[TRTLLM-14026][feat] BREAKING: Remove TensorRT from the C++ tree

.github/CODEOWNERS

@@ -37,8 +37,6 @@
 /tests/unittest @NVIDIA/trt-llm-devs
 
 # ===== TensorRT backend (will be deprecated soon) — also on the trt-llm-devs fallback =====
-/cpp/include/tensorrt_llm/plugins @NVIDIA/trt-llm-devs
-/cpp/tensorrt_llm/plugins @NVIDIA/trt-llm-devs
 /tensorrt_llm/builder.py @NVIDIA/trt-llm-devs
 /tensorrt_llm/commands/build.py @NVIDIA/trt-llm-devs
 /tensorrt_llm/commands/prune.py @NVIDIA/trt-llm-devs
@@ -105,7 +103,6 @@
 /cpp/tensorrt_llm/batch_manager @NVIDIA/trt-llm-runtime-devs
 /cpp/tensorrt_llm/common @NVIDIA/trt-llm-runtime-devs
 /cpp/tensorrt_llm/executor @NVIDIA/trt-llm-runtime-devs
-/cpp/tensorrt_llm/executor_worker @NVIDIA/trt-llm-runtime-devs
 /cpp/tensorrt_llm/layers @NVIDIA/trt-llm-runtime-devs
 /cpp/tensorrt_llm/nanobind @NVIDIA/trt-llm-runtime-devs
 /cpp/tensorrt_llm/runtime @NVIDIA/trt-llm-runtime-devs
@@ -169,7 +166,6 @@
 /examples/llm-api/quickstart_multimodal.py @NVIDIA/trt-llm-models-devs @NVIDIA/trt-llm-doc-owners
 /examples/models @NVIDIA/trt-llm-models-devs @NVIDIA/trt-llm-doc-owners
 /examples/serve/*multimodal* @NVIDIA/trt-llm-models-devs @NVIDIA/trt-llm-doc-owners
-/scripts/build_cpp_examples.py @NVIDIA/trt-llm-models-devs
 /scripts/generate_config_database_tests.py @NVIDIA/trt-llm-models-devs @NVIDIA/trt-llm-doc-owners
 /scripts/generate_config_table.py @NVIDIA/trt-llm-models-devs @NVIDIA/trt-llm-doc-owners
 /tensorrt_llm/_torch/models @NVIDIA/trt-llm-models-devs

.pre-commit-config.yaml

@@ -9,14 +9,14 @@ common-files: &common_files |
         .devcontainer/make_env.py |
         .github/scripts/label_community_user.py |
         .github/scripts/pr_checklist_check.py |
-        benchmarks/cpp/__init__.py |
-        benchmarks/cpp/prepare_dataset.py |
-        benchmarks/cpp/utils/__init__.py |
-        benchmarks/cpp/utils/convert_nemo_dataset.py |
-        benchmarks/cpp/utils/generate_rand_loras.py |
-        benchmarks/cpp/utils/prepare_real_data.py |
-        benchmarks/cpp/utils/prepare_synthetic_data.py |
-        benchmarks/cpp/utils/utils.py |
+        benchmarks/__init__.py |
+        benchmarks/prepare_dataset.py |
+        benchmarks/utils/__init__.py |
+        benchmarks/utils/convert_nemo_dataset.py |
+        benchmarks/utils/generate_rand_loras.py |
+        benchmarks/utils/prepare_real_data.py |
+        benchmarks/utils/prepare_synthetic_data.py |
+        benchmarks/utils/utils.py |
         cpp/conanfile.py |
         cpp/kernels/fmha_v2/conftest.py |
         cpp/kernels/fmha_v2/fmha_test.py |
@@ -1350,14 +1350,14 @@ legacy-files: &legacy_files |
         .devcontainer/make_env.py |
         .github/scripts/label_community_user.py |
         .github/scripts/pr_checklist_check.py |
-        benchmarks/cpp/__init__.py |
-        benchmarks/cpp/prepare_dataset.py |
-        benchmarks/cpp/utils/__init__.py |
-        benchmarks/cpp/utils/convert_nemo_dataset.py |
-        benchmarks/cpp/utils/generate_rand_loras.py |
-        benchmarks/cpp/utils/prepare_real_data.py |
-        benchmarks/cpp/utils/prepare_synthetic_data.py |
-        benchmarks/cpp/utils/utils.py |
+        benchmarks/__init__.py |
+        benchmarks/prepare_dataset.py |
+        benchmarks/utils/__init__.py |
+        benchmarks/utils/convert_nemo_dataset.py |
+        benchmarks/utils/generate_rand_loras.py |
+        benchmarks/utils/prepare_real_data.py |
+        benchmarks/utils/prepare_synthetic_data.py |
+        benchmarks/utils/utils.py |
         cpp/conanfile.py |
         cpp/kernels/fmha_v2/conftest.py |
         cpp/kernels/fmha_v2/fmha_test.py |

benchmarks/README.md

@@ -1,10 +1,36 @@
-# TensorRT-LLM Benchmarks
+# TensorRT-LLM Benchmarking
 
-## Overview
+For benchmarking TensorRT-LLM, use
+[`trtllm-bench`](../docs/source/developer-guide/perf-benchmarking.md) and
+`trtllm-serve`. The legacy C++ runtime benchmarks (`gptManagerBenchmark`,
+`bertBenchmark`, `disaggServerBenchmark`) were removed together with the
+TensorRT backend.
 
-There are currently two workflows to benchmark TensorRT-LLM:
-* [`trtllm-bench`](../docs/source/developer-guide/perf-benchmarking.md)
-  - `trtllm-bench` is native to TensorRT-LLM and is a Python benchmarker for reproducing and testing the performance of TensorRT-LLM.
-  - _NOTE_: This benchmarking suite is a current work in progress and is prone to large changes.
-* [C++ benchmarks](./cpp)
-  - The recommended workflow that uses TensorRT-LLM C++ API and can take advantage of the latest features of TensorRT-LLM.
+This directory keeps the dataset preparation tools consumed by `trtllm-bench`:
+
+- `prepare_dataset.py` — generate benchmark datasets from real data or with
+  synthetic normal/uniform token-length distributions:
+
+  ```bash
+  python3 prepare_dataset.py \
+      --tokenizer <path/to/tokenizer> \
+      --output preprocessed_dataset.json \
+      dataset \
+      --dataset-name <name of the dataset> \
+      --dataset-split <split of the dataset to use> \
+      --dataset-input-key <dataset dictionary key for input> \
+      --dataset-prompt-key <dataset dictionary key for prompt> \
+      --dataset-output-key <dataset dictionary key for output> \
+      [--num-requests 100] \
+      [--max-input-len 1000] \
+      [--output-len-dist 100,10]
+  ```
+
+  Synthetic variants: `python3 prepare_dataset.py ... token-norm-dist ...` and
+  `... token-unif-dist ...`. Run with `--help` for the full option list.
+
+- `utils/prepare_real_data.py`, `utils/prepare_synthetic_data.py` — the
+  subcommand implementations.
+- `utils/generate_rand_loras.py` — generate random LoRA adapters for
+  LoRA benchmarking.
+- `utils/convert_nemo_dataset.py` — convert NeMo chat datasets.