我正在研究一个numpy矩阵 adj ,该矩阵表示某些networkx图的邻接矩阵。当我按以下方式构造 adj 时:
adj = sparse.csr_matrix(nx.adjacency_matrix(graph), dtype='longdouble').todense()
然后运行adj = adj ** 2
,然后在htop
中可以看到numpy使用了所有可用线程。
但是,由于精度损失,我试图在两者之间的某个位置集成mpmath
。
我是这样做的:
mp.dps = 120
adj = sparse.csr_matrix(nx.adjacency_matrix(graph), dtype='longdouble').todense()
# ... just like before
adjmp = mp.matrix(adj)
# this casts all values to mpf
adj = np.matrix(adjmp, dtype=object)
# and get back the np matrix, now with mpfs inside
生成的 adj 看起来像这样
matrix([[mpf('0.0'), mpf('0.0'), mpf('0.0'), ..., mpf('0.0'), mpf('0.0'),
mpf('0.125')], # [...]
这是我的期望。
计算包括两个步骤:第一步是对 adj 求平方,第二步是实际计算。从结果中,我可以看到精度要高得多,但是htop
表明由于某种原因,平方步骤仅在一个线程上运行。
当我运行np.show_config()时,我得到:
blas_mkl_info:
NOT AVAILABLE
blis_info:
NOT AVAILABLE
openblas_info:
libraries = ['openblas', 'openblas']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None)]
blas_opt_info:
libraries = ['openblas', 'openblas']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None)]
lapack_mkl_info:
NOT AVAILABLE
openblas_lapack_info:
libraries = ['openblas', 'openblas']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None)]
lapack_opt_info:
libraries = ['openblas', 'openblas']
library_dirs = ['/usr/local/lib']
language = c
define_macros = [('HAVE_CBLAS', None)]