Release 0.10.4

.github/workflows/deploy.yml

@@ -134,8 +134,40 @@ jobs:
           python -m pip install cibuildwheel==2.2.2
       - name: Build wheels
         env:
-          CIBW_BEFORE_ALL: "yum install -y yum-utils wget && wget https://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda-repo-rhel6-10-1-local-10.1.243-418.87.00-1.0-1.x86_64.rpm && rpm -i cuda-repo-rhel6-10-1-local-10.1.243-418.87.00-1.0-1.x86_64.rpm && yum clean all && yum -y install cuda-10-1 openblas-devel"
-          CIBW_SKIP: "*-manylinux_i686 pp* *musllinux*"
+          CIBW_BEFORE_ALL: "yum install -y yum-utils wget && wget -q https://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda-repo-rhel6-10-1-local-10.1.243-418.87.00-1.0-1.x86_64.rpm && rpm -i cuda-repo-rhel6-10-1-local-10.1.243-418.87.00-1.0-1.x86_64.rpm && yum clean all && yum -y install cuda-10-1 openblas-devel"
+          CIBW_SKIP: "*-manylinux_i686 cp310* pp* cp36* *musllinux*"
+          CIBW_ENVIRONMENT: QISKIT_AER_PACKAGE_NAME=qiskit-aer-gpu AER_THRUST_BACKEND=CUDA CUDACXX=/usr/local/cuda/bin/nvcc
+        run: |
+          python -m cibuildwheel --output-dir wheelhouse
+      - uses: actions/upload-artifact@v2
+        with:
+          path: ./wheelhouse/*.whl
+      - name: Publish Wheels
+        env:
+          TWINE_PASSWORD: ${{ secrets.TWINE_PASSWORD }}
+          TWINE_USERNAME: qiskit
+        run : |
+          pip install -U twine
+          twine upload wheelhouse/*
+  gpu-build-310:
+    name: Build qiskit-aer-gpu wheels
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - uses: actions/setup-python@v2
+        name: Install Python
+        with:
+          python-version: '3.7'
+      - name: Add msbuild to PATH
+        uses: microsoft/setup-msbuild@v1.0.2
+        if: runner.os == 'Windows'
+      - name: Install cibuildwheel
+        run: |
+          python -m pip install cibuildwheel==2.2.2
+      - name: Build wheels
+        env:
+          CIBW_BEFORE_ALL: "yum install -y yum-utils wget && wget -q https://developer.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda-repo-rhel7-10-1-local-10.1.105-418.39-1.0-1.x86_64.rpm && rpm -i cuda-repo-rhel7-10-1-local-10.1.105-418.39-1.0-1.x86_64.rpm && yum clean all && yum -y install cuda-10-1 openblas-devel"
+          CIBW_BUILD: "cp310-manylinux_x86_64"
           CIBW_ENVIRONMENT: QISKIT_AER_PACKAGE_NAME=qiskit-aer-gpu AER_THRUST_BACKEND=CUDA CUDACXX=/usr/local/cuda/bin/nvcc
         run: |
           python -m cibuildwheel --output-dir wheelhouse

CMakeLists.txt

@@ -110,7 +110,6 @@ if(STATIC_LINKING)
 endif()
 
 if(NOT MSVC)
-	enable_cxx_compiler_flag_if_supported("-ffast-math")
 	if(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le")
 		# PowerPC builds are not meant to be redistributable, we build them
 		# in place, so we can have CPU = native.

constraints.txt

@@ -7,3 +7,8 @@ scipy>=1.0
 # with modern importlib-metadata (4.8.1). importlib-metadata is only needed on
 # Python <3.8.
 importlib-metadata==4.6.4
+
+# Jinja2 3.1.0 is incompatible with sphinx and/or jupyter until they are updated
+# to work with the new jinja version (the jinja maintainers aren't going to
+# fix things) pin to the previous working version.
+jinja2==3.0.3

docs/conf.py

@@ -46,7 +46,7 @@
 # The short X.Y version
 version = ''
 # The full version, including alpha/beta/rc tags
-release = '0.10.3'
+release = '0.10.4'
 
 # -- General configuration ---------------------------------------------------
 

qiskit/providers/aer/VERSION.txt

@@ -1 +1 @@
-0.10.3
+0.10.4

qiskit/providers/aer/backends/aerbackend.py

@@ -288,9 +288,30 @@ def _run(self, qobj, job_id='', format_result=True):
         # Start timer
         start = time.time()
 
+        # Take metadata from headers of experiments to work around JSON serialization error
+        metadata_list = []
+        metadata_index = 0
+        for expr in qobj.experiments:
+            if hasattr(expr.header, "metadata"):
+                metadata_copy = expr.header.metadata.copy()
+                metadata_list.append(metadata_copy)
+                expr.header.metadata.clear()
+                if "id" in metadata_copy:
+                    expr.header.metadata["id"] = metadata_copy["id"]
+                expr.header.metadata["metadata_index"] = metadata_index
+                metadata_index += 1
+
         # Run simulation
         output = self._execute(qobj)
 
+        # Recover metadata
+        metadata_index = 0
+        for expr in qobj.experiments:
+            if hasattr(expr.header, "metadata"):
+                expr.header.metadata.clear()
+                expr.header.metadata.update(metadata_list[metadata_index])
+                metadata_index += 1
+
         # Validate output
         if not isinstance(output, dict):
             logger.error("%s: simulation failed.", self.name())
@@ -305,6 +326,14 @@ def _run(self, qobj, job_id='', format_result=True):
         output["backend_name"] = self.name()
         output["backend_version"] = self.configuration().backend_version
 
+        # Push metadata to experiment headers
+        for result in output["results"]:
+            if ("header" in result and
+                    "metadata" in result["header"] and
+                    "metadata_index" in result["header"]["metadata"]):
+                metadata_index = result["header"]["metadata"]["metadata_index"]
+                result["header"]["metadata"] = metadata_list[metadata_index]
+
         # Add execution time
         output["time_taken"] = time.time() - start
 

qiskit/providers/aer/noise/device/models.py

@@ -245,6 +245,7 @@ def _device_thermal_relaxation_error(qubits,
     error = None
     for qubit in qubits:
         t1, t2, freq = relax_params[qubit]
+        t2 = _truncate_t2_value(t1, t2)
         population = _excited_population(freq, temperature)
         if first:
             error = thermal_relaxation_error(t1, t2, gate_time, population)
@@ -255,6 +256,17 @@ def _device_thermal_relaxation_error(qubits,
     return error
 
 
+def _truncate_t2_value(t1, t2):
+    """Return t2 value truncated to 2 * t1 (for t2 > 2 * t1)"""
+    new_t2 = t2
+    if t2 > 2 * t1:
+        new_t2 = 2 * t1
+        warn("Device model returned an invalid T_2 relaxation time greater than"
+             f" the theoretical maximum value 2 * T_1 ({t2} > 2 * {t1})."
+             " Truncating to maximum value.", UserWarning)
+    return new_t2
+
+
 def _excited_population(freq, temperature):
     """Return excited state population"""
     population = 0

qiskit/providers/aer/noise/device/parameters.py

@@ -180,11 +180,6 @@ def thermal_relaxation_values(properties):
                 # Convert to Gigahertz
                 freq *= _GHZ_UNITS.get(freq_params.unit, 1)
 
-        # NOTE: T2 cannot be larger than 2 * T1 for a physical noise
-        # channel, however if a backend erroneously reports such a value we
-        # truncated it here:
-        t2 = min(2 * t1, t2)
-
         values.append((t1, t2, freq))
     return values
 

qiskit/providers/aer/noise/noise_model.py

@@ -21,11 +21,10 @@
 from numpy import ndarray
 
 from qiskit.circuit import Instruction, Delay
-from qiskit.providers import BaseBackend, BackendV1, BackendV2
 from qiskit.providers.exceptions import BackendPropertyError
 from qiskit.providers.models import BackendProperties
 from qiskit.transpiler import PassManager
-from .device.models import _excited_population
+from .device.models import _excited_population, _truncate_t2_value
 from .device.models import basic_device_gate_errors
 from .device.models import basic_device_readout_errors
 from .errors.quantum_error import QuantumError
@@ -303,10 +302,14 @@ def from_backend(cls, backend,
         Raises:
             NoiseError: If the input backend is not valid.
         """
-        if isinstance(backend, BackendV2):
+        backend_interface_version = getattr(backend, "version", None)
+        if not isinstance(backend_interface_version, int):
+            backend_interface_version = 0
+
+        if backend_interface_version == 2:
             raise NoiseError(
                 "NoiseModel.from_backend does not currently support V2 Backends.")
-        if isinstance(backend, (BaseBackend, BackendV1)):
+        if backend_interface_version <= 1:
             properties = backend.properties()
             configuration = backend.configuration()
             basis_gates = configuration.basis_gates
@@ -373,9 +376,11 @@ def from_backend(cls, backend,
             except BackendPropertyError:
                 excited_state_populations = None
             try:
+                t1s = [properties.t1(q) for q in range(num_qubits)]
+                t2s = [properties.t2(q) for q in range(num_qubits)]
                 delay_pass = RelaxationNoisePass(
-                    t1s=[properties.t1(q) for q in range(num_qubits)],
-                    t2s=[properties.t2(q) for q in range(num_qubits)],
+                    t1s=t1s,
+                    t2s=[_truncate_t2_value(t1, t2) for t1, t2 in zip(t1s, t2s)],
                     dt=dt,
                     op_types=Delay,
                     excited_state_populations=excited_state_populations

qiskit/providers/aer/pulse/system_models/pulse_system_model.py

@@ -17,7 +17,7 @@
 
 from warnings import warn
 from collections import OrderedDict
-from qiskit.providers import BaseBackend, Backend
+from qiskit.providers import Backend
 from ...aererror import AerError
 from .hamiltonian_model import HamiltonianModel
 
@@ -94,7 +94,7 @@ def from_backend(cls, backend, subsystem_list=None):
             AerError: If channel or u_channel_lo are invalid.
         """
 
-        if not isinstance(backend, (BaseBackend, Backend)):
+        if not isinstance(backend, Backend):
             raise AerError("{} is not a Qiskit backend".format(backend))
 
         # get relevant information from backend

releasenotes/notes/density-multi-chunk-fix-e9effc67d0365418.yaml

@@ -0,0 +1,13 @@
+---
+fixes:
+  - |
+    Fix cache blocking transpiler to recognize superop to be cache blocked.
+    This is fix for
+    `issue 1479 <https://github.com/Qiskit/qiskit-aer/issues/1479>`
+    now density_matrix with noise models can be parallelized.
+    New test, test_noise.TestNoise.test_kraus_gate_noise_on_QFT_cache_blocking
+    is added to verify this issue.
+    Also this fix include fix for
+    `issue 1483 <https://github.com/Qiskit/qiskit-aer/issues/1483>`
+    discovered by adding new test case.
+    This fixes measure over chunks for statevector.

releasenotes/notes/fix-invalid-t2-error-a3685e4a3ad0a1e7.yaml

@@ -0,0 +1,12 @@
+---
+fixes:
+  - |
+    Fixes a bug in ``NoiseModel.from_backend()`` that raised an error when
+    T2 value greater than 2 * T1 was supplied by the backend.
+    After this fix, it becomes to truncate T2 value up to 2 * T1 and
+    issue a user warning if truncates.
+    The bug was introduced at #1391 and, before that, ``NoiseModel.from_backend()`` had
+    truncated the T2 value up to 2 * T1 silently.
+
+    See `Issue 1464 <https://github.com/Qiskit/qiskit-aer/issues/1464>`__
+    for details.

releasenotes/notes/fix-thrust-cpu-threads-67db86b2edcf06b3.yaml

@@ -0,0 +1,6 @@
+---
+fixes:
+  - |
+    device=Thrust was very slow for small number of qubits because OpenMP
+    threading was always applied. This fix applies OpenMP threads as same
+    as device=CPU by using statevector_parallel_threshold.

releasenotes/notes/no-fast-math-1de357a9650094f3.yaml

@@ -0,0 +1,18 @@
+---
+upgrade:
+  - |
+    Qiskit Aer is no longer compiled with unsafe floating-point optimisations.
+    While most of the effects should have been localised to Qiskit Aer, some
+    aspects of subnormal handling may previously have been leaked into user code
+    by the library incorrectly setting the "flush to zero" mode.  This will not
+    happen any more.
+fixes:
+  - |
+    Qiskit Aer will no longer set the floating-point mode to "flush to zero"
+    when loaded.  Downstream users may previously have seen warnings from Numpy
+    such as:
+
+      The value of the smallest subnormal for <class 'numpy.float64'> type is zero.
+
+    These will now no longer be emitted, and the floating-point handling will be
+    correct.

releasenotes/notes/remove_circuit_metadata_from_qobj-324e7ea9b369ee67.yaml

@@ -0,0 +1,13 @@
+---
+fixes:
+  - |
+    Fixed a potential issue with running simulations on circuits that have the
+    :attr:`.QuantumCircuit.metadata` attribute set. The :attr:`~.QuantumCircuit.metadata`
+    attribute can be any python dictionary and previously qiskit-aer would attempt to
+    JSON serialize the contents of the attribute to process it with the rest of the rest
+    of the circuit input, even if the contents were not JSON serializable. This no longer
+    occurs as the :attr:`.QuantumCircuit.metadata`  attribute is not used to run the
+    simulation so now the contents are no serialized and instead are directly attached
+    to the :class:`qiskit.result.Result` object without attempting to JSON serialize
+    the contents.
+    Fixed `#1435 <https://github.com/Qiskit/qiskit-aer/issues/1435>`__

src/controllers/aer_controller.hpp

@@ -1930,10 +1930,14 @@ bool Controller::validate_state(const state_t &state, const Circuit &circ,
     size_t required_mb = state.required_memory_mb(circ.num_qubits, circ.ops) / num_process_per_experiment_;                                        
     size_t mem_size = (sim_device_ == Device::GPU) ? max_memory_mb_ + max_gpu_memory_mb_ : max_memory_mb_;
     memory_valid = (required_mb <= mem_size);
-  }
-  if (throw_except && !memory_valid) {
-    error_msg << "Insufficient memory to run circuit " << circ_name;
-    error_msg << " using the " << state.name() << " simulator.";
+    if (throw_except && !memory_valid) {
+      error_msg << "Insufficient memory to run circuit " << circ_name;
+      error_msg << " using the " << state.name() << " simulator.";
+      error_msg << " Required memory: " << required_mb << "M, max memory: " << max_memory_mb_ << "M";
+      if (sim_device_ == Device::GPU) {
+        error_msg << " (Host) + " << max_gpu_memory_mb_ << "M (GPU)";
+      }
+    }
   }
 
   if (noise_valid && circ_valid && memory_valid) {

src/simulators/extended_stabilizer/ch_runner.hpp

@@ -664,11 +664,7 @@ void Runner::metropolis_step(AER::RngEngine &rng)
   }
   complex_t ampsum(real_part, imag_part);
   double p_threshold = std::norm(ampsum)/std::norm(old_ampsum_);
-  #ifdef  __FAST_MATH__ //isnan doesn't behave well under fastmath, so use absolute tolerance check instead
-  if(std::isinf(p_threshold) || std::abs(std::norm(old_ampsum_)-0.) < 1e-8)
-  #else
   if(std::isinf(p_threshold) || std::isnan(p_threshold))
-  #endif
   {
     accept_ = true;
     old_ampsum_ = ampsum;

src/simulators/statevector/chunk/chunk_container.hpp

@@ -487,6 +487,10 @@ class ChunkContainer : public std::enable_shared_from_this<ChunkContainer<data_t
   uint_t num_cmemory_;
   mutable int_t conditional_bit_;
   bool keep_conditional_bit_;         //keep conditional bit alive
+  int_t num_pow2_qubits_;             //largest number of qubits that meets num_chunks_ = m*(2^num_pow2_qubits_)
+  bool density_matrix_;
+
+  int_t omp_threads_;                 //number of threads can be used for parallelization on CPU
 public:
   ChunkContainer()
   {
@@ -498,6 +502,8 @@ class ChunkContainer : public std::enable_shared_from_this<ChunkContainer<data_t
     conditional_bit_ = -1;
     keep_conditional_bit_ = false;
     matrix_bits_ = AER_DEFAULT_MATRIX_BITS;
+    density_matrix_ = false;
+    omp_threads_ = 1;
   }
   virtual ~ChunkContainer(){}
 
@@ -569,6 +575,11 @@ class ChunkContainer : public std::enable_shared_from_this<ChunkContainer<data_t
     keep_conditional_bit_ = keep;
   }
 
+  void set_omp_threads(int_t nthreads)
+  {
+    omp_threads_ = nthreads;
+  }
+
   virtual thrust::complex<data_t>& operator[](uint_t i) = 0;
 
   virtual uint_t Allocate(int idev,int chunk_bits,int num_qubits,uint_t chunks,uint_t buffers = AER_MAX_BUFFERS,bool multi_shots = false,int matrix_bit = AER_DEFAULT_MATRIX_BITS) = 0;
@@ -820,7 +831,10 @@ void ChunkContainer<data_t>::Execute(Function func,uint_t iChunk,uint_t count)
 #else
   uint_t size = count * func.size(chunk_bits_);
   auto ci = thrust::counting_iterator<uint_t>(0);
-  thrust::for_each_n(thrust::device, ci , size, func);
+  if(omp_threads_ > 1)
+    thrust::for_each_n(thrust::device, ci , size, func);
+  else
+    thrust::for_each_n(thrust::seq, ci , size, func);
 #endif
 
 }
@@ -971,7 +985,10 @@ void ChunkContainer<data_t>::ExecuteSum(double* pSum,Function func,uint_t iChunk
     auto ci = thrust::counting_iterator<uint_t>(0);
 
     double sum;
-    sum = thrust::transform_reduce(thrust::device, ci, ci + size, func,0.0,thrust::plus<double>());
+    if(omp_threads_ > 1)
+      sum = thrust::transform_reduce(thrust::device, ci, ci + size, func,0.0,thrust::plus<double>());
+    else
+      sum = thrust::transform_reduce(thrust::seq, ci, ci + size, func,0.0,thrust::plus<double>());
     if(count == 1 && pSum){
       *pSum = sum;
     }
@@ -1108,7 +1125,10 @@ void ChunkContainer<data_t>::ExecuteSum2(double* pSum,Function func,uint_t iChun
 
     auto ci = thrust::counting_iterator<uint_t>(0);
 
-    ret = thrust::transform_reduce(thrust::device, ci, ci + size, func,zero,complex_sum());
+    if(omp_threads_ > 1)
+      ret = thrust::transform_reduce(thrust::device, ci, ci + size, func,zero,complex_sum());
+    else
+      ret = thrust::transform_reduce(thrust::seq, ci, ci + size, func,zero,complex_sum());
 
     if(count == 1 && pSum){
       *((thrust::complex<double>*)pSum) = ret;