Update

src/qibotn/test_multinode.py

@@ -1,126 +1,65 @@
-import qibo
+import os
-#import qibotn.cutn as cutn
+import sys
-from cuquantum import cutensornet as cutn
+from timeit import default_timer as timer
-
-from qibo import gates
-from qibo.models import Circuit, QFT
 import numpy as np
-from mpi4py import MPI  # this line initializes MPI
 import cupy as cp
+import cuquantum
+from cuquantum import cutensornet as cutn
+from qibo import gates
+from qibo.models import QFT
+from mpi4py import MPI  # this line initializes MPI
 from cupy.cuda.runtime import getDeviceCount
 from qibotn.QiboCircuitConvertor import QiboCircuitToEinsum
-import cuquantum
 
 def qibo_qft(nqubits, swaps):
     circ_qibo = QFT(nqubits, swaps)
     state_vec = np.array(circ_qibo())
     return circ_qibo, state_vec
 
-print("QiboTN")
+args = sys.argv
+
+if len(args) < 2:
+    print("Usage: python script.py [nqubits] ")
+    sys.exit(1)
+    
+nqubits = int(args[1])
 
 root = 0
 comm = MPI.COMM_WORLD
 rank, size = comm.Get_rank(), comm.Get_size()
-print("Andy: Rank ", rank," size ", size)
-# Assign the device for each process.
 device_id = rank % getDeviceCount()
 cp.cuda.Device(device_id).use()
+#print("Andy: Rank ", rank," size ", size, 'Device count',getDeviceCount())
+
+# Check if the env var is set
+if not "CUTENSORNET_COMM_LIB" in os.environ:
+    raise RuntimeError("need to set CUTENSORNET_COMM_LIB to the path of the MPI wrapper library")
+
+if not os.path.isfile(os.environ["CUTENSORNET_COMM_LIB"]):
+    raise RuntimeError("CUTENSORNET_COMM_LIB does not point to the path of the MPI wrapper library")
 
 datatype = 'complex128'
-nqubits = 10
-'''
-qibo_circ = Circuit(nqubits)
-qibo_circ.add(gates.H(0))
-#qibo_circ.add(gates.CZ(3,4))
-qibo_circ.add(gates.CZ(2,4))
-#qibo_circ.add(gates.CNOT(0,4))
-#qibo_circ.add(gates.SWAP(0,4))
-qibo_circ.add(gates.H(2))
-qibo_circ.add(gates.H(4))
-'''
 qibo_circ = QFT(nqubits)
-
-'''
-expr = 'ehl,gj,edhg,bif,d,c,k,iklj,cf,a->ba'
-shapes = [(8, 2, 5), (5, 7), (8, 8, 2, 5), (8, 6, 3), (8,), (6,), (5,), (6, 5, 5, 7), (6, 3), (3,)]
-print("Andy: expr =",expr)
-if rank == root:
-    operands = [cp.random.rand(*shape) for shape in shapes]
-else:
-    operands = [cp.empty(shape) for shape in shapes]
-'''
-
 myconvertor = QiboCircuitToEinsum(qibo_circ, dtype=datatype)
-expr, mode_label, q_frontier, operands = myconvertor.state_vector()
-shapes = [tensor.shape for tensor in operands]
-print("expr ", expr)
-print("Operands ", operands)
-print("Shape", shapes)
-# Set the operand data on root. Since we use the buffer interface APIs offered by mpi4py for communicating array
-#  objects, we can directly use device arrays (cupy.ndarray, for example) as we assume mpi4py is built against
-#  a CUDA-aware MPI.
-if rank != root:
-    operands = [cp.empty(shape,dtype="complex128") for shape in shapes]
-
-'''    
-if rank == root:
-    operands = [cp.random.rand(*shape) for shape in shapes]
-    print("Operands random", operands)
-
-else:
-    operands = [cp.empty(shape) for shape in shapes]
-'''
-
-for operand in operands:
-    print("Is CUPY array? ", cp.get_array_module(operand), " Operand size = ", operand.nbytes)
-
-for operand in operands:
-    comm.Bcast(operand, root)
 
 # Bind the communicator to the library handle
 handle = cutn.create()
-print("Andy cutn.create()")
-print("Andy ", cutn.get_mpi_comm_pointer(comm))
 cutn.distributed_reset_configuration(
     handle, *cutn.get_mpi_comm_pointer(comm)
 )
-print("Andy cutn.distributed_reset_configuration")
 
-operands_interleave = myconvertor.get_interleave_format( mode_label, q_frontier, operands)
-print("new function interkeave ", operands_interleave)
-print("Ori function interleave", myconvertor.state_vector_operands())
-
-result = cuquantum.contract(*operands_interleave, options={'device_id' : device_id, 'handle': handle})
-#result = cuquantum.contract(expr, *operands, options={'device_id' : device_id, 'handle': handle})
-
-'''
-
-# Create a new GPU buffer for verification
-result_cp = cp.empty_like(result)
-
-# Sum the partial contribution from each process on root, with GPU
 if rank == root:
-    comm.Reduce(sendbuf=MPI.IN_PLACE, recvbuf=result_cp, op=MPI.SUM, root=root)
+    start = timer()
-else:
+    
-    comm.Reduce(sendbuf=result_cp, recvbuf=None, op=MPI.SUM, root=root)
+result = cuquantum.contract(*myconvertor.state_vector_operands(), options={'device_id' : device_id, 'handle': handle})
-'''
+
+if rank == root:
+    end = timer()
+
 # Check correctness.
 if rank == root:
-    #operands = myconvertor.state_vector_operands()
+    #(qibo_circ, result_sv) =  qibo_qft(nqubits, swaps=True)
-    #result_cp = cp.einsum(*operands, optimize=True)
+    time = end - start
-    #result_cp = np.einsum(*operands, optimize=True)
+    #print("Does the cuQuantum parallel contraction result match the cupy.einsum result?", cp.allclose(result.flatten(), result_sv))
-    (qibo_circ, result_sv) =  qibo_qft(nqubits, swaps=True)
+    print("nqubit", nqubits, "time taken = ", time, 's')
-    print("Does the cuQuantum parallel contraction result match the cupy.einsum result?", cp.allclose(result.flatten(), result_sv))
+    
-
-
-'''
-result_tn = cutn.eval(qibo_circ, datatype)
-
-qibo.set_backend(backend="qibojit", platform="numpy")
-(qibo_circ, result_sv) = qibo_qft(nqubits, swaps=True)
-#print(result_tn)
-#print(result_sv)
-
-assert np.allclose(
-        result_sv, result_tn.flatten()), "Resulting dense vectors do not match"
-'''