NVIDIA/TensorRT-Incubator mlir-tensorrt-v0.4.4.dev202512190

mlir-tensorrt-v0.4.4.dev202512190

Repository: NVIDIA/TensorRT-Incubator

Tag: mlir-tensorrt-v0.4.4.dev202512190

Published: 2025-12-20T03:35:47Z

Prerelease: yes

Release notes:

Major Features

• Open-source the kernel generation backend
• Open-source the PJRT (XLA/JAX) integration
• Various optimizations targeting StableHLO while loops and raising to for-style loops
• Improve C++ host code generation pipeline; fill in some missing EmitC converters

Generated Release Notes

• Update CI workflows and devcontainer configs for CUDA 12.9 and Ubuntu 24.04 by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/723
• migrate internal changes by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/722
• migrate internal changes by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/724
• [CI] Adding Test Pipeline for PRs by @lanluo-nvidia in https://github.com/NVIDIA/TensorRT-Incubator/pull/725
• [mlir-tensorrt] Migrate internal change by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/726
• Add conversion of JAX shape_assertion custom calls to cf.assert by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/728
• Add Math dialect to EmitC conversion support by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/729
• Refactor: Move StableHLO matchers to dialect utils and reorganize namespaces by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/730
• Fix zero-rank tensor handling in shape functions by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/731
• [compiler] Fix missing test file by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/732
• [compiler] Add stablehlo.get_dimension_size conversion in stablehlo-to-linalg by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/733
• [compiler] Add cf.assert support to convert-host-to-emitc by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/734
• NFC: make minor updates to pass naming by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/735
• [tensorrt] Make new transpose/reshape elimination pass as v2 pass. by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/736
• [compiler] Add mtrt-scf-strength-reduction pass by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/737
• integrate internal changes by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/738
• integrate internal changes by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/739
• [mlir-tensorrt] Integrate Kernel backend by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/740
• [mlir-tensorrt] Migrate PJRT integration by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/741
• [CMake] Fix TRT CUDA version selection and add tests by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/745
• [CI] Add release wheel pipeline by @lanluo-nvidia in https://github.com/NVIDIA/TensorRT-Incubator/pull/742
• [mlir-tensorrt] NFC: Integrate internal changes by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/746
• [compiler] Add mtrt-scf-float-strength-reduce pass to preprocessing pipeline by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/747
• NFC: Fix formatting in MLIR-TRT README.md by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/749
• [integrations/PJRT] Implement executable metadata queries by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/748
• [integrations/PJRT] Implement additional PJRT methods for executable serialization/deserialization by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/750
• NFC: Update MLIR-TRT build instructions for PJRT integration wheels by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/751
• Upgrade GitHub Actions for Node 24 compatibility by @salmanmkc in https://github.com/NVIDIA/TensorRT-Incubator/pull/743
• integrate internal changes by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/752
• [mlir-tensorrt] Update LLVM patch to add support for PTX ISA v9.0 by @christopherbate in https://github.com/NVIDIA/TensorRT-Incubator/pull/753

Full Changelog: https://github.com/NVIDIA/TensorRT-Incubator/compare/mlir-tensorrt-v0.4.2dev20251112...mlir-tensorrt-v0.4.4.dev202512190