def estimate_blocksize(A,efficiency=0.7):
"""Attempt to determine the blocksize of a sparse matrix
Returns a blocksize=(r,c) such that
- A.nnz / A.tobsr( (r,c) ).nnz > efficiency
"""
if not (isspmatrix_csr(A) or isspmatrix_csc(A)):
A = csr_matrix(A)
if A.nnz == 0:
return (1,1)
if not 0 < efficiency < 1.0:
raise ValueError,'efficiency must satisfy 0.0 < efficiency < 1.0'
high_efficiency = (1.0 + efficiency) / 2.0
nnz = float(A.nnz)
M,N = A.shape
if M % 2 == 0 and N % 2 == 0:
e22 = nnz / ( 4 * count_blocks(A,(2,2)) )
else:
e22 = 0.0
if M % 3 == 0 and N % 3 == 0:
e33 = nnz / ( 9 * count_blocks(A,(3,3)) )
else:
e33 = 0.0
if e22 > high_efficiency and e33 > high_efficiency:
e66 = nnz / ( 36 * count_blocks(A,(6,6)) )
if e66 > efficiency:
return (6,6)
else:
return (3,3)
else:
if M % 4 == 0 and N % 4 == 0:
e44 = nnz / ( 16 * count_blocks(A,(4,4)) )
else:
e44 = 0.0
if e44 > efficiency:
return (4,4)
elif e33 > efficiency:
return (3,3)
elif e22 > efficiency:
return (2,2)
else:
return (1,1)
def estimate_blocksize(A, efficiency=0.7):
"""Attempt to determine the blocksize of a sparse matrix
Returns a blocksize=(r,c) such that
- A.nnz / A.tobsr( (r,c) ).nnz > efficiency
"""
if not (isspmatrix_csr(A) or isspmatrix_csc(A)):
A = csr_matrix(A)
if A.nnz == 0:
return (1, 1)
if not 0 < efficiency < 1.0:
raise ValueError('efficiency must satisfy 0.0 < efficiency < 1.0')
high_efficiency = (1.0 + efficiency) / 2.0
nnz = float(A.nnz)
M, N = A.shape
if M % 2 == 0 and N % 2 == 0:
e22 = nnz / (4 * count_blocks(A, (2, 2)))
else:
e22 = 0.0
if M % 3 == 0 and N % 3 == 0:
e33 = nnz / (9 * count_blocks(A, (3, 3)))
else:
e33 = 0.0
if e22 > high_efficiency and e33 > high_efficiency:
e66 = nnz / (36 * count_blocks(A, (6, 6)))
if e66 > efficiency:
return (6, 6)
else:
return (3, 3)
else:
if M % 4 == 0 and N % 4 == 0:
e44 = nnz / (16 * count_blocks(A, (4, 4)))
else:
e44 = 0.0
if e44 > efficiency:
return (4, 4)
elif e33 > efficiency:
return (3, 3)
elif e22 > efficiency:
return (2, 2)
else:
return (1, 1)
def _mul_sparse_matrix(self, other):
M, K1 = self.shape
K2, N = other.shape
indptr = np.empty_like(self.indptr)
R, n = self.blocksize
# convert to this format
if isspmatrix_bsr(other):
C = other.blocksize[1]
else:
C = 1
from csr import isspmatrix_csr
if isspmatrix_csr(other) and n == 1:
other = other.tobsr(blocksize=(n, C), copy=False) # lightweight conversion
else:
other = other.tobsr(blocksize=(n, C))
csr_matmat_pass1(M // R, N // C, self.indptr, self.indices, other.indptr, other.indices, indptr)
bnnz = indptr[-1]
indices = np.empty(bnnz, dtype=np.intc)
data = np.empty(R * C * bnnz, dtype=upcast(self.dtype, other.dtype))
bsr_matmat_pass2(
M // R,
N // C,
R,
C,
n,
self.indptr,
self.indices,
np.ravel(self.data),
other.indptr,
other.indices,
np.ravel(other.data),
indptr,
indices,
data,
)
data = data.reshape(-1, R, C)
# TODO eliminate zeros
return bsr_matrix((data, indices, indptr), shape=(M, N), blocksize=(R, C))
def count_blocks(A,blocksize):
"""For a given blocksize=(r,c) count the number of occupied
blocks in a sparse matrix A
"""
r,c = blocksize
if r < 1 or c < 1:
raise ValueError,'r and c must be positive'
if isspmatrix_csr(A):
M,N = A.shape
return csr_count_blocks(M,N,r,c,A.indptr,A.indices)
elif isspmatrix_csc(A):
return count_blocks(A.T,(c,r))
else:
return count_blocks(csr_matrix(A),blocksize)
def count_blocks(A, blocksize):
"""For a given blocksize=(r,c) count the number of occupied
blocks in a sparse matrix A
"""
r, c = blocksize
if r < 1 or c < 1:
raise ValueError('r and c must be positive')
if isspmatrix_csr(A):
M, N = A.shape
return csr_count_blocks(M, N, r, c, A.indptr, A.indices)
elif isspmatrix_csc(A):
return count_blocks(A.T, (c, r))
else:
return count_blocks(csr_matrix(A), blocksize)
def _mul_sparse_matrix(self, other):
M, K1 = self.shape
K2, N = other.shape
indptr = np.empty_like(self.indptr)
R, n = self.blocksize
#convert to this format
if isspmatrix_bsr(other):
C = other.blocksize[1]
else:
C = 1
from csr import isspmatrix_csr
if isspmatrix_csr(other) and n == 1:
other = other.tobsr(blocksize=(n, C),
copy=False) #lightweight conversion
else:
other = other.tobsr(blocksize=(n, C))
csr_matmat_pass1( M/R, N/C, \
self.indptr, self.indices, \
other.indptr, other.indices, \
indptr)
bnnz = indptr[-1]
indices = np.empty(bnnz, dtype=np.intc)
data = np.empty(R * C * bnnz, dtype=upcast(self.dtype, other.dtype))
bsr_matmat_pass2( M/R, N/C, R, C, n, \
self.indptr, self.indices, np.ravel(self.data), \
other.indptr, other.indices, np.ravel(other.data), \
indptr, indices, data)
data = data.reshape(-1, R, C)
#TODO eliminate zeros
return bsr_matrix((data, indices, indptr),
shape=(M, N),
blocksize=(R, C))
