PySparse

PySparse

ll_mat.c

In short

Last update: June 3rd, 2015

The next table gives a rough idea of what part of PySparse has been implemented in CySparse. Don't forget that CySparse is much more general and that other methods/functionalities were added to LLSparseMatrix.

Roughly speaking, there are 42 interesting routines defined in ll_mat.c, of which 33 are implemented (or existing code must be adapated to the new version with types).

This is roughly 79%

• col_scale: 8e3ed9bf93bd568264456af82b8321de82b39d3d
• row_scale: 1fa447c71c70d62160fd6198b19529a7383ee848
• norm: 9873423e4e88d1021f4498accf26ae041bc14f1d
• generalize: 18090304d9dc9425d9daa54ccc5af28f3f2e73cd
• scale: 5f78e6b80e2e62ece0122821c407b1fc365c424c
• take_triplet: 00f83e9b85374ed0fb9d4d47030af85a2f3272ba
• shift: 8ebcd28553a32a0d3b7708dc4683ed97f9f9aace
• copy: 4744a3b3fc15ca403c43d6cd8eecb80d7b071230
• clear_submatrix+ delete_rows: 525e50e72d34ac296dacdc3c4c301de2fd7e54ae
• to_csr: a4114859c97f1e82393c919843c85a02c42b9469

Table

Function ll_mat.c	Function ll_mat.pyx	Comments	Done
PysparseIterator	-	Not needed for the moment	🆗
PysparseIterator_*	-	"	🆗
SpMatrix_LLMatBuildColIndex	-	We use CSC and CSR	🆗
SpMatrix_LLMatDestroyColIndex	-	"	🆗
SpMatrix_LLMatGetItem	at, safe_at, __getitem__	Improved version + bound testing	✅
SpMatrix_LLMatSetItem	put, safe_put, __setitem__	Same + bound testing	✅
SpMatrix_LLMatUpdateItemAdd	update_ll_mat_item_ add_@index@_@type@	Idem	✅
LLMat_Compress	compress	Idem	✅
create_indexlist	create_c_ array_ indices_from_ python_ object_@index@	More elaborate with arrays	✅
copySubMatrix_FromList	-	Not needed, we use views	🆗
getSubMatrix_FromList	-	Not needed, we use views	🆗
clear_submatrix	clear_submatrix	Idem. PySparse has never exported this method...	✅
setSubMatrix_FromList	-	Not needed, we use views	🆗
ll_matvec_kernel	multiply_ ll_mat_ with_ numpy_vector_ kernel_@index@ _@type@	Idem	✅
ll_matvec_kernel_stride	multiply_ ll_mat_ with_ strided_ numpy_vector_ kernel_@index@ _@type@	Idem	✅
ll_matvec_kernel_sym	multiply_sym_ ll_mat_ with_ numpy_vector_ kernel_@index@_ @type@	Idem	✅
ll_matvec_kernel_stride_sym	multiply_ ll_mat_ with_ strided_ numpy_vector_ kernel_@index@_ @type@	Idem	✅
ll_matvec_transp_kernel	multiply_tranposed_ ll_mat_ with_ numpy_vector_ kernel_@index@_ @type@	Idem	✅
ll_matvec_transp_kernel_stride	multiply_tranposed_ ll_mat_ with_ with_ strided_ numpy_vector_ kernel_@index@_ @type@	Idem	✅
LLMat_matvec_transp	matvec_transp, A.T * b	Idem + another method	✅
LLMat_matvec	matvec, A * b	Idem + another method	✅
LLMat_col_scale	col_scale	This is one place where SpMatrix_ LLMat Build Col Index might be handy	✅
LLMat_row_scale	row_scale		✅
LLMat_to_csr	to_csr	Idem except column indices are not sorted	✅
LLMat_to_sss	-	Not needed	🆗
LLMat_generalize	generalize	Idem	✅
LLMat_export_mtx	-	was done before	🔸
LLMat_copy	copy	use of memcpy	✅
LLMat_update_add_at	-		🔴
LLMat_norm	norm	Works for symmetric matrices	✅
LLMat_shift	shift	Idem	✅
LLMat_keys	keys	Idem	✅
LLMat_values	values	Idem	✅
LLMat_items	items	Idem	✅
LLMat_scale	scale, A *= b	Idem	✅
LLMat_update_add_mask	-		🔴
LLMat_update_add_mask_sym	-		🔴
LLMat_take	take_triplet	Somewhat different	✅
LLMat_put	put_triplet	Somewhat different	✅
LLMat_delete_rows	delete_rows_ with_mask	Only int8 Numpy array for mask	✅
LLMat_delete_cols	-		🔴
LLMat_delete_rowcols	-		🔴
LLMat_Find	find	Idem	✅
LLMatType_dealloc	__dealloc__		✅
LLMatType_print	print_to	Idem but not satisfied	🔸
LLMatType_getattr	use of properties and attributes directly	No needed	🆗
LLMat_length	-	Do we need this?	❓
LLMat_subscript	-	No needed , we use views	🆗
LLMat_ass_subscript	-	No needed, we use views	🆗
LLMat_as_mapping	-	No needed, we use views	🆗
LLMatType	LLSparseMatrix		🆗
SpMatrix_NewLLMatObject	__cinit__		🆗
LLMat_from_mtx	-	Done before	🔸
helper_getCArrayObject	-	No needed	🆗
LLMat_from_csr	-		🔴
LLMat_from_csc	-		🔴
LLMat_dot	multiply_ transposed_ ll_mat_ by_ ll_mat_ @index@_@type@, A.T * B, matdot_ transp	Idem	✅
LLMat_symdot	-		🔴
LLMat_parse_index	-	No needed	🆗