2888 Commits

Author SHA1 Message Date
DRayX 4c22ac5fe4 Update lombok imports to improve delombok compatibility (#10416)
* Replace wildcard lombok imports with single class imports

This helps improve compatibility with delombok

* Removed unused lombok imports

Delombok doesn't remove unused imports
2026-02-24 05:13:17 +09:00
Adam Gibson 63f649dc93 Update helper functions (#10374)
* Update helper functions

Updates to helper implementations:
- MmulHelper CPU/CUDA
- ShapeUtils, ShapeBuilders
- Attention, BLAS helpers
- Matrix operations (LU, QR, Hessenberg)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* Update libnd4j/include/helpers/impl/ShapeUtils.cpp

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update libnd4j/include/helpers/ShapeBuilders.h

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update libnd4j/include/helpers/impl/hhColPivQR.cpp

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update libnd4j/include/helpers/impl/ShapeUtils.cpp

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Fix COORDS2INDEX macro to use strides instead of shapes (#10393)

* Initial plan

* Fix COORDS2INDEX macro usage to use strides instead of shapes

Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

* Fix memory leak in MmulHelper::mmulNxN from undeleted cExpectedShape (#10394)

* Initial plan

* Fix memory leak in MmulHelper.cpp by deleting cExpectedShape

Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

* Initial plan (#10392)

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2025-12-28 01:17:58 +09:00
Adam Gibson 5b6278b27d Update NativeOps and NativeOpExecutioner (#10372)
* Update NativeOps and NativeOpExecutioner

Refactored NativeOps layer:
- Split NativeOpExecutioner into focused compilation units
- Added lifecycle tracking for DataBuffer and NDArray
- Updated CUDA implementations

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* Update libnd4j/include/legacy/impl/NativeOpsHelpers_Arrays.cpp

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update libnd4j/include/legacy/cpu/NativeOpExecutioner_transform.cpp

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Add JavaCPP signature mapping documentation for setGraphContextOutputArraysArr (#10408)

* Initial plan

* Add comprehensive comment explaining signature change for setGraphContextOutputArraysArr

Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

* Update libnd4j/include/legacy/cuda/NativeOpExecutioner_transform.cu

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2025-12-28 01:17:12 +09:00
Adam Gibson bbdc8358b9 Remove unused imports from DL4J and ND4J layer classes (#10362)
Removes unused import statements from 15 Java files.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:09:19 +09:00
Adam Gibson 5348dc2702 Standardize CUDA macro naming: __CUDABLAS__ to SD_CUDA (#10363)
Renames __CUDABLAS__ to SD_CUDA for consistency with other SD_ prefixed macros.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:08:57 +09:00
Adam Gibson a9ea5c2e80 Fix memory leaks in C++ ops and add lifecycle tracking (#10365)
- Fixes memory leaks in 131 op implementations by adding proper delete calls
- Adds lifecycle tracking infrastructure for NDArray, DataBuffer, and caches
- Pattern: converts stack temporaries to pointer allocation with cleanup

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:08:35 +09:00
Adam Gibson 1f8167ecd2 Update loop indexing macros and type handling (#10364)
Updates INDEX2COORDS/COORDS2INDEX usage and type handling in loop implementations.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:08:13 +09:00
Adam Gibson 966ab50cdc Fix C++ helper implementations (#10367)
* Fix C++ helper implementations

- TAD calculation fixes in gather, softmax
- Thread safety improvements (BLAS lock in batched_gemm)
- Edge case handling for empty arrays
- Various algorithm fixes in CPU/CUDA helpers

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* Fix memory leak in cell_contains on early return

Restructure loop to avoid early returns that bypassed delete calls.
Uses result variable to track outcome and ensures cleanup always runs.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:07:49 +09:00
Adam Gibson ec480fad4e Add ONNX transformer model import support (#10366)
Adds support for importing transformer/BERT models from ONNX:
- Attention, LayerNormalization, SkipLayerNormalization
- EmbedLayerNormalization, FastGelu, BiasGelu
- Microsoft ONNX Runtime custom ops

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:07:19 +09:00
Adam Gibson 95d5708792 Add SameDiff control flow execution framework (#10368)
New classes for frame-aware loop and conditional execution:
- Frame analysis (FrameAnalyzer, FrameExecutionOptimizer)
- Loop state tracking (LoopInfo, LoopState, LoopAnalysisHelpers)
- Cross-frame references (CrossFrameReference, CrossLoopAnalysis)
- Loop termination analysis and error reporting
- Execution visualization (FrameVisualizer, SameDiffExecutionVisualizer)

Updates to existing execution infrastructure:
- AbstractSession, InferenceSession
- ForwardExecutionDAG, ExecutionNode
- FlatBuffersMapper, SDZSerializer

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:06:55 +09:00
Adam Gibson 47757e7dc5 Update DL4J layer implementations (#10369)
Updates to neural network layers:
- Convolution layers (1D, 2D, Deconv, Depthwise, Separable)
- Normalization (BatchNorm, LocalResponseNorm)
- Recurrent layers (RNN, LSTM, Bidirectional)
- Output layers (Loss, RNN, OCNN, VAE)
- Base layer classes

Additional updates:
- MultiLayerNetwork
- ModelSerializer
- WordVectorSerializer
- TFOpLayerImpl

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-23 21:06:33 +09:00
Adam Gibson 9fb4ace726 Modernize LibND4J CMake build system (#10245)
* Add files for pr_2_1_cmake_core

* CMake build system improvements: clang support, memory sanitizers, parallelization

* Remove unrelated submodule references

* Fix oneDNN git repository URL
2025-12-19 04:40:18 +09:00
Adam Gibson 196d7b2f4e Simplify batched GEMM API (#10361)
* Add files for pr_6_1_op_headers

* Add files for pr_6_2_generic_blas_basic

* Where op: memory handling improvements

* Remove unrelated submodule references
2025-12-18 21:13:59 +09:00
Adam Gibson 2fb4ea96ca Add files for pr_9_4_autodiff_execution (#10283) 2025-12-18 20:49:21 +09:00
Adam Gibson 072bf96f53 Better error messages when ops fail. Replaces some print only error messages/warnings in the core op classes. (#10268)
* Add files for pr_6_13_op_impl_base

* DeclarableOp: execution context memory management

* Remove engineers-guide-to-llms and pp-step-clone directories

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-16 18:19:12 +09:00
Adam Gibson e07b9d6707 BUILD_TEMPLATE* calls changes, reshape changes in various cpu ops. (#10262)
* Add files for pr_6_7_cpu_helpers_1

* CPU helpers: GEMM and convolution memory fixes

* Remove engineers-guide-to-llms and pp-step-clone directories

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-16 18:18:45 +09:00
Adam Gibson ced73f1c80 Updates github actions builds. (#10242)
* Add files for pr_1_1_github_actions

* Update GitHub Actions workflows with CUDA 12.3 build fixes

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* Update .github/actions/install-cmake-linux/action.yml

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update .github/workflows/build-deploy-linux-cuda-12.6.yml

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update .github/workflows/build-deploy-linux-cuda-12.6.yml

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update .github/workflows/build-deploy-android-arm64.yml

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update .github/workflows/build-deploy-android-arm64.yml

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update .github/workflows/build-deploy-android-arm64.yml

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Initial plan (#10355)

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

* Initial plan (#10357)

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

* [WIP] Address feedback on GitHub Actions builds updates (#10359)

* Initial plan

* Add missing releaseRepoId input and matrix variable to build-deploy-linux-x86_64.yml

Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

* [WIP] Address feedback on GitHub Actions builds updates (#10358)

* Initial plan

* Remove redundant MAVEN_GPG_KEY and GPG_SIGNING_KEY env vars from Windows workflow

Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: agibsonccc <1144306+agibsonccc@users.noreply.github.com>

* Initial plan (#10360)

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
2025-12-16 18:18:05 +09:00
Adam Gibson e1bda9dbfb Maven version updates for lombok for java 25 support (#10243)
* Add files for pr_1_3_maven_poms

* Update Maven POMs for CUDA backend and platform tests

* Remove engineers-guide-to-llms and pp-step-clone directories

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-16 18:17:31 +09:00
Adam Gibson 73752a3723 Better type handling with random generator. (#10252)
* Add files for pr_4_2_graph_execution

* Graph execution context: memory management improvements

* Remove CRITICAL style comments, remove engineers-guide-to-llms and pp-step-clone

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* Fix truncated RNG.h - redirect to RandomGenerator.h

The file was truncated mid-license header. Since RandomGenerator.h
contains the actual implementation, RNG.h now simply includes it.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-16 18:16:26 +09:00
Adam Gibson ef04c8bfad Autodiff core improvements (#10280)
* Add files for pr_9_1_autodiff_core

* Autodiff core: memory management improvements

* Remove engineers-guide-to-llms and pp-step-clone directories

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

* Fix ControlFlow.java - restore full implementation with outputVarNames support

Adds overloaded methods for loopBody and invokeParams to support
passing outputVarNames to the parent graph.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-16 18:15:40 +09:00
Adam Gibson 2fda04f557 Improved SIMD detection for various ops to remove warnings. (#10272)
* Add files for pr_7_4_ops_core

* Ops header: type definitions update

* Remove engineers-guide-to-llms and pp-step-clone directories

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-16 15:29:29 +09:00
Adam Gibson 9d360daed8 Add files for pr_5_3_cuda_loops (#10255) 2025-12-16 15:17:39 +09:00
Adam Gibson 6b708bb384 CPU side selective rendering, BUILD_TEMPLATE* call changes, SD_CUDA changes. (#10250)
* Add files for pr_3_2_array_impl

* Array implementations: buffer cleanup fixes, memory leak patches

* Remove engineers-guide-to-llms and pp-step-clone

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-14 03:50:45 +09:00
Adam Gibson 32e4e96c70 Improved type coverage for core cpu ops using ITERATE_COMBINATIONS, (#10254)
* Add files for pr_5_2_cpu_loops

* CPU loops: broadcasting and summary stats fixes

* Remove engineers-guide-to-llms and pp-step-clone

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-14 03:50:13 +09:00
Adam Gibson b035cd6e67 Add files for pr_6_4_generic_parity (#10259) 2025-12-14 03:49:44 +09:00
Adam Gibson 82b176533c Add files for pr_6_12_helper_impl (#10267) 2025-12-14 03:49:04 +09:00
Adam Gibson 45d1546fbc Add files for pr_7_7_helpers_cuda (#10274) 2025-12-14 03:48:26 +09:00
Adam Gibson 2ab89cdaf9 Add files for pr_6_9_cpu_compilation_units (#10264) 2025-12-14 03:47:53 +09:00
Adam Gibson ce2c14c2e0 Cuda side BUILD_TEMPLATE* changes. (#10265)
* Add files for pr_6_10_cuda_helpers_1

* CUDA helpers: extract_patches memory fix

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 05:07:18 +09:00
Adam Gibson df66f7a758 Add files for pr_6_3_generic_nn_core (#10258) 2025-12-12 05:06:42 +09:00
Adam Gibson b9006ccf6b Misc casting, reshape pointer changes across files. (#10275)
* Add files for pr_7_8_helpers_impl

* Helper implementations: memory management and reshape fixes

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 05:06:10 +09:00
Adam Gibson 3c8d0cc8f5 Enhanced dependency tracking for building an execution graph. (#10284)
* Add files for pr_9_5_autodiff_internal

* InferenceSession: execution memory handling

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 05:05:01 +09:00
Adam Gibson 6ae78382c6 Misc casting, BUILD_TEMPLATE, reshape cleanup (#10263)
* Add files for pr_6_8_cpu_helpers_2

* CPU helpers: LRN, skip-gram, softmax fixes

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 05:03:32 +09:00
Adam Gibson e1403bccbb Add files for pr_4_1_graph_schema (#10251) 2025-12-12 05:03:01 +09:00
Adam Gibson 4bfb945095 CPU side casting/reduction loops templates fixes (#10273)
* Add files for pr_7_6_helpers_cpu

* CPU helpers: TAD calculator and SVD memory fixes

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 05:01:42 +09:00
Adam Gibson 9ec0be84a9 Misc clean up moving long long definitions to sd::LongType, more BUILD_TEMPLATE* call keyword changes. (#10253)
* Add files for pr_5_1_loop_headers

* Loop headers: type system updates, template fixes

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 05:01:11 +09:00
Adam Gibson ae47f4daee Cuda definitions refactoring, native ops data type conversions implementation. (#10270)
* Add files for pr_7_2_legacy_impl

* Legacy impl: environment and cnpy updates

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 05:00:41 +09:00
Adam Gibson 365bc4aa32 Improved type checking for native ops: checks for types being present when using certain types. (#10269)
* Add files for pr_7_1_legacy_native

* NativeOps: lifecycle tracking integration, memory fixes

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 04:59:50 +09:00
Adam Gibson e845155545 Array data type, rehsape pointer and scalar refactoring. (#10249)
* Add files for pr_3_1_array_core

* Array core headers: memory management fixes, NaN prevention

* Remove pp-step-clone and engineers-guide-to-llms folders

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-12 04:59:14 +09:00
Adam Gibson fa19efd587 Add files for pr_6_5_generic_reduce (#10260) 2025-12-11 13:21:50 +09:00
Adam Gibson 6a633a622c Add files for pr_6_6_generic_shape_transforms (#10261) 2025-12-11 13:21:25 +09:00
Adam Gibson ed63dd71e3 Add files for pr_6_11_cuda_helpers_2 (#10266) 2025-12-10 15:54:54 +09:00
Adam Gibson f522a3e417 Add files for pr_9_8_nd4j_resources (#10287) 2025-12-10 15:45:08 +09:00
Adam Gibson c42cf91a65 Build system and type system: selective rendering, template processing, type validation (#10347)
- Implements selective rendering CMake modules for type-based compilation
- Adds template processing infrastructure for code generation
- Implements type validation for compile-time type checking
- Updates buildnativeoperations.sh with new build options

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Claude <noreply@anthropic.com>
2025-12-09 17:28:32 +09:00
Adam Gibson 61a0c89fc9 System types and macros: type definitions, boilerplate generation (#10348)
- Updates core type definitions in types.h
- Improves type boilerplate expansion macros
- Adds type instantiation boilerplate helpers
- Updates op_boilerplate.h with new template macros

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Claude <noreply@anthropic.com>
2025-12-09 17:28:07 +09:00
Adam Gibson 753a1269b0 Array memory management: buffer lifecycle, deallocator fixes (#10349)
- Adds ConstantOffsetsBuffer and ConstantShapeBuffer headers
- Fixes InteropDataBuffer memory handling
- Improves DataBuffer implementation with proper cleanup
- Adds PrimaryPointerDeallocator for safe deallocation

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Claude <noreply@anthropic.com>
2025-12-09 17:27:27 +09:00
Adam Gibson 597cc598cd Shape cache helpers: trie structures, shape buffer creators (#10350)
- Adds ConstantShapeHelper and ConstantTadHelper headers
- Implements DirectShapeTrie and DirectTadTrie for fast lookups
- Adds CPU and CUDA ShapeBufferCreator implementations
- Implements StripedLocks for thread-safe cache access

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Claude <noreply@anthropic.com>
2025-12-09 17:27:01 +09:00
Adam Gibson d9ddbc8f32 NativeOps lifecycle tracking: split helpers, sanitizer support (#10351)
- Splits NativeOpExecutioner into broadcast, integer, reduce modules
- Adds NativeOpsHelpers for Arrays, Context, DataBuffers
- Implements LifecycleTracking helper for memory leak detection
- Adds Sanitizers helper for ASAN/MSAN support

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Claude <noreply@anthropic.com>
2025-12-09 17:26:32 +09:00
Adam Gibson d45252f71e Loops and ops: broadcasting, indexreduce, misc fixes (#10352)
- Adds broadcasting_bool and broadcasting_int loop implementations
- Fixes indexreduce loop for better memory handling
- Fixes various op implementations (loss, tensor, transforms, tsne)
- Adds QR decomposition helpers for CPU and CUDA

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Claude <noreply@anthropic.com>
2025-12-09 17:26:08 +09:00
Adam Gibson aad61ec839 ND4J memory management: deallocator service, opaque buffers, backend executioners (#10353)
- Fixes ArrayCacheMemoryMgr for proper cache cleanup
- Improves DeallocatorService for reliable resource cleanup
- Updates NativeOps interface and opaque buffer handling
- Fixes CPU and CUDA backend executioners

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Claude <noreply@anthropic.com>
2025-12-09 17:25:31 +09:00