Comparisons

Comparison benchmarks

⚠️ ALL benchmarks must be redone! ⚠️

What is it?

Simple comparisons between:

• CySparse (https://github.com/PythonOptimizers/cysparse);
• PySparse (http://pysparse.sourceforge.net/);
• SpPy (http://pythonhosted.org/sppy/);
• SciPy.sparse(http://docs.scipy.org/doc/scipy/reference/sparse.html);
• CVXOPT (http://cvxopt.org/);
• Spark (http://spark.apache.org/docs/latest/index.html);
• Theano (http://deeplearning.net/software/theano/);
• SymPy (https://github.com/sympy/sympy);
• PyCuda (?) (http://wiki.tiker.net/PyCuda);
• minieigen (https://github.com/eudoxos/minieigen)

See the *.bench.txt and the corresponding *.py files in performance/comparisons for more details.

Results

Legend

• CySparse : ✅
• PySparse : 🔴
• SpPy: 💛
• SciPy.sparse : 🔸
• Worth further investigation: ❓

More symbols means more performance gains between the winner and the second one.

Table LLSparseMatrix

What	Link	Contestants	Winner
A * x	matvec	✅ 🔴 🔸	-
A^t * x	matvec_transp		-
A[i,j] = x	simple element assignment		-
x = A[i,j]	simple element retrieval		✅
col_scale	col_scale		✅ ✅
col_scale strided	col_scale_strided		✅ ✅
row_scale	row_scale		✅
row_scale strided	row_scale_strided		✅
norm_inf	norm_inf		-
norm_frob	norm_frob		-
norm_frob sym	norm_frob_sym		-
norm_one	norm_one		✅
generalize	generalize		-
scale	scale		-
copy	copy		✅ ✅
take	take		✅
shift	shift		✅
matdot	matdot		- ❓
delete_rows	delete_rows		❓
update_add_at	update_add_at		✅ ✅
put	put	✅ 🔴 💛	✅
find	find		🔴
to_csr	to_csr		✅

Table CSRSparseMatrix

What	Link	Winner

Table CSCSparseMatrix

What	Link	Winner