Repository navigation
Conversation
Register pointer-based CuTe DSL kernels as TensorRT AOT QDP plugins with dynamic shapes, constexpr parameters, multiple outputs, and shared memory. Validate compiler ABI and launch requirements before registering operators. Reuse signature, dtype, target, PTX, and pointer-launch helpers across the kernel frontends. Include a dynamic-shape example, API documentation, optional dependencies, CI coverage, and focused ABI and GPU regressions. Validation: 222 focused tests passed, including FP16/FP32 dynamic execution, shared memory, and standalone engines with CuTe imports blocked. CuTe and Triton examples also passed direct TensorRT GPU execution. All applicable pre-commit hooks and the CI manifest checks passed. Eleven existing integration cases were excluded from the focused run; the installed native runtime predates this checkout's Engine.from_engine_tensor API. GPU validation used the serialized-engine path.
github-actions
Bot
added
documentation
Recompile with a compiler-selected PTX ISA when CuTe's default output is newer than the loaded CUDA driver supports. Preserve architecture-specific targets such as sm_90a, retain ABI and driver-load validation, and derive runtime arguments from the validated signature layout. Replace the brittle CI setup source assertion with behavioral coverage for required and optional dependencies. Document compatibility requirements and add regression tests for target selection, compiler ISA retries, Hopper warpgroup instructions, and serialized dynamic-shape engines. Validation: 39 CuTe tests passed with both CUDA 12 and CUDA 13 compiler variants; 199 shared/frontend tests passed; 288 pin/updater tests passed (1 skipped). Pre-commit checks and the 32-suite CI manifest check passed.
This branch has not been deployed
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
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.
Register pointer-based CuTe DSL kernels as TensorRT AOT QDP plugins with dynamic shapes, constexpr parameters, multiple outputs, and shared memory. Validate compiler ABI and launch requirements before registering operators.
Reuse signature, dtype, target, PTX, and pointer-launch helpers across the kernel frontends. Include a dynamic-shape example, API documentation, optional dependencies, CI coverage, and focused ABI and GPU regressions.
Type of change
Checklist: