Document CUDA library dependencies for python wheel

docs/sphinx/api/languages/python_api.rst

@@ -38,10 +38,10 @@ Program Construction
     .. automethod:: my
     .. automethod:: mz
     .. automethod:: c_if
+    .. automethod:: for_loop
     .. automethod:: adjoint
     .. automethod:: control
     .. automethod:: apply_call
-    .. automethod:: for_loop
 
 Kernel Execution
 =============================

docs/sphinx/install.rst

@@ -118,42 +118,45 @@ or run the Python examples using the Python interpreter.
 Python wheels
 --------------------
 
-CUDA Quantum Python wheels are available on [PyPI.org](https://pypi.org/project/cuda-quantum). 
-The CUDA Quantum Python wheels contain the Python API and core components of
-CUDA Quantum. For more information about available packages and documentation,
+CUDA Quantum Python wheels are available on `PyPI.org <https://pypi.org/project/cuda-quantum>`__. Installation instructions can be found in the `project description <https://pypi.org/project/cuda-quantum/#description>`__.
+For more information about available versions and documentation,
 see :doc:`versions`.
 
-To install the latest release using `pip <https://pypi.org/project/pip/>`__, run
+At this time, wheels are distributed for Linux operating systems only. 
 
-.. code-block:: console
+There are currently no source distributions available on PyPI, but you can download the source code for the latest version of the CUDA Quantum Python wheels from our `GitHub repository <https://github.com/NVIDIA/cuda-quantum>`__. The source code for previous versions can be downloaded from the respective `GitHub Release <https://github.com/NVIDIA/cuda-quantum/releases>`__.
 
-    python3 -m pip install cuda-quantum
+To build the CUDA Quantum Python API from source using pip, run the following commands:
 
-There are currently no source distributions available on PyPI, but you can download the source code 
-for all releases `here <https://github.com/NVIDIA/cuda-quantum/releases>`__. 
-For more information about building a CUDA Quantum Python wheel from source, see the 
-`README <https://github.com/NVIDIA/cuda-quantum/blob/main/python/README.md>`__.
+```console
+git clone https://github.com/NVIDIA/cuda-quantum.git
+cd cuda-quantum && ./scripts/install_prerequisites.sh
+pip install .
+```
 
+For more information about building the entire C++ and Python API's, please see `Building from Source`_.
 
-Build CUDA Quantum from Source
+Building from Source
 ------------------------------
 
-For more information about building CUDA Quantum from source, 
-we refer to the `CUDA Quantum GitHub repository`_.
+Instructions for building the Python wheels from source are given in the section :ref:`install-python-wheels`.
+For more information about building the entire C++ and Python API from source, we refer to the `CUDA Quantum GitHub repository`_.
 
 .. _CUDA Quantum GitHub repository: https://github.com/NVIDIA/cuda-quantum/blob/main/Building.md
 
 
-CUDA Quantum Dependencies
--------------------------
+.. _dependencies-and-compatibility:
+
+Dependencies and Compatibility
+--------------------------------
 
-CUDA Quantum can be used to simulate quantum programs (see :doc:`using/simulators`) on a CPU-only system, but a GPU is highly recommended.
+CUDA Quantum can be used to compile and run quantum programs on a CPU-only system, but a GPU is highly recommended and necessary to use the GPU-based simulators, see also :doc:`using/simulators`.
 
 The supported CPUs include x86_64 (x86-64-v3 architecture and newer) and ARM64 architectures.
 
 .. note:: 
 
-   The CUDA Quantum Python wheels depend on an existing CUDA installation on your system. For more information about installing the CUDA Quantum Python wheels, take a look at :ref:`this page <install-python-wheels>`.
+   Some of the components included in the CUDA Quantum Python wheels depend on an existing CUDA installation on your system. For more information about installing the CUDA Quantum Python wheels, take a look at :ref:`this section <install-python-wheels>`.
 
 The following table summarizes the required components.
 
@@ -167,6 +170,8 @@ The following table summarizes the required components.
       - Linux
     * - Tested Distributions
       - CentOS 8; Debian 11, 12; Fedora 38; OpenSUSE/SELD/SLES 15.5; RHEL 8, 9; Rocky 8, 9; Ubuntu 22.04
+    * - Python versions
+      - 3.8+
 
 .. list-table:: Requirements for GPU Simulation
     :widths: 30 50

docs/sphinx/releases.rst

@@ -8,6 +8,7 @@ The latest version of CUDA Quantum is on the main branch of our `GitHub reposito
 
 - `Docker image (nightly builds) <https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nightly/containers/cuda-quantum>`__
 - `Documentation <https://nvidia.github.io/cuda-quantum/latest>`__
+- `Examples <https://github.com/NVIDIA/cuda-quantum/tree/main/docs/sphinx/examples>`__
 
 **0.4.1**
 
@@ -16,6 +17,7 @@ The 0.4.1 release adds support for ARM processors in the form of multi-platform
 - `Docker image <https://catalog.ngc.nvidia.com/orgs/nvidia/containers/cuda-quantum>`__
 - `Python wheel <https://pypi.org/project/cuda-quantum/>`__
 - `Documentation <https://nvidia.github.io/cuda-quantum/0.4.1>`__
+- `Examples <https://github.com/NVIDIA/cuda-quantum/tree/releases/v0.4.1/docs/sphinx/examples>`__
 
 The full change log can be found `here <https://github.com/NVIDIA/cuda-quantum/releases>`__.
 
@@ -30,6 +32,7 @@ The fully featured version is available as a Docker image for `linux/amd64` plat
 - `Docker image <https://catalog.ngc.nvidia.com/orgs/nvidia/containers/cuda-quantum/tags>`__
 - `Python wheel <https://pypi.org/project/cuda-quantum/0.4.0>`__
 - `Documentation <https://nvidia.github.io/cuda-quantum/0.4.0>`__
+- `Examples <https://github.com/NVIDIA/cuda-quantum/tree/0.4.0/docs/sphinx/examples>`__
 
 The full change log can be found `here <https://github.com/NVIDIA/cuda-quantum/releases/tag/0.4.0>`__.
 
@@ -39,3 +42,4 @@ The 0.3.0 release of CUDA Quantum is available as a Docker image for `linux/amd6
 
 - `Docker image <https://catalog.ngc.nvidia.com/orgs/nvidia/containers/cuda-quantum/tags>`__
 - `Documentation <https://nvidia.github.io/cuda-quantum/0.3.0>`__
+- `Examples <https://github.com/NVIDIA/cuda-quantum/tree/0.3.0/docs/sphinx/examples>`__

docs/sphinx/using/simulators.rst

@@ -28,6 +28,10 @@ In python, this can be specified with
 By default, this will leverage :code:`FP32` floating point types for the simulation. To 
 switch to :code:`FP64`, specify the :code:`nvidia-fp64` target instead. 
 
+.. note:: 
+
+    This backend requires an NVIDIA GPU and CUDA runtime libraries. See the section :ref:`dependencies-and-compatibility` for more information.
+
 cuQuantum multi-node multi-GPU
 ++++++++++++++++++++++++++++++++++
 
@@ -48,6 +52,10 @@ In python, this can be specified with
 
     cudaq.set_target('nvidia-mgpu')
 
+.. note:: 
+
+    This backend requires an NVIDIA GPU, CUDA runtime libraries, as well as an MPI installation. See the section :ref:`dependencies-and-compatibility` for more information.
+
 OpenMP CPU-only
 ++++++++++++++++++++++++++++++++++
 
@@ -66,6 +74,10 @@ The :code:`tensornet` target provides a tensor-network simulator accelerated wit
 the :code:`cuTensorNet` library. This backend is currently available for use from C++ and supports 
 Multi-Node, Multi-GPU distribution of tensor operations required to evaluate and simulate the circuit.
 
+.. note:: 
+
+    This backend requires an NVIDIA GPU and CUDA runtime libraries. See the section :ref:`dependencies-and-compatibility` for more information.
+
 This backend exposes a set of environment variables to configure specific aspects of the simulation:
 
 * **`CUDAQ_CUTN_HOST_RAM=8`**: Prescribes the size of the CPU Host RAM allocated by each MPI process (defaults to 4 GB). A rule of thumb is to give each MPI process the same amount of CPU Host RAM as the RAM size of the GPU assigned to it. If there is more CPU RAM available, it is fine to further increase this number.

pyproject.toml

@@ -28,13 +28,6 @@ Releases = "https://nvidia.github.io/cuda-quantum/latest/releases.html"
 
 [project.optional-dependencies]
 chemistry = [ "scipy==1.10.1", "openfermionpyscf==0.5" ]
-cudart = [
-  # cublas is (the only library) need for the simulators but available for x86_64 only
-  'nvidia-cublas-cu11 ~= 11.11',
-  'nvidia-cusolver-cu11 ~= 11.4',
-  'nvidia-cusparse-cu11 ~= 11.7',
-  'nvidia-cuda-runtime-cu11 ~= 11.8'
-]
 
 [build-system]
 requires = ["scikit-build-core", "cmake>=3.26"]

python/README.md

@@ -13,33 +13,52 @@ computing, including CUDA, ISO standard parallelism, OpenMP, and OpenACC
   cuQuantum and a number of different physical quantum processors (QPUs)
 
 The CUDA Quantum Python wheels contain the Python API and core components of
-CUDA Quantum. For more information about available packages and documentation,
-see our [release
-notes](https://nvidia.github.io/cuda-quantum/latest/releases.html).
+CUDA Quantum. More information about available packages as well as a link to the
+documentation and examples for each version can be found in the [release
+notes][cudaq_docs_releases]. System requirements and compatibility are listed in
+the Getting Started section of the linked documentation.
+
+[cudaq_docs_releases]:
+    https://nvidia.github.io/cuda-quantum/latest/releases.html
 
 ## Installing CUDA Quantum
 
-CUDA Quantum Python wheels are available on
-[PyPI.org](https://pypi.org/project/cuda-quantum). To install the latest
-release, simply run
+To install the latest stable version of CUDA Quantum, run
 
 ```console
-pip install cuda-quantum
+python3 -m pip install cuda-quantum
 ```
 
-At this time, wheels are distributed for Linux operating systems only. To build
-the CUDA Quantum Python API from source using pip:
+CUDA Quantum can be used to compile and run quantum programs on a CPU-only
+system, but a GPU is highly recommended and necessary to use the some of the
+simulators. The GPU-based simulators included in the CUDA Quantum Python wheels
+require an existing CUDA installation. Additionally, multi-GPU simulators
+require an existing MPI installation.
+
+In most cases, the CUDA and MPI dependencies can be installed via package
+manager. On Ubuntu 22.04, for example, the following commands install all
+optional CUDA dependencies:
 
 ```console
-git clone https://github.com/NVIDIA/cuda-quantum.git
-cd cuda-quantum && ./scripts/install_prerequisites.sh
-pip install .
+  arch=x86_64 # set this to sbsa for ARM processors
+  sudo apt-get update && sudo apt-get install -y wget
+  wget -q https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/$arch/cuda-keyring_1.0-1_all.deb
+  sudo dpkg -i cuda-keyring_1.0-1_all.deb
+  sudo apt-get update && sudo apt-get install -y cuda-11-8
 ```
 
-For more information about building the entire C++ and Python API's, please see
-the CUDA Quantum [documentation][official_install].
+Detailed instructions for how to install the complete CUDA toolkit on different
+operating systems can be found in the [CUDA
+documentation](https://docs.nvidia.com/cuda/).
 
-[official_install]: https://nvidia.github.io/cuda-quantum/latest/install.html
+If you have several GPUs available but no MPI installation yet, we recommend
+taking a look at the [OpenMPI documentation](https://docs.open-mpi.org/). On
+Ubuntu 22.04, for example, the following commands install the necessary MPI
+libraries:
+
+```console
+  sudo apt-get update && sudo apt-get install openmpi
+```
 
 ## Running CUDA Quantum
 
@@ -57,15 +76,8 @@ kernel.mz(qubit)
 result = cudaq.sample(kernel)
 ```
 
-## Documentation
-
-To see more examples, go to [python examples][python_examples], or check out the
-[Python API reference][python_api_reference].
-
-[python_examples]:
-    https://nvidia.github.io/cuda-quantum/latest/using/python.html
-[python_api_reference]:
-    https://nvidia.github.io/cuda-quantum/latest/api/languages/python_api.html
+Additional examples and documentation are linked in the [release
+notes][cudaq_docs_releases].
 
 ## Contributing