

Diff since v1.2.0

Merged pull requests:

• Update manifest (#307) (@github-actions[bot])
• Update manifest (#309) (@github-actions[bot])
• Generate IR validation errors on unsupported floating-point values. (#310) (@maleadt)
• Bump the binary toolchain. (#311) (@maleadt)

v1.2.0

oneAPI v1.2.0

Diff since v1.1.0

Merged pull requests:

• Add support for KernelAbstraction 0.9 (#300) (@vchuravy)

v1.1.0

oneAPI v1.1.0

Diff since v1.0.2

Closed issues:

• New version of oneAPI_Support_jll? (#298)

Merged pull requests:

• Support batch primitives (gemm_batch, trsm_batch) (#287) (@kballeda)
• Update manifest (#290) (@github-actions[bot])
• Support for creating oneArrays backed by shard memory (#291) (@maleadt)
• Use broadcast instead of copies to initialize mapreduce buffers. (#292) (@maleadt)
• Update manifest (#293) (@github-actions[bot])
• Adapt to GPUCompiler 0.18. (#295) (@maleadt)
• support gemm_batch_stride (#297) (@kballeda)
• Bump oneAPI support library. (#299) (@maleadt)
• Update manifest (#301) (@github-actions[bot])
• Choose parallel tests based on CPUs, not threads. (#304) (@maleadt)
• Adapt to LLVM.jl 5 and GPUCompiler.jl 0.19. (#306) (@maleadt)