Skip to content

Instantly share code, notes, and snippets.

@mattjj
Last active May 26, 2016 22:53
Show Gist options
  • Save mattjj/ed5ac08786100833c691811d87f677a9 to your computer and use it in GitHub Desktop.
Save mattjj/ed5ac08786100833c691811d87f677a9 to your computer and use it in GitHub Desktop.
# distutils: extra_compile_args = -O2 -w
# cython: boundscheck=False, nonecheck = False, wraparound=False, cdivision=True
import numpy as np
cimport numpy as np
import operator as op
from cython.parallel import prange
from cython import boundscheck, nonecheck, wraparound
def csr_dot(rows, cols, B):
out = np.zeros_like(B)
_csr_dot(rows, cols, B, out)
return out
def to_sparse_format(dct):
rows, cols = zip(*sorted(dct.items(), key=op.itemgetter(0)))
rows = np.repeat(rows, map(len, cols))
cols = np.concatenate(cols)
return rows.astype('int32'), cols.astype('int32')
@nonecheck(False)
@wraparound(False)
@boundscheck(False)
cdef inline void _csr_dot(int[::1] rows, int[::1] cols, double[:,::1] B, double[:,::1] out):
cdef int idx, i, j, k
for idx in range(rows.shape[0]):
i = rows[idx]
k = cols[idx]
for j in range(B.shape[1]):
out[i,j] += B[k,j]
from __future__ import division
import numpy as np
import numpy.random as npr
from bindot import to_sparse_format, csr_dot
def dct_to_dense(dct):
rows, cols = to_sparse_format(dct)
out = np.zeros((1 + max(dct.keys()), 1 + max(max(row) for row in dct.values())))
for i, j in zip(rows, cols):
out[i,j] = 1
return out
if __name__ == "__main__":
npr.seed(0)
dct = {0: [0, 1, 2], 1: [1, 0, 3], 2: [2, 0, 3], 3: [3, 1, 2], 4: [4, 5], 5: [5, 4]}
B = npr.randn(6, 6)
rows, cols = to_sparse_format(dct)
print np.allclose(csr_dot(rows, cols, B), np.dot(dct_to_dense(dct), B))
from setuptools import setup
import numpy as np
from Cython.Build import cythonize
setup(
ext_modules=cythonize('**/*.pyx'),
include_dirs=[np.get_include(),],
)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment