feat(runtime): TRT-RTX runtime cache, dynamic shapes strategy, and native CUDA graph support on the C++ runtime#2

Open

tp5uiuc wants to merge 9 commits into

reintroduce-rtx-featuresfrom

feat/trtrtx-cpp-runtime-v2

Commits on Jun 2, 2026

feat(runtime): add TensorRT-RTX runtime cache, dynamic shapes strategy, and native CUDA graph support to C++ runtime
tp5uiuc
committed
fix(runtime): route post-NCCL-release exec context through recreate_execution_context
tp5uiuc
committed
fix(runtime): validate strategy strings on every build, not only RTX
tp5uiuc
committed
fix(runtime): gate IRuntimeConfig usage for older TRT (Jetpack)
tp5uiuc
committed
test(runtime): smoke tests for TRT-RTX features on C++ runtime
tp5uiuc
committed
fix(runtime): add RTX-only entries to kSerializedInfoIndexNames
tp5uiuc
committed
refactor(runtime): address PR review comments on TRTEngine and TRTRuntimeConfig
tp5uiuc
committed
refactor(runtime): unify serialized engine layout across TRT and TRT-RTX
tp5uiuc
committed
test(runtime): parameterize TRT-RTX runtime tests over python/cpp
tp5uiuc
committed