Merge branch 'release-v24.03' into 'main'

Release v24.03 See merge request cuda-hpc-libraries/cuquantum-sdk/cuquantum-public!26
NVIDIA · Mar 26, 2024 · 788c048 · 788c048
2 parents 2ac5645 + eacdc0e
commit 788c048
Show file tree

Hide file tree

Showing 104 changed files with 14,133 additions and 8,537 deletions.
diff --git a/LICENSE b/LICENSE
@@ -1,4 +1,4 @@
-Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
 BSD-3-Clause
 

diff --git a/benchmarks/LICENSE b/benchmarks/LICENSE
@@ -1,4 +1,4 @@
-Copyright (c) 2021-2023 NVIDIA CORPORATION & AFFILIATES.
+Copyright (c) 2021-2024 NVIDIA CORPORATION & AFFILIATES.
 
 BSD-3-Clause
 

diff --git a/benchmarks/cuquantum_benchmarks/__init__.py b/benchmarks/cuquantum_benchmarks/__init__.py
@@ -2,4 +2,4 @@
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
-__version__ = '0.3.1'
+__version__ = '0.3.2'
diff --git a/benchmarks/cuquantum_benchmarks/backends/backend_qiskit.py b/benchmarks/cuquantum_benchmarks/backends/backend_qiskit.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES
+# Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
@@ -43,7 +43,11 @@ def __init__(self, ngpus, ncpu_threads, precision, *args, identifier=None, **kwa
     def find_version(self, identifier):
         if identifier == 'cusvaer':
             return version('cusvaer')
-        return qiskit.__qiskit_version__['qiskit-aer']
+
+        if hasattr(qiskit_aer, "__version__"):
+            return qiskit_aer.__version__
+        else:
+            return qiskit.__qiskit_version__['qiskit-aer']
 
     def preprocess_circuit(self, circuit, *args, **kwargs):
         if _internal_utils is not None:
@@ -77,7 +81,10 @@ def create_aer_backend(self, identifier, ngpus, ncpu_threads, *args, **kwargs):
         try:
             # we defer importing Aer as late as possible, due to a bug it has that
             # could init all GPUs prematurely
-            from qiskit.providers.aer import AerSimulator
+            if hasattr(qiskit, "__version__") and qiskit.__version__ >= "1.0.0":
+                from qiskit_aer import AerSimulator
+            else:
+                from qiskit.providers.aer import AerSimulator
         except ImportError as e:
             raise RuntimeError("qiskit-aer (or qiskit-aer-gpu) is not installed") from e
 

diff --git a/benchmarks/cuquantum_benchmarks/frontends/frontend_qiskit.py b/benchmarks/cuquantum_benchmarks/frontends/frontend_qiskit.py
@@ -1,12 +1,15 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES
+# Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
 from math import pi
 
 try:
     import qiskit
-    from qiskit.extensions import UnitaryGate
+    if hasattr(qiskit, "__version__") and qiskit.__version__ >= "1.0.0":
+        from qiskit.circuit.library import UnitaryGate
+    else:
+        from qiskit.extensions import UnitaryGate
 except ImportError:
     qiskit = UnitaryGate = None
 
@@ -35,7 +38,7 @@ def generateCircuit(self, gateSeq):
                 circuit.x(g.targets)
 
             elif g.id == 'cnot':
-                circuit.cnot(g.controls, g.targets)
+                circuit.cx(g.controls, g.targets)
 
             elif g.id == 'cz':
                 circuit.cz(g.controls, g.targets)

diff --git a/benchmarks/cuquantum_benchmarks/run_interface.py b/benchmarks/cuquantum_benchmarks/run_interface.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES
+# Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
@@ -206,7 +206,10 @@ def _fix_filename_for_cutn(self, circuit_filename, nqubits):
     def extract_frontend_version(self):
         if self.frontend == 'qiskit':
             import qiskit
-            version = qiskit.__qiskit_version__['qiskit-terra']
+            if hasattr(qiskit, "__version__") and qiskit.__version__ >= "1.0.0":
+                version = qiskit.__version__
+            else:
+                version = qiskit.__qiskit_version__['qiskit-terra']
         elif self.frontend == 'cirq':
             import cirq
             version = cirq.__version__

diff --git a/python/LICENSE b/python/LICENSE
@@ -1,4 +1,4 @@
-Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
 BSD-3-Clause
 

diff --git a/python/README.md b/python/README.md
@@ -4,47 +4,18 @@
 
 Please visit the [NVIDIA cuQuantum Python documentation](https://docs.nvidia.com/cuda/cuquantum/latest/python).
 
+For instructions on installing *cuQuantum Python*, refer to our 
+[getting started section](../getting_started/getting_started.rst)
 
-## Installation
+## Building and installing cuQuantum Python from source
 
-### Install cuQuantum Python from conda-forge
-
-If you already have a Conda environment set up, it is the easiest to install cuQuantum Python from the conda-forge channel:
-```
-conda install -c conda-forge cuquantum-python
-```
-The Conda solver will install all required dependencies for you. If you need to select a particular CUDA version, say CUDA 12.0, please issue the following command:
-```
-conda install -c conda-forge cuquantum-python cuda-version=12.0
-```
-
-### Install cuQuantum Python from PyPI
-
-Alternatively, assuming you already have a Python environment set up (it doesn't matter if it's a Conda env or not),
-you can also install cuQuantum Python this way:
-
-```
-pip install cuquantum-python-cuXX
-```
-with `XX` being `11` (for CUDA 11) or `12` (for CUDA 12).
-The `pip` solver will also install all required dependencies for you (including both cuTENSOR and cuQuantum wheels).
-
-Notes:
-
-- Users can install cuQuantum Python using `pip install --no-cache-dir cuquantum-python`, which will attempt to detect the current CUDA environment and choose the appropriate wheel to install. In the event of detection failure, CUDA 11 is assumed. This is subject to change in the future. Installing wheels with the `-cuXX` suffix is encouraged. `--no-cache-dir` is required when using `pip` 23.1+.
-- CuPy also uses a similar auto-detection mechanism to determine the correct wheel to install. If in doubt, or if installing `cuquantum-python-cu11`, please follow [CuPy's installation guide](https://docs.cupy.dev/en/stable/install.html) and install it manually.
-- To manually manage all Python dependencies, append `--no-deps` to `pip install` to bypass the `pip` solver, see below.
-
-### Building and installing cuQuantum Python from source
-
-#### Requirements
+### Requirements
 
 The build-time dependencies of the cuQuantum Python package include:
 
 * CUDA Toolkit 11.x or 12.x
 * cuStateVec 1.4.0+
-* cuTensorNet 2.2.0+
-* cuTENSOR 1.6.1+
+* cuTensorNet 2.4.0+
 * Python 3.9+
 * Cython >=0.29.22,<3
 * pip 21.3.1+
@@ -86,11 +57,10 @@ Runtime dependencies of the cuQuantum Python package include:
 * Driver: Linux (450.80.02+ for CUDA 11, 525.60.13+ for CUDA 12)
 * CUDA Toolkit 11.x or 12.x
 * cuStateVec 1.4.0+
-* cuTensorNet 2.2.0+
-* cuTENSOR 1.6.1+
+* cuTensorNet 2.4.0+
 * Python 3.9+
 * NumPy v1.21+
-* CuPy v10.0.0+ (see [installation guide](https://docs.cupy.dev/en/stable/install.html))
+* CuPy v13.0.0+ (see [installation guide](https://docs.cupy.dev/en/stable/install.html))
 * PyTorch v1.10+ (optional, see [installation guide](https://pytorch.org/get-started/locally/))
 * Qiskit v0.24.0+ (optional, see [installation guide](https://qiskit.org/documentation/getting_started.html))
 * Cirq v0.6.0+ (optional, see [installation guide](https://quantumai.google/cirq/install))

diff --git a/python/builder/pep517.py b/python/builder/pep517.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES
+# Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
@@ -30,8 +30,8 @@ def get_requires_for_build_wheel(config_settings=None):
     # set up version constraints: note that CalVer like 22.03 is normalized to
     # 22.3 by setuptools, so we must follow the same practice in the constraints;
     # also, we don't need the patch number here
-    cuqnt_require = [f'custatevec-cu{utils.cuda_major_ver}~=1.5',   # ">=1.5.0,<2"
-                     f'cutensornet-cu{utils.cuda_major_ver}~=2.3',  # ">=2.3.0,<3"
+    cuqnt_require = [f'custatevec-cu{utils.cuda_major_ver}~=1.6',   # ">=1.6.0,<2"
+                     f'cutensornet-cu{utils.cuda_major_ver}~=2.4',  # ">=2.4.0,<3"
                     ]
 
     return _build_meta.get_requires_for_build_wheel(config_settings) + cuqnt_require

diff --git a/python/builder/utils.py b/python/builder/utils.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES
+# Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
@@ -79,7 +79,7 @@ def run(self):
 class build_ext(_build_ext):
 
     def _set_library_roots(self):
-        custatevec_root = cutensornet_root = cutensor_root = None
+        custatevec_root = cutensornet_root = cuquantum_root = None
         # Note that we need sys.path because of build isolation (since PEP 517)
         py_paths = sys.path + [site.getusersitepackages()] + site.getsitepackages()
 
@@ -102,83 +102,71 @@ def _set_library_roots(self):
                 if cuquantum_root is None:
                     raise RuntimeError('cuStateVec is not found, please set $CUQUANTUM_ROOT '
                                        'or $CUSTATEVEC_ROOT') from e
-                else:
-                    custatevec_root = cuquantum_root
             try:
                 cutensornet_root = os.environ['CUTENSORNET_ROOT']
             except KeyError as e:
                 if cuquantum_root is None:
                     raise RuntimeError('cuTensorNet is not found, please set $CUQUANTUM_ROOT '
                                        'or $CUTENSORNET_ROOT') from e
-                else:
-                    cutensornet_root = cuquantum_root
 
-        return custatevec_root, cutensornet_root
+        return custatevec_root, cutensornet_root, cuquantum_root
 
     def _prep_includes_libs_rpaths(self):
         """
-        Set global vars cusv_incl_dir, cutn_incl_dir, cusv_lib_dir, cutn_lib_dir,
-        cusv_lib, cutn_lib, and extra_linker_flags.
+        Set global vars cusv_incl_dir, cutn_incl_dir, and extra_linker_flags.
+
+        With the new bindings, we no longer need to link to cuQuantum DSOs.
         """
-        custatevec_root, cutensornet_root = self._set_library_roots()
-
-        global cusv_incl_dir, cutn_incl_dir
-        cusv_incl_dir = [os.path.join(cuda_path, 'include'),
-                         os.path.join(custatevec_root, 'include')]
-        cutn_incl_dir = [os.path.join(cuda_path, 'include'),
-                         os.path.join(cutensornet_root, 'include')]
-
-        global cusv_lib_dir, cutn_lib_dir
-        # we include both lib64 and lib to accommodate all possible sources
-        cusv_lib_dir = [os.path.join(custatevec_root, 'lib'),
-                        os.path.join(custatevec_root, 'lib64')]
-        cutn_lib_dir = [os.path.join(cutensornet_root, 'lib'),
-                        os.path.join(cutensornet_root, 'lib64')]
-
-        global cusv_lib, cutn_lib, extra_linker_flags
+        custatevec_root, cutensornet_root, cuquantum_root = self._set_library_roots()
+
+        global cusv_incl_dir, cutn_incl_dir, cuqnt_incl_dir
+        cusv_incl_dir = cutn_incl_dir = cuqnt_incl_dir = None
+        base_incl_dir = (os.path.join(cuda_path, 'include'),)
+        if cuquantum_root is not None:
+            cuqnt_incl_dir = base_incl_dir + (os.path.join(cuquantum_root, 'include'),)
+        if custatevec_root is not None:
+            cusv_incl_dir = base_incl_dir + (os.path.join(custatevec_root, 'include'),)
+        if cutensornet_root is not None:
+            cutn_incl_dir = base_incl_dir + (os.path.join(cutensornet_root, 'include'),)
+
+        global extra_linker_flags
         if not building_wheel:
             # Note: with PEP-517 the editable mode would not build a wheel for installation
             # (and we purposely do not support PEP-660).
-            cusv_lib = ['custatevec']
-            cutn_lib = ['cutensornet']
             extra_linker_flags = []
         else:
             # Note: soname = library major version
             # We don't need to link to cuBLAS/cuSOLVER/cuTensor at build time
-            cusv_lib = [':libcustatevec.so.1']
-            cutn_lib = [':libcutensornet.so.2']
             # The rpaths must be adjusted given the following full-wheel installation:
-            # - cuquantum-python: site-packages/cuquantum/{custatevec, cutensornet}/  [=$ORIGIN]
+            # - cuquantum-python: site-packages/cuquantum/{custatevec, cutensornet}/_internal/  [=$ORIGIN]
             # - cusv & cutn:      site-packages/cuquantum/lib/
             # - cutensor:         site-packages/cutensor/lib/
             # - cublas:           site-packages/nvidia/cublas/lib/
             # - cusolver:         site-packages/nvidia/cusolver/lib/
             # (Note that starting v22.11 we use the new wheel format, so all lib wheels have suffix -cuXX,
             #  and cuBLAS/cuSOLVER additionally have prefix nvidia-.)
             ldflag = "-Wl,--disable-new-dtags,"
-            ldflag += "-rpath,$ORIGIN/../lib,"
-            ldflag += "-rpath,$ORIGIN/../../cutensor/lib,"
-            ldflag += "-rpath,$ORIGIN/../../nvidia/cublas/lib,"
-            ldflag += "-rpath,$ORIGIN/../../nvidia/cusolver/lib"
+            ldflag += "-rpath,$ORIGIN/../../lib,"
+            ldflag += "-rpath,$ORIGIN/../../../cutensor/lib,"
+            ldflag += "-rpath,$ORIGIN/../../../nvidia/cublas/lib,"
+            ldflag += "-rpath,$ORIGIN/../../../nvidia/cusolver/lib"
             extra_linker_flags = [ldflag]
 
         print("\n"+"*"*80)
         print("CUDA version:", cuda_ver)
         print("CUDA path:", cuda_path)
-        print("cuStateVec path:", custatevec_root)
-        print("cuTensorNet path:", cutensornet_root)
+        print("cuStateVec path:", custatevec_root if custatevec_root else cuquantum_root)
+        print("cuTensorNet path:", cutensornet_root if cutensornet_root else cuquantum_root)
         print("*"*80+"\n")
 
     def build_extension(self, ext):
+        ext.include_dirs = ()
+        for include_dir in (cusv_incl_dir, cutn_incl_dir, cuqnt_incl_dir):
+            if include_dir is not None:
+                ext.include_dirs += include_dir
         if ext.name.endswith("custatevec"):
-            ext.include_dirs = cusv_incl_dir
-            ext.library_dirs = cusv_lib_dir
-            ext.libraries = cusv_lib
             ext.extra_link_args = extra_linker_flags
         elif ext.name.endswith("cutensornet"):
-            ext.include_dirs = cutn_incl_dir
-            ext.library_dirs = cutn_lib_dir
-            ext.libraries = cutn_lib
             ext.extra_link_args = extra_linker_flags
 
         super().build_extension(ext)

diff --git a/python/cuquantum/__init__.py b/python/cuquantum/__init__.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES
+# Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
@@ -7,7 +7,7 @@
 from cuquantum.cutensornet import (
     contract, contract_path, einsum, einsum_path, tensor, tensor_qualifiers_dtype, Network, BaseCUDAMemoryManager, MemoryPointer,
     NetworkOptions, OptimizerInfo, OptimizerOptions, PathFinderOptions, ReconfigOptions, SlicerOptions, CircuitToEinsum)
-from cuquantum.utils import ComputeType, cudaDataType, libraryPropertyType
+from cuquantum._utils import ComputeType, cudaDataType, libraryPropertyType
 from cuquantum._version import __version__
 
 
@@ -43,4 +43,4 @@
         ):
     cutensornet._internal.enum_utils.add_enum_class_doc(enum, chomp="_ATTRIBUTE|_PREFERENCE_ATTRIBUTE")
 
-del enum, utils
+del enum
diff --git a/python/cuquantum/__main__.py b/python/cuquantum/__main__.py
@@ -1,4 +1,4 @@
-# Copyright (c) 2021-2023, NVIDIA CORPORATION & AFFILIATES
+# Copyright (c) 2021-2024, NVIDIA CORPORATION & AFFILIATES
 #
 # SPDX-License-Identifier: BSD-3-Clause
 
@@ -7,14 +7,21 @@
 import site
 import sys
 
-import cuquantum  # get the shared libraries loaded
-
 
 def get_lib_path(name):
     """Get the loaded shared library path."""
     # Ideally we should call dl_iterate_phdr or dladdr to do the job, but this
     # is simpler and not bad; the former two are not strictly portable anyway
     # (not part of POSIX). Obviously this only works on Linux!
+
+    # We have switched to use dlopen, force library loading via internal API
+    if "custatevec" in name:
+        from cuquantum import custatevec as cusv
+        cusv._internal.custatevec._inspect_function_pointers()
+    elif "cutensor" in name:  # cutensor or cutensornet
+        from cuquantum import cutensornet as cutn
+        cutn._internal.cutensornet._inspect_function_pointers()
+
     try:
         with open('/proc/self/maps') as f:
             lib_map = f.read()