Replace all occurences of pynvbench with cuda-bench

ci/build_cuda_bench_wheel_for_cuda.sh

@@ -78,7 +78,7 @@ python -m pip wheel --no-deps --verbose --wheel-dir dist .
 
 # Temporarily rename wheel to include CUDA version to avoid collision during multi-CUDA build
 # The merge script will combine these into a single wheel
-for wheel in dist/pynvbench-*.whl; do
+for wheel in dist/cuda_bench-*.whl; do
     if [[ -f "$wheel" ]]; then
         base_name=$(basename "$wheel" .whl)
         new_name="${base_name}.cu${cuda_version_major}.whl"
@@ -89,4 +89,4 @@ done
 
 # Move wheel to output directory
 mkdir -p /workspace/wheelhouse
-mv dist/pynvbench-*.cu*.whl /workspace/wheelhouse/
+mv dist/cuda_bench-*.cu*.whl /workspace/wheelhouse/

ci/build_multi_cuda_wheel.sh

@@ -43,7 +43,7 @@ if [[ -z "${HOST_WORKSPACE:-}" ]]; then
   echo "Setting HOST_WORKSPACE to: $HOST_WORKSPACE"
 fi
 
-# pynvbench must be built in a container that can produce manylinux wheels,
+# cuda-bench must be built in a container that can produce manylinux wheels,
 # and has the CUDA toolkit installed. We use the rapidsai/ci-wheel image for this.
 # We build separate wheels using separate containers for each CUDA version,
 # then merge them into a single wheel.
@@ -75,7 +75,7 @@ for ctk in 12 13; do
         --mount type=bind,source=${HOST_WORKSPACE},target=/workspace/ \
         --env py_version=${py_version} \
         $image \
-        /workspace/ci/build_pynvbench_wheel_for_cuda.sh
+        /workspace/ci/build_cuda_bench_wheel_for_cuda.sh
     # Prevent GHA runners from exhausting available storage with leftover images:
     if [[ -n "${GITHUB_ACTIONS:-}" ]]; then
       docker rmi -f $image
@@ -134,13 +134,13 @@ cd wheelhouse_merged
 
 # Unpack CUDA 12 wheel (this will be our base)
 $PYTHON -m wheel unpack "$cu12_wheel"
-base_dir=$(find . -maxdepth 1 -type d -name "pynvbench-*" | head -1)
+base_dir=$(find . -maxdepth 1 -type d -name "cuda_bench-*" | head -1)
 
 # Unpack CUDA 13 wheel into a temporary subdirectory
 mkdir cu13_tmp
 cd cu13_tmp
 $PYTHON -m wheel unpack "$cu13_wheel"
-cu13_dir=$(find . -maxdepth 1 -type d -name "pynvbench-*" | head -1)
+cu13_dir=$(find . -maxdepth 1 -type d -name "cuda_bench-*" | head -1)
 
 # Copy the cu13/ directory from CUDA 13 wheel into the base wheel
 cp -r "$cu13_dir"/cuda/bench/cu13 "../$base_dir/cuda/bench/"
@@ -159,7 +159,7 @@ cd ..
 
 # Install auditwheel and repair the merged wheel
 $PYTHON -m pip install --break-system-packages auditwheel
-for wheel in wheelhouse_merged/pynvbench-*.whl; do
+for wheel in wheelhouse_merged/cuda_bench-*.whl; do
     echo "Repairing merged wheel: $wheel"
     $PYTHON -m auditwheel repair \
         --exclude 'libcuda.so.1' \
@@ -177,12 +177,12 @@ rm -rf wheelhouse/*  # Clean existing wheelhouse
 mkdir -p wheelhouse
 
 # Move only the final repaired merged wheel
-if ls wheelhouse_final/pynvbench-*.whl 1> /dev/null 2>&1; then
-    mv wheelhouse_final/pynvbench-*.whl wheelhouse/
+if ls wheelhouse_final/cuda_bench-*.whl 1> /dev/null 2>&1; then
+    mv wheelhouse_final/cuda_bench-*.whl wheelhouse/
     echo "Final merged wheel moved to wheelhouse"
 else
     echo "No final repaired wheel found, moving unrepaired merged wheel"
-    mv wheelhouse_merged/pynvbench-*.whl wheelhouse/
+    mv wheelhouse_merged/cuda_bench-*.whl wheelhouse/
 fi
 
 # Clean up temporary directories

ci/test_cuda_bench.sh

@@ -74,7 +74,7 @@ echo "::group::🧪 Testing CUDA ${cuda_version} wheel on ${cuda_image}"
       --env cuda_version=${cuda_version} \
       --env cuda_extra="${cuda_extra}" \
       $cuda_image \
-      /workspace/ci/test_pynvbench_inner.sh
+      /workspace/ci/test_cuda_bench_inner.sh
   # Prevent GHA runners from exhausting available storage with leftover images:
   if [[ -n "${GITHUB_ACTIONS:-}" ]]; then
     docker rmi -f $cuda_image

ci/test_cuda_bench_inner.sh

@@ -1,7 +1,7 @@
 #!/bin/bash
 set -euo pipefail
 
-# Target script for `docker run` command in test_pynvbench.sh
+# Target script for `docker run` command in test_cuda_bench.sh
 # The /workspace pathnames are hard-wired here.
 
 # Install GCC 13 toolset (needed for builds that might happen during testing)
@@ -24,15 +24,15 @@ echo "CUDA version: $(nvcc --version | grep release)"
 # Wheel should be in /workspace/wheelhouse (downloaded by workflow or built locally)
 WHEELHOUSE_DIR="/workspace/wheelhouse"
 
-# Find the pynvbench wheel (multi-CUDA wheel)
+# Find the cuda-bench wheel (multi-CUDA wheel)
 # Prefer manylinux wheels, fall back to any wheel
-PYNVBENCH_WHEEL_PATH="$(ls ${WHEELHOUSE_DIR}/pynvbench-*manylinux*.whl 2>/dev/null | head -1)"
-if [[ -z "$PYNVBENCH_WHEEL_PATH" ]]; then
-    PYNVBENCH_WHEEL_PATH="$(ls ${WHEELHOUSE_DIR}/pynvbench-*.whl 2>/dev/null | head -1)"
+CUDA_BENCH_WHEEL_PATH="$(ls ${WHEELHOUSE_DIR}/cuda_bench-*manylinux*.whl 2>/dev/null | head -1)"
+if [[ -z "$CUDA_BENCH_WHEEL_PATH" ]]; then
+    CUDA_BENCH_WHEEL_PATH="$(ls ${WHEELHOUSE_DIR}/cuda_bench-*.whl 2>/dev/null | head -1)"
 fi
 
-if [[ -z "$PYNVBENCH_WHEEL_PATH" ]]; then
-    echo "Error: No pynvbench wheel found in ${WHEELHOUSE_DIR}"
+if [[ -z "$CUDA_BENCH_WHEEL_PATH" ]]; then
+    echo "Error: No cuda-bench wheel found in ${WHEELHOUSE_DIR}"
     echo "Contents of ${WHEELHOUSE_DIR}:"
     ls -la ${WHEELHOUSE_DIR}/ || true
     exit 1
@@ -42,9 +42,9 @@ fi
 CUDA_EXTRA="${cuda_extra:-cu${cuda_version}}"
 TEST_EXTRA="test-cu${cuda_version}"
 
-echo "Installing wheel: $PYNVBENCH_WHEEL_PATH with extras: ${TEST_EXTRA}"
-python -m pip install "${PYNVBENCH_WHEEL_PATH}[${TEST_EXTRA}]"
+echo "Installing wheel: $CUDA_BENCH_WHEEL_PATH with extras: ${TEST_EXTRA}"
+python -m pip install "${CUDA_BENCH_WHEEL_PATH}[${TEST_EXTRA}]"
 
 # Run tests
 cd "/workspace/python/test/"
-python -m pytest -v test_nvbench.py
+python -m pytest -v test_cuda_bench.py

python/cuda/bench/__init__.py

@@ -23,11 +23,11 @@ from cuda.pathfinder import (  # type: ignore[import-not-found]
 )
 
 try:
-    __version__ = importlib.metadata.version("pynvbench")
+    __version__ = importlib.metadata.version("cuda-bench")
 except Exception as e:
     __version__ = "0.0.0dev"
     warnings.warn(
-        "Could not retrieve version of pynvbench package dynamically from its metadata. "
+        "Could not retrieve version of cuda-bench package dynamically from its metadata. "
         f"Exception {e} was raised. "
         f"Version is set to fall-back value '{__version__}' instead."
     )
@@ -47,7 +47,7 @@ def _get_cuda_major_version():
     except ImportError:
         raise ImportError(
             "cuda-bindings is required for runtime CUDA version detection. "
-            "Install with: pip install pynvbench[cu12] or pip install pynvbench[cu13]"
+            "Install with: pip install cuda-bench[cu12] or pip install cuda-bench[cu13]"
         )
 
 
@@ -59,7 +59,7 @@ try:
     _nvbench_module = importlib.import_module(_module_fullname)
 except ImportError as e:
     raise ImportError(
-        f"No pynvbench extension found for CUDA {_cuda_major}.x. "
+        f"No cuda-bench extension found for CUDA {_cuda_major}.x. "
         f"This wheel may not include support for your CUDA version. "
         f"Supported CUDA versions: 12, 13. "
         f"Original error: {e}"

python/pyproject.toml

@@ -3,7 +3,7 @@ requires = ["scikit-build-core>=0.10", "setuptools_scm"]
 build-backend = "scikit_build_core.build"
 
 [project]
-name = "pynvbench"
+name = "cuda-bench"
 description = "CUDA Kernel Benchmarking Package"
 authors = [{ name = "NVIDIA Corporation" }]
 license = { text = "Apache-2.0 WITH LLVM-exception" }
@@ -36,10 +36,10 @@ cu12 = ["cuda-bindings>=12.0.0,<13.0.0", "nvidia-cuda-cupti-cu12"]
 cu13 = ["cuda-bindings>=13.0.0,<14.0.0", "nvidia-cuda-cupti>=13.0"]
 
 # Test dependencies for CUDA 12
-test-cu12 = ["pynvbench[cu12]", "pytest", "cupy-cuda12x", "numba"]
+test-cu12 = ["cuda-bench[cu12]", "pytest", "cupy-cuda12x", "numba"]
 
 # Test dependencies for CUDA 13
-test-cu13 = ["pynvbench[cu13]", "pytest", "cupy-cuda13x", "numba"]
+test-cu13 = ["cuda-bench[cu13]", "pytest", "cupy-cuda13x", "numba"]
 
 # Generic test dependencies (defaults to CUDA 12)
 test = ["pytest", "cupy-cuda12x", "numba"]