Skip to content

DynamicShapePlan Runtime Snapshot: Native Execution, Java Runtime, and CUDA Bindings#10421

Open
agibsonccc wants to merge 1 commit intomasterfrom
pr/dsp-runtime-execution
Open

DynamicShapePlan Runtime Snapshot: Native Execution, Java Runtime, and CUDA Bindings#10421
agibsonccc wants to merge 1 commit intomasterfrom
pr/dsp-runtime-execution

Conversation

@agibsonccc
Copy link
Copy Markdown
Contributor

Summary

Delivers DynamicShapePlan runtime and execution-path changes across native + Java layers, including execution mode controls, runtime caching/allocation behavior, and CUDA/native binding updates used in DSP inference flows.

Source snapshot commit: b5893454f08c91e7bafb47478757c84a5f1cab04 from ag_new_release_updates_2.

Included Files

  • libnd4j/include/array/DataBuffer.h
  • libnd4j/include/array/NDArray.hXX
  • libnd4j/include/array/cpu/DataBuffer.cpp
  • libnd4j/include/array/cuda/DataBuffer.cu
  • libnd4j/include/execution/cuda/CudaGraphScheduler.h
  • libnd4j/include/graph/NativeDynamicShapePlan.h
  • libnd4j/include/graph/impl/NativeDynamicShapePlan.cpp
  • libnd4j/include/graph/impl/NativePlanCompiler.cpp
  • libnd4j/include/helpers/MmulHelper.h
  • libnd4j/include/helpers/cuda/MmulHelper.cu
  • libnd4j/include/helpers/cuda/PointersManager.cu
  • libnd4j/include/legacy/NativeOps.h
  • libnd4j/include/legacy/cpu/NativeOps_dsp.cpp
  • libnd4j/include/legacy/cuda/NativeOps_dsp.cu
  • libnd4j/include/legacy/impl/Environment.cpp
  • libnd4j/include/memory/cuda/CudaMemoryPool.cu
  • libnd4j/include/mlir/runtime/MLIREngine.cpp
  • libnd4j/include/mlir/runtime/MLIREngine.h
  • libnd4j/include/system/Environment.h
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/SameDiff.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/CollectiveCommunicator.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/CollectiveCommunicatorFactory.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/DspCompilationMode.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/DynamicShapePlanCompiler.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/DynamicShapePlanExecutor.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/GraphExecutionMode.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/LocalCollectiveCommunicator.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/NcclCommunicator.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/PipelineParallelRunner.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/TensorParallelConfig.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/execution/TensorParallelRunner.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/internal/InferenceSession.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/optimize/GraphOptimizer.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/optimize/optimizations/AttentionFusionOptimizations.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/optimize/optimizations/NormalizationFusionOptimizations.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/optimize/optimizations/QuantizationOptimizations.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/autodiff/samediff/peft/LoraAdapterCache.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/api/buffer/DataType.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/api/ndarray/LazyINDArray.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/factory/CompiledGraphFunction.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/factory/Environment.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/factory/GraphFunction.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/factory/GraphScope.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/factory/Nd4j.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/util/DeviceLocal.java
  • nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/nativeblas/NativeOps.java
  • nd4j/nd4j-backends/nd4j-backend-impls/nd4j-cuda/src/main/java/org/nd4j/linalg/jcublas/CudaEnvironment.java
  • nd4j/nd4j-backends/nd4j-backend-impls/nd4j-cuda/src/main/java/org/nd4j/linalg/jcublas/bindings/Nd4jCuda.java
  • nd4j/nd4j-common/src/main/java/org/nd4j/common/config/ND4JEnvironmentVars.java
  • nd4j/nd4j-common/src/main/java/org/nd4j/common/config/ND4JSystemProperties.java
  • nd4j/samediff-llm/src/main/java/org/eclipse/deeplearning4j/llm/generation/DecoderUtils.java
  • nd4j/samediff-llm/src/main/java/org/eclipse/deeplearning4j/llm/generation/DraftModelSpeculator.java
  • nd4j/samediff-llm/src/main/java/org/eclipse/deeplearning4j/llm/generation/KVCacheHostOffloader.java
  • nd4j/samediff-llm/src/main/java/org/eclipse/deeplearning4j/llm/generation/KVCachePrefixTree.java
  • nd4j/samediff-llm/src/main/java/org/eclipse/deeplearning4j/llm/generation/Speculator.java
  • nd4j/samediff-llm/src/main/java/org/eclipse/deeplearning4j/llm/generation/StaticKvCacheDecodeLoop.java
  • nd4j/samediff-llm/src/main/java/org/eclipse/deeplearning4j/llm/generation/TreeAttentionVerifier.java

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant