Repository navigation
Avoid redundant host array copy during NDArray conversion - #223
Merged
Merged
Conversation
ejmeitz
added a commit
that referenced
this pull request
Oct 14, 2026
v0.3.0 (#182) - Added random number generation for all relevant types (rand, randn, randexp, etc.), mirroring cuPyNumeric (#176). - Made NDArray a subtype of AbstractArray and added LinearAlgebra.Diagonal support (#165). - Added argmin, argmax, var, std, and mean (#178). - Added more unary and binary operators (e.g. <<) and == / != as 0-d Bool reductions (#177, #187). - Added more linear-algebra operators (e.g. eigen) and vector linear-algebra ops (#183, #202). - Added cuSolverMp support (#192). - Added the new @accelerate macro with explicit Julia scoping semantics (escape analysis) and multi-line broadcast fusion, plus fused compound broadcasts and compile-time argument buffers (#174, #200, #226). - Added FFT/iFFT/bFFT, in-place and batched variants (#185, #204, #220). - Added sort, sort!, sortperm, searchsortedfirst, and searchsortedlast (#181). - Added tensor contractions and a TensorOperations.jl @tensor extension (#186, #188, #190). - Added mapreduce, NDArray map!, host copyto!, and muladd support (#196, #204, #206). - Added the NDScalar type with opt-in scalar unwrapping via allowfetch (#203, #230). - Added a Krylov.jl extension (#205). - Added AoS (native Legate) and SoA (experimental StructArrays.jl extension) support (#207). - Added undef initialization, zero-backed for users (#213, #232). - Added in-place and 1-d slice assignment (#210, #215). - Added a Dynamic Mode Decomposition example and benchmark (#184). - Fixed Int32 overflow for arrays with more than 2^31 elements. - Improved startup and conversion cost by loading CUDACore before Legate, owning Julia array conversions, and avoiding redundant host copies (#193, #209, #223). - Revamped the benchmark suite with fewer hyperparameters and removed the in-package Benchmarking module (#191, #195). - Improved CI with new GPU JLL triggers, regression CI via [regression-ci <branch>], opt-in container builds, a Julia 1.13 matrix, and tagging CI (#194, #197, #201, #216, #218, #221, #222). - Updated cuPyNumeric compat and documentation for v0.3.0 (#217, #231). - Additional CI adjustments, benchmark compat bumps, JLL wrapper updates, and docs fixes landed as direct commits without PR numbers. --------- Co-authored-by: krasow <krasow@u.northwestern.edu>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This causes a lot of overheads in benchmarks which use huge arrays.