Merge branch 'main' into raultorres/remove_kokkos_plugin

.github/workflows/build-wheel-linux-x86_64.yaml

@@ -194,6 +194,11 @@ jobs:
       # building with LLD is a strong requirement for mhlo
       run: |
         export PATH=$GITHUB_WORKSPACE/llvm-build/bin:$PATH
+        
+        export TARGET_FILE=mlir/mlir-hlo/mhlo/transforms/CMakeLists.txt
+        export PATCH_FILE=mlir/patches/mhlo-Add-PassesIncGen-in-transforms-CMakeList.patch
+        if patch --dry-run -p1 $TARGET_FILE $PATCH_FILE > /dev/null 2>&1; then patch -p1 $TARGET_FILE $PATCH_FILE; fi
+        
         cmake -S mlir/mlir-hlo -B mhlo-build -G Ninja \
               -DCMAKE_BUILD_TYPE=Release \
               -DLLVM_ENABLE_ASSERTIONS=ON \
@@ -204,7 +209,7 @@ jobs:
               -DCMAKE_CXX_VISIBILITY_PRESET=protected \
               -DLLVM_ENABLE_LLD=ON
 
-        cmake --build mhlo-build --target check-mlir-hlo
+        LIT_FILTER_OUT="chlo_legalize_to_mhlo" cmake --build mhlo-build --target check-mlir-hlo
 
     - name: Build Enzyme
       if: steps.cache-enzyme-build.outputs.cache-hit != 'true'

.github/workflows/build-wheel-macos-arm64.yaml

@@ -183,6 +183,11 @@ jobs:
       if: steps.cache-mhlo-build.outputs.cache-hit != 'true'
       run: |
         export PATH=${{ steps.setup_env.outputs.dependency_build_dir }}/llvm-build/bin:$PATH
+        
+        export TARGET_FILE=mlir/mlir-hlo/mhlo/transforms/CMakeLists.txt
+        export PATCH_FILE=mlir/patches/mhlo-Add-PassesIncGen-in-transforms-CMakeList.patch
+        if patch --dry-run -p1 $TARGET_FILE $PATCH_FILE > /dev/null 2>&1; then patch -p1 $TARGET_FILE $PATCH_FILE; fi
+        
         cmake -S ${{ steps.setup_env.outputs.dependency_build_dir }}/mlir/mlir-hlo -B ${{ steps.setup_env.outputs.dependency_build_dir }}/mhlo-build -G Ninja \
               -DCMAKE_BUILD_TYPE=Release \
               -DLLVM_ENABLE_ASSERTIONS=ON \

.github/workflows/build-wheel-macos-x86_64.yaml

@@ -155,6 +155,11 @@ jobs:
       if: steps.cache-mhlo-build.outputs.cache-hit != 'true'
       run: |
         export PATH=$GITHUB_WORKSPACE/llvm-build/bin:$PATH
+        
+        export TARGET_FILE=mlir/mlir-hlo/mhlo/transforms/CMakeLists.txt
+        export PATCH_FILE=mlir/patches/mhlo-Add-PassesIncGen-in-transforms-CMakeList.patch
+        if patch --dry-run -p1 $TARGET_FILE $PATCH_FILE > /dev/null 2>&1; then patch -p1 $TARGET_FILE $PATCH_FILE; fi
+        
         cmake -S mlir/mlir-hlo -B mhlo-build -G Ninja \
               -DCMAKE_BUILD_TYPE=Release \
               -DLLVM_ENABLE_ASSERTIONS=ON \

.github/workflows/scripts/linux_arm64/rh8/build_llvm.sh

@@ -48,4 +48,4 @@ cmake -S /catalyst/mlir/llvm-project/llvm -B /catalyst/llvm-build -G Ninja \
     -DPython3_EXECUTABLE=/usr/bin/python3 \
     -DPython3_NumPy_INCLUDE_DIRS=/opt/_internal/cpython-${PYTHON_VERSION}.${PYTHON_SUBVERSION}/lib/python${PYTHON_VERSION}/site-packages/numpy/core/include \
     -DCMAKE_CXX_VISIBILITY_PRESET=hidden
-LIT_FILTER_OUT=\"Bytecode\" cmake --build /catalyst/llvm-build --target check-mlir llvm-symbolizer
+LIT_FILTER_OUT="Bytecode|tosa-to-tensor" cmake --build /catalyst/llvm-build --target check-mlir llvm-symbolizer

.github/workflows/scripts/linux_arm64/rh8/build_mhlo.sh

@@ -36,6 +36,10 @@ export PATH=/catalyst/llvm-build/bin:/opt/_internal/cpython-${PYTHON_VERSION}.${
 # Patch MHLO for correct linking
 sed -i -e 's/LINK_LIBS PUBLIC/LINK_LIBS PUBLIC MLIRDeallocationUtils/g' mlir/mlir-hlo/deallocation/transforms/CMakeLists.txt
 
+export TARGET_FILE=mlir/mlir-hlo/mhlo/transforms/CMakeLists.txt
+export PATCH_FILE=mlir/patches/mhlo-Add-PassesIncGen-in-transforms-CMakeList.patch
+if patch --dry-run -p1 $TARGET_FILE $PATCH_FILE > /dev/null 2>&1; then patch -p1 $TARGET_FILE $PATCH_FILE; fi
+
 # Build MHLO
 cmake -S /catalyst/mlir/mlir-hlo -B /catalyst/mhlo-build -G Ninja \
     -DCMAKE_BUILD_TYPE=Release \
@@ -46,4 +50,4 @@ cmake -S /catalyst/mlir/mlir-hlo -B /catalyst/mhlo-build -G Ninja \
     -DLLVM_ENABLE_ZLIB=OFF \
     -DLLVM_ENABLE_ZSTD=FORCE_ON \
     -DCMAKE_CXX_VISIBILITY_PRESET=hidden
-cmake --build /catalyst/mhlo-build --target check-mlir-hlo
+LIT_FILTER_OUT="chlo_legalize_to_mhlo" cmake --build /catalyst/mhlo-build --target check-mlir-hlo

doc/changelog.md

@@ -266,6 +266,12 @@
       return y
   ```
 
+<h3>Documentation</h3>
+
+* A page has been added to the documentation, listing devices that are
+  Catalyst compatible.
+  [(#966)](https://github.com/PennyLaneAI/catalyst/pull/966)
+
 <h3>Internal changes</h3>
 
 * llvm O2 and Enzyme passes are only run when needed (gradients presents). Async execution of QNodes triggers now triggers a
@@ -303,10 +309,11 @@ This release contains contributions from (in alphabetical order):
 
 Alessandro Cosentino,
 Lillian M. A. Frederiksen,
+Josh Izaac,
 Kunwar Maheep Singh,
 Mehrdad Malekmohammadi,
 Romain Moyard,
-Erick Ochoa,
+Erick Ochoa Lopez,
 Mudit Pandey,
 nate stemen,
 Raul Torres,

doc/dev/devices.rst

@@ -0,0 +1,89 @@
+
+.. raw:: html
+
+    <style>
+    tr {
+        border-top: #d3d3d3 1px solid;
+        border-bottom: #d3d3d3 1px solid;
+    }
+    .row-odd {
+        background-color: #f7f7f7;
+    }
+    </style>
+
+Supported devices
+=================
+
+Not all PennyLane devices currently work with Catalyst, and for those that do, their supported
+feature set may not necessarily match supported features when used without :func:`~.qjit`.
+
+Supported backend devices include:
+
+.. list-table::
+  :widths: 20 80
+
+  * - ``lightning.qubit``
+
+    - A fast state-vector qubit simulator written with a C++ backend. See the
+      `Lightning documentation <https://docs.pennylane.ai/projects/lightning/en/stable/lightning_qubit/device.html>`__
+      for more details, as well as its
+      `Catalyst configuration file <https://github.com/PennyLaneAI/pennylane-lightning/blob/master/pennylane_lightning/lightning_qubit/lightning_qubit.toml>`__
+      for natively supported instructions.
+
+  * - ``lightning.kokkos``
+
+    - A fast state-vector qubit simulator utilizing the Kokkos library for CPU and GPU accelerated
+      circuit simulation. See the
+      `Lightning documentation <https://docs.pennylane.ai/projects/lightning/en/stable/lightning_kokkos/device.html>`__
+      for more details, as well as its
+      `Catalyst configuration file <https://github.com/PennyLaneAI/pennylane-lightning/blob/master/pennylane_lightning/lightning_kokkos/lightning_kokkos.toml>`__
+      for natively supported instructions.
+
+  * - ``braket.aws.qubit``
+
+    - Interact with quantum computing hardware devices and simulators through Amazon Braket. To use
+      this device with Catalyst, make sure to install the
+      `PennyLane-Braket plugin <https://amazon-braket-pennylane-plugin-python.readthedocs.io/en/latest/>`__.
+      See the
+      `Catalyst configuration file <https://github.com/PennyLaneAI/catalyst/blob/main/runtime/lib/backend/openqasm/braket_aws_qubit.toml>`__
+      for natively supported instructions.
+
+  * - ``braket.local.qubit``
+
+    - Run gate-based circuits on the Braket SDK's local simulator. To use
+      this device with Catalyst, make sure to install the
+      `PennyLane-Braket plugin <https://amazon-braket-pennylane-plugin-python.readthedocs.io/en/latest/>`__.
+      See the
+      `Catalyst configuration file <https://github.com/PennyLaneAI/catalyst/blob/main/runtime/lib/backend/openqasm/braket_local_qubit.toml>`__
+      for natively supported instructions.
+
+  * - ``qrack.simulator``
+
+    - `Qrack <https://github.com/unitaryfund/qrack>`__ is a GPU-accelerated quantum computer
+      simulator with many novel optimizations including hybrid stabilizer simulation. To use this
+      device with Catalyst, make sure to install the
+      `PennyLane-Qrack plugin <https://pennylane-qrack.readthedocs.io/en/latest/>`__, and check out
+      the `QJIT compilation with Qrack and Catalyst tutorial <https://pennylane.ai/qml/demos/qrack/>`__.
+
+      See the `Catalyst configuration file <https://github.com/unitaryfund/pennylane-qrack/blob/master/pennylane_qrack/QrackDeviceConfig.toml>`__
+      for natively supported instructions.
+
+  * - ``oqc.cloud``
+
+    - Execute on `Oxford Quantum Circuits (OQC) <https://www.oqc.tech/>`__ superconducting hardware,
+      via `OQC Cloud <https://docs.oqc.app>`__. To use OQC Cloud with Catalyst, simply `install the
+      client <https://docs.oqc.app/installation.html>`__, ensure your credentials are set as
+      environment variables, and use the ``backend`` argument to specify the OQC backend to use:
+
+      .. code-block:: python
+
+          import os
+          os.environ["OQC_EMAIL"] = "your_email"
+          os.environ["OQC_PASSWORD"] = "your_password"
+          os.environ["OQC_URL"] = "oqc_url"
+
+          dev = qml.device("oqc.cloud", backend="lucy", shots=2012, wires=2)
+
+      See the `Catalyst configuration file <https://github.com/PennyLaneAI/catalyst/blob/main/frontend/catalyst/third_party/oqc/src/oqc.toml>`__
+      for natively supported instructions.
+

doc/dev/quick_start.rst

@@ -40,8 +40,9 @@ PennyLane. However, some of PennyLane's features may not be fully supported yet,
 
 .. warning::
 
-    The supported backend devices are currently ``lightning.qubit``, ``lightning.kokkos``,
-    ``braket.local.qubit``, ``braket.aws.qubit``, and ``oqc.cloud`` but future plans include the addition of more.
+    Not all PennyLane devices currently work with Catalyst. Supported backend devices include
+    ``lightning.qubit``, ``lightning.kokkos``, and ``braket.aws.qubit``. For
+    a full of supported devices, please see :doc:`/dev/devices`.
 
 PennyLane tapes are still used internally by Catalyst and you can express your circuits in the
 way you are used to, as long as you ensure that all operations are added to the main tape.
@@ -101,101 +102,19 @@ PennyLane are also supported via the decomposition mechanism in Catalyst. For ex
         qml.adjoint(qml.SingleExcitation(jnp.pi / 3, wires=[0, 1]))
         return qml.state()
 
-In addition, you can jit most of :doc:`PennyLane templates <introduction/templates>` to easily construct and evaluate
-more complex quantum circuits; see below for the list of currently supported operations and templates.
+In addition, you can qjit most :doc:`PennyLane templates <introduction/templates>` to easily construct and evaluate
+more complex quantum circuits.
 
 .. important::
 
    Decomposition will generally happen in accordance with the specification provided by devices,
-   which can vary from device to device (e.g. ``default.qubit`` and ``lightning.qubit`` might
-   decompose quite differently.)
+   which can vary from device to device (e.g., ``default.qubit`` and ``lightning.qubit`` might
+   decompose quite differently).
    However, Catalyst's decomposition logic will differ in the following cases:
 
    1. All :class:`qml.Controlled <pennylane.ops.op_math.Controlled>` operations will decompose to :class:`qml.QubitUnitary <pennylane.QubitUnitary>` operations.
    2. The set of operations supported by Catalyst itself can in some instances lead to additional decompositions compared to the device itself.
 
-.. raw:: html
-
-    <div class="summary-table">
-
-.. details::
-    :title: List of supported native operations
-
-    .. autosummary::
-        :nosignatures:
-
-        ~pennylane.Identity
-        ~pennylane.PauliX
-        ~pennylane.PauliY
-        ~pennylane.PauliZ
-        ~pennylane.Hadamard
-        ~pennylane.S
-        ~pennylane.T
-        ~pennylane.PhaseShift
-        ~pennylane.RX
-        ~pennylane.RY
-        ~pennylane.RZ
-        ~pennylane.CNOT
-        ~pennylane.CY
-        ~pennylane.CZ
-        ~pennylane.SWAP
-        ~pennylane.IsingXX
-        ~pennylane.IsingYY
-        ~pennylane.IsingXY
-        ~pennylane.IsingZZ
-        ~pennylane.ControlledPhaseShift
-        ~pennylane.CRX
-        ~pennylane.CRY
-        ~pennylane.CRZ
-        ~pennylane.CRot
-        ~pennylane.CSWAP
-        ~pennylane.MultiRZ
-        ~pennylane.QubitUnitary
-
-.. raw:: html
-
-    </div>
-    <div class="summary-table">
-
-.. details::
-    :title: List of supported templates
-
-    .. autosummary::
-        :nosignatures:
-
-        ~pennylane.AllSinglesDoubles
-        ~pennylane.AmplitudeEmbedding
-        ~pennylane.AngleEmbedding
-        ~pennylane.ApproxTimeEvolution
-        ~pennylane.ArbitraryStatePreparation
-        ~pennylane.BasicEntanglerLayers
-        ~pennylane.BasisEmbedding
-        ~pennylane.BasisStatePreparation
-        ~pennylane.broadcast
-        ~pennylane.FermionicDoubleExcitation
-        ~pennylane.FermionicSingleExcitation
-        ~pennylane.FlipSign
-        ~pennylane.GateFabric
-        ~pennylane.GroverOperator
-        ~pennylane.IQPEmbedding
-        ~pennylane.kUpCCGSD
-        ~pennylane.MERA
-        ~pennylane.MottonenStatePreparation
-        ~pennylane.MPS
-        ~pennylane.Permute
-        ~pennylane.QAOAEmbedding
-        ~pennylane.QFT
-        ~pennylane.QuantumMonteCarlo
-        ~pennylane.QuantumPhaseEstimation
-        ~pennylane.RandomLayers
-        ~pennylane.SimplifiedTwoDesign
-        ~pennylane.StronglyEntanglingLayers
-        ~pennylane.TTN
-        ~pennylane.UCCSD
-
-.. raw:: html
-
-    </div>
 
 Observables
 -----------
@@ -218,28 +137,6 @@ a tensor product of a :class:`qml.PauliX <pennylane.PauliX>`, :class:`qml.Hadama
         )
         return qml.expval(qml.PauliX(0) @ qml.Hadamard(1) @ qml.Hermitian(h_matrix, 2))
 
-.. raw:: html
-
-    <div class="summary-table">
-
-.. details::
-    :title: List of supported observables
-
-    .. autosummary::
-        :nosignatures:
-
-        ~pennylane.Identity
-        ~pennylane.PauliX
-        ~pennylane.PauliY
-        ~pennylane.PauliZ
-        ~pennylane.Hadamard
-        ~pennylane.Hermitian
-        ~pennylane.Hamiltonian
-
-.. raw:: html
-
-    </div>
-
 Measurements
 ------------
 Most PennyLane :doc:`measurement processes <introduction/measurements>`

doc/index.rst

@@ -67,6 +67,7 @@ Catalyst
 
    dev/installation
    dev/quick_start
+   dev/devices
    dev/autograph
    dev/sharp_bits
    dev/jax_integration

frontend/catalyst/jit.py

@@ -90,9 +90,9 @@ def qjit(
 
     .. note::
 
-        The supported backend devices are currently ``lightning.qubit``, ``lightning.kokkos``,
-        ``braket.local.qubit``, ``braket.aws.qubit``, and ``oqc.cloud``. For a list of supported
-        operations, observables, and measurements, please see the :doc:`/dev/quick_start`.
+        Not all PennyLane devices currently work with Catalyst. Supported backend devices include
+        ``lightning.qubit``, ``lightning.kokkos``, and ``braket.aws.qubit``. For
+        a full of supported devices, please see :doc:`/dev/devices`.
 
     Args:
         fn (Callable): the quantum or classical function

frontend/test/async_tests/test_async.py

@@ -26,6 +26,7 @@
 # pylint: disable=missing-function-docstring
 
 
+@pytest.mark.skip(reason="Skip mlir async destruction errors.")
 def test_qnode_execution(backend):
     """The two first QNodes are executed in parrallel."""
     dev = qml.device(backend, wires=2)
@@ -67,6 +68,7 @@ def circuit3(params):
 # ("parameter-shift", "auto"), ("adjoint", "auto")]
 @pytest.mark.parametrize("diff_methods", [("finite-diff", "fd")])
 @pytest.mark.parametrize("inp", [(1.0), (2.0), (3.0), (4.0)])
+@pytest.mark.skip(reason="Skip mlir async destruction errors.")
 def test_gradient(inp, diff_methods, backend):
     """Parameter shift and finite diff generate multiple QNode that are run async."""
 

frontend/test/lit/test_instrumentation_console.py

@@ -15,12 +15,17 @@
 # Redirect stderr to stdout to make the output visible to FileCheck.
 # RUN: %PYTHON %s 2>&1 | FileCheck %s
 
+import jax
 import numpy as np
 import pennylane as qml
 
 from catalyst import qjit
 from catalyst.debug import instrumentation
 
+# Test only on CPU execution platform
+original_jax_platforms = jax.config.jax_platforms if hasattr(jax.config, "jax_platforms") else None
+jax.config.update("jax_platforms", "cpu")
+
 num_layers = 1
 num_wires = 1
 dev = qml.device("lightning.qubit", wires=num_wires)
@@ -77,3 +82,6 @@ def circuit(weights):
 
 with instrumentation(circuit.__name__, filename=None, detailed=True):
     qjit(circuit)(weights)
+
+# Restore original execution platforms
+jax.config.update("jax_platforms", original_jax_platforms)