Add qibojit backend

.github/workflows/rules.yml

@@ -26,14 +26,14 @@ jobs:
         python -m pip install --upgrade pip
         pip install pylint
         pip install pytest-cov
-        pip install .[qibotf,docs,tests]
+        pip install .[qibojit,qibotf,docs,tests]
     - name: Install package on Windows
       if: startsWith(matrix.os, 'windows')
       run: |
         python -m pip install --upgrade pip
         pip install pylint
         pip install pytest-cov
-        pip install .[docs,tests]
+        pip install .[qibojit,docs,tests]
     - name: Test with pylint
       run: |
         pylint src -E -d E1123,E1120
@@ -43,7 +43,7 @@ jobs:
     - name: Test examples
       if: startsWith(matrix.os, 'ubuntu') && matrix.python-version == '3.9'
       run: |
-        pytest examples/
+        OMP_NUM_THREADS=1 pytest examples/
     - name: Upload coverage to Codecov
       if: startsWith(matrix.os, 'ubuntu')
       uses: codecov/codecov-action@v1

.github/workflows/wheels.yml

@@ -38,15 +38,15 @@ jobs:
       if: startsWith(matrix.os, 'windows') == 0
       run: |
         python -m pip install --upgrade pip
-        pip install qibotf pytest cirq ply sklearn
+        pip install qibojit qibotf pytest cirq ply sklearn
         pip install -r requirements.txt
         pip install qibo --no-index --find-links ./dist/
         pytest --pyargs qibo
     - name: Test wheels on windows
       if: startsWith(matrix.os, 'windows')
       run: |
         python -m pip install --upgrade pip
-        pip install pytest cirq ply sklearn
+        pip install qibojit pytest cirq ply sklearn
         pip install -r requirements.txt
         pip install qibo --no-index --find-links ./dist/
         pytest --pyargs qibo

doc/source/advancedexamples.rst

@@ -64,10 +64,11 @@ raised prompting the user to switch the default device using ``qibo.set_device``
 Setting the number of CPU threads
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
-Qibo by default uses the ``"custom"`` operators backend. This backend uses
-OpenMP instructions for the parallelization. In most cases will utilize all
-available CPU threads. For small circuits the parallelization overhead may
-decrease performance making single thread execution preferrable.
+Qibo by default uses the ``qibojit`` or ``qibotf`` backends which are based on
+custom operators. These backends uses OpenMP instructions for parallelization.
+In most cases, utilizing all available CPU threads provides better performance.
+However, for small circuits the parallelization overhead may decrease
+performance making single thread execution preferrable.
 
 You can restrict the number of threads by exporting the ``OMP_NUM_THREADS``
 environment variable with the requested number of threads before launching Qibo,
@@ -81,8 +82,8 @@ or programmatically, during runtime, as follows:
     # retrieve the current number of threads
     current_threads = qibo.get_threads()
 
-On the other hand, when using the ``"matmuleinsum"`` or ``"defaulteinsum"``
-backends Qibo inherits Tensorflow's defaults for CPU thread configuration.
+On the other hand, when using the ``tensorflow`` backend Qibo inherits
+Tensorflow's defaults for CPU thread configuration.
 Tensorflow allows restricting the number of threads as follows:
 
 .. code-block::  python
@@ -561,7 +562,7 @@ refer to the :ref:`Optimizers <Optimizers>` section of the documentation.
 Note that if the Stochastic Gradient Descent optimizer is used then the user
 has to use a backend based on tensorflow primitives and not the default custom
 backend, as custom operators currently do not support automatic differentiation.
-To switch the backend one can do ``qibo.set_backend("matmuleinsum")``.
+To switch the backend one can do ``qibo.set_backend("tensorflow")``.
 Check the :ref:`How to use automatic differentiation? <autodiff-example>`
 section for more details.
 
@@ -707,9 +708,8 @@ function.
 
 .. code-block:: python
 
-    # switch backend to "matmuleinsum" or "defaulteinsum"
     import qibo
-    qibo.set_backend("matmuleinsum")
+    qibo.set_backend("tensorflow")
     import tensorflow as tf
     from qibo import gates, models
 
@@ -733,9 +733,8 @@ function.
         optimizer.apply_gradients(zip([grads], [params]))
 
 
-Note that a backend that uses tensorflow primitives gates
-(either ``"matmuleinsum"`` or ``"defaulteinsum"``) has to be used because
-the default ``"custom"`` backend does not support automatic differentiation.
+Note that the ``"tensorflow"`` backend has to be used here because ``"qibotf"``
+and other custom backends do not support automatic differentiation.
 
 The optimization procedure may also be compiled, however in this case it is not
 possible to use :meth:`qibo.abstractions.circuit.AbstractCircuit.set_parameters` as the
@@ -745,7 +744,7 @@ For example:
 .. code-block:: python
 
     import qibo
-    qibo.set_backend("matmuleinsum")
+    qibo.set_backend("tensorflow")
     import tensorflow as tf
     from qibo import gates, models
 

doc/source/applications.rst

@@ -1,27 +1,143 @@
-Application tutorials
-=====================
+Applications
+============
 
 In this section we present some examples of quantum circuits applied to specific
 problems.
 
+_______________________
+
+Applications by topic
+---------------------
+
+Complexity theory
+^^^^^^^^^^^^^^^^^
 
 .. toctree::
     :maxdepth: 1
 
-    tutorials/aavqe/README.md
     tutorials/grover3sat/README.md
+    tutorials/adiabatic3sat/README.md
+
+
+Condensed Matter Physics
+^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/aavqe/README.md
+
+
+Cryptography
+^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
     tutorials/hash-grover/README.md
+
+
+Finance
+^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/unary/README.md
+
+
+High-Energy Physics
+^^^^^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/qPDF/qPDF.ipynb
+
+
+Quantum Physics
+^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/qsvd/README.md
+    tutorials/3_tangle/README.md
+    tutorials/adiabatic/README.md
+    tutorials/shor/README.md
+    tutorials/bell-variational/README.md
+    tutorials/falqon/README.md
+    tutorials/grover/README.md
+
+Quantum Machine Learning
+^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/variational_classifier/README.md
+    tutorials/reuploading_classifier/README.md
+    tutorials/autoencoder/README.md
+    tutorials/EF_QAE/README.md
+
+_______________________
+
+Applications by algorithm
+-------------------------
+
+Variational Quantum Circuits
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/aavqe/README.md
     tutorials/variational_classifier/README.md
     tutorials/reuploading_classifier/README.md
     tutorials/autoencoder/README.md
     tutorials/qsvd/README.md
     tutorials/3_tangle/README.md
+    tutorials/EF_QAE/README.md
     tutorials/unary/README.md
+    tutorials/bell-variational/README.md
+    tutorials/qPDF/qPDF.ipynb
+
+
+Grover's Algorithm
+^^^^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/grover3sat/README.md
+    tutorials/hash-grover/README.md
+    tutorials/grover/README.md
+
+
+Shor's Factorization Algorithms
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
+    tutorials/shor/README.md
+
+
+Adiabatic Evolution
+^^^^^^^^^^^^^^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
     tutorials/adiabatic/README.md
     tutorials/adiabatic3sat/README.md
-    tutorials/EF_QAE/README.md
-    tutorials/shor/README.md
-    tutorials/qPDF/qPDF.ipynb
-    tutorials/bell-variational/README.md
+
+
+FALQON
+^^^^^^
+
+.. toctree::
+    :maxdepth: 1
+
     tutorials/falqon/README.md
-    tutorials/grover/README.md
+

doc/source/examples.rst

@@ -35,10 +35,9 @@ evaluation performance, e.g.:
 .. code-block:: python
 
     import numpy as np
-    # switch backend to "matmuleinsum" or "defaulteinsum"
-    # (slower than default "custom" backend)
+    # switch backend to "tensorflow"
     import qibo
-    qibo.set_backend("matmuleinsum")
+    qibo.set_backend("tensorflow")
     from qibo.models import Circuit
     from qibo import gates
 
@@ -52,10 +51,10 @@ evaluation performance, e.g.:
         init_state = np.ones(4) / 2.0 + i
         c(init_state)
 
-Note that compiling is only supported when native tensorflow gates are used.
-This happens when the calculation backend is switched to ``"matmuleinsum"``
-or ``"defaulteinsum"``. This backend is much slower than the default ``"custom"``
-backend which uses custom tensorflow operators to apply gates.
+Note that compiling is only supported when the native ``tensorflow`` backend
+is used. This backend is much slower than ``qibotf`` which uses custom
+tensorflow operators to apply gates.
+
 
 How to print a circuit summary?
 -------------------------------