workspace_gpu = gpuarray.zeros(lwork, dtype=A.dtype)
info = gpuarray.zeros(batchSize, dtype=np.int32)
# Compute:
solver.cusolverDnCheevjBatched(handle, 'CUSOLVER_EIG_MODE_VECTOR', 'u', n,
x_gpu.gpudata, n, w_gpu.gpudata,
workspace_gpu.gpudata, lwork, info.gpudata,
params, batchSize)
# Print info
tmp = info.get()
if any(tmp):
print "the following job did not converge:", np.nonzero(tmp)[0]
else:
print "all jobs converged"
# Destroy handle
solver.cusolverDnDestroySyevjInfo(params)
solver.cusolverDnDestroy(handle)
Q = x_gpu.get()
W = w_gpu.get()
print 'maximum error in A * Q - Q * Lambda is:'
for i in range(batchSize):
q = Q[i * n:(i + 1) * n, :].T.copy()
x = A[i * n:(i + 1) * n, :].copy()
w = W[i, :].copy()
print '{}th matrix'.format(i), np.abs(
np.dot(x, q) - np.dot(q, np.diag(w))).max()
lwork = solver.cusolverDnDsyevjBatched_bufferSize(handle, 'CUSOLVER_EIG_MODE_VECTOR',
'u', n, x_gpu.gpudata, n,
w_gpu.gpudata, params, batchSize)
workspace_gpu = gpuarray.zeros(lwork, dtype = A.dtype)
info = gpuarray.zeros(batchSize, dtype = np.int32)
# Compute:
solver.cusolverDnDsyevjBatched(handle, 'CUSOLVER_EIG_MODE_VECTOR',
'u', n, x_gpu.gpudata, n,
w_gpu.gpudata, workspace_gpu.gpudata,
lwork, info.gpudata, params, batchSize)
# print info
tmp = info.get()
if any(tmp):
print("the following job did not converge: %r" % np.nonzero(tmp)[0])
# Destroy handle
solver.cusolverDnDestroySyevjInfo(params)
solver.cusolverDnDestroy(handle)
Q = x_gpu.get()
W = w_gpu.get()
print('maximum error in A * Q - Q * Lambda is:')
for i in range(batchSize):
q = Q[i*n:(i+1)*n,:].T.copy()
x = A[i*n:(i+1)*n,:].copy()
w = W[i, :].copy()
print('{}th matrix %r'.format(i) % np.abs(np.dot(x, q) - np.dot(q, np.diag(w))).max())