1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
|
function test251(tasks)
%TEST251 test dot4 for plus-pair semirings
% GB_AxB_dot4 computes C+=A'*B when C is dense.
% SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2025, All Rights Reserved.
% SPDX-License-Identifier: Apache-2.0
% NOTE: test coverage should start with an empty JIT cache.
fprintf ('test251 ------------ C+=A''*B when C is dense (plus-pair)\n') ;
if (nargin < 1)
tasks = [ ] ;
end
if (isempty (tasks))
tasks = {
{ 0, 1, 1}, ... % ( 98, 98)
{ 1, 1, 1}, ... % ( 22, 120)
{ 2, 1, 1}, ... % ( 13, 133)
{ 3, 1, 1}, ... % ( 31, 164)
{ 4, 1, 1}, ... % ( 11, 175)
{ 5, 1, 1}, ... % ( 11, 186)
{ 6, 1, 1}, ... % ( 13, 199)
{ 7, 1, 1}, ... % ( 18, 217)
{ 8, 1, 1}, ... % ( 13, 230)
{ 9, 1, 1}, ... % ( 9, 239)
{10, 1, 1}, ... % ( 9, 248)
{11, 1, 1}, ... % ( 4, 252)
{12, 1, 1}, ... % ( 16, 268)
{13, 1, 1}, ... % ( 17, 285)
{ 0, 1, 2}, ... % ( 7, 292)
{ 0, 1, 4}, ... % ( 16, 308)
{ 1, 1, 4}, ... % ( 2, 310)
{ 2, 1, 4}, ... % ( 1, 311)
{ 3, 1, 4}, ... % ( 1, 312)
{ 4, 1, 4}, ... % ( 1, 313)
{ 5, 1, 4}, ... % ( 1, 314)
{ 6, 1, 4}, ... % ( 1, 315)
{ 7, 1, 4}, ... % ( 1, 316)
{ 8, 1, 4}, ... % ( 1, 317)
{ 9, 1, 4}, ... % ( 1, 318)
{10, 1, 4}, ... % ( 1, 319)
{11, 1, 4}, ... % ( 1, 320)
{12, 1, 4}, ... % ( 1, 321)
{13, 1, 4}, ... % ( 1, 322)
{ 0, 1, 8}, ... % ( 5, 327)
{ 1, 1, 8}, ... % ( 1, 328)
{ 2, 1, 8}, ... % ( 2, 330)
{ 4, 1, 8}, ... % ( 2, 332)
{ 5, 1, 8}, ... % ( 2, 334)
{ 0, 2, 1}, ... % ( 5, 339)
{ 0, 2, 2}, ... % ( 1, 340)
{ 0, 4, 1}, ... % ( 8, 348)
{ 1, 4, 1}, ... % ( 3, 351)
{ 2, 4, 1}, ... % ( 1, 352)
{ 3, 4, 1}, ... % ( 1, 353)
{ 4, 4, 1}, ... % ( 1, 354)
{ 5, 4, 1}, ... % ( 1, 355)
{ 6, 4, 1}, ... % ( 1, 356)
{ 7, 4, 1}, ... % ( 1, 357)
{ 8, 4, 1}, ... % ( 1, 358)
{ 9, 4, 1}, ... % ( 1, 359)
{10, 4, 1}, ... % ( 1, 360)
{11, 4, 1}, ... % ( 1, 361)
{12, 4, 1}, ... % ( 1, 362)
{13, 4, 1}, ... % ( 1, 363)
{ 1, 4, 2}, ... % ( 2, 365)
{ 0, 4, 4}, ... % ( 14, 379)
{ 2, 4, 4}, ... % ( 1, 380)
{ 3, 4, 4}, ... % ( 1, 381)
{ 4, 4, 4}, ... % ( 1, 382)
{ 5, 4, 4}, ... % ( 1, 383)
{ 6, 4, 4}, ... % ( 1, 384)
{ 7, 4, 4}, ... % ( 1, 385)
{ 8, 4, 4}, ... % ( 1, 386)
{ 9, 4, 4}, ... % ( 1, 387)
{10, 4, 4}, ... % ( 1, 388)
{11, 4, 4}, ... % ( 1, 389)
{12, 4, 4}, ... % ( 1, 390)
{13, 4, 4}, ... % ( 1, 391)
{ 0, 4, 8}, ... % ( 3, 394)
{ 0, 8, 1}, ... % ( 5, 399)
{ 1, 8, 1}, ... % ( 4, 403)
{ 2, 8, 1}, ... % ( 2, 405)
{ 4, 8, 1}, ... % ( 2, 407)
{ 5, 8, 1}, ... % ( 2, 409)
{ 0, 8, 2}, ... % ( 1, 410)
{ 1, 8, 2}, ... % ( 1, 411)
{ 0, 8, 4}, ... % ( 3, 414)
{ 0, 8, 8}, ... % ( 3, 417)
{ 1, 8, 8}, ... % ( 1, 418)
{ 2, 8, 8}, ... % ( 2, 420)
{ 4, 8, 8}, ... % ( 2, 422)
{ 5, 8, 8}, ... % ( 2, 424)
} ;
end
track_coverage = false ;
if (track_coverage)
global GraphBLAS_grbcov
track_coverage = ~isempty (GraphBLAS_grbcov) ;
clast = sum (GraphBLAS_grbcov > 0) ;
cfirst = clast ;
end
rng ('default') ;
plus_pair.add = 'plus' ;
plus_pair.multiply = 'oneb' ; % same as pair
[~, ~, ~, types, ~, ~, ~,] = GB_spec_opsall ;
types = types.all ;
add_op.opname = 'plus' ;
dtn_dot = struct ('axb', 'dot', 'inp0', 'tran') ;
dtn_sax = struct ('axb', 'saxpy', 'inp0', 'tran') ;
n = 20 ;
C = GB_spec_random (n, n, inf, 100, 'double') ;
C.sparsity = 8 ;
C0.matrix = sparse (n, n) ;
% create the test matrices
A_matrices = cell (8) ;
B_matrices = cell (8,8) ;
for A_sparsity = [1 2 4 8]
if (A_sparsity == 8)
A = GB_spec_random (n, n, inf, 100, 'double') ;
else
A = GB_spec_random (n, n, 0.1, 100, 'double') ;
end
A_matrices {A_sparsity} = A ;
for B_sparsity = [1 2 4 8]
if (B_sparsity == 8)
B = GB_spec_random (n, n, inf, 100, 'double') ;
else
B = GB_spec_random (n, n, 0.1, 100, 'double') ;
end
B_matrices {A_sparsity,B_sparsity} = B ;
end
end
for kk = 1:length(tasks)
task = tasks {kk} ;
k = task {1} ;
A_sparsity = task {2} ;
B_sparsity = task {3} ;
A = A_matrices {A_sparsity} ;
A.sparsity = A_sparsity ;
B = B_matrices {A_sparsity,B_sparsity} ;
B.sparsity = B_sparsity ;
if (k == 0)
type = 'logical' ;
add_op.opname = 'xor' ;
plus_pair.add = 'xor' ;
else
type = types {k} ;
add_op.opname = 'plus' ;
plus_pair.add = 'plus' ;
end
plus_pair.class = type ;
add_op.optype = type ;
if (test_contains (type, 'single'))
tol = 1e-5 ;
else
tol = 1e-10 ;
end
A.class = type ;
B.class = type ;
C0.class = type ;
C.class = type ;
% X = C + A'*B using dot4
X2 = GB_mex_mxm (C, [ ], add_op, plus_pair, A, B, dtn_dot) ;
X1 = GB_spec_mxm (C, [ ], add_op, plus_pair, A, B, dtn_dot) ;
GB_spec_compare (X1, X2, 0, tol) ;
% X = A'*B using dot2/dot3
X2 = GB_mex_mxm (C0, [ ], [ ], plus_pair, A, B, dtn_dot) ;
X1 = GB_spec_mxm (C0, [ ], [ ], plus_pair, A, B, dtn_dot) ;
GB_spec_compare (X1, X2, 0, tol) ;
% X = C + A'*B using saxpy
X2 = GB_mex_mxm (C, [ ], add_op, plus_pair, A, B, dtn_sax) ;
X1 = GB_spec_mxm (C, [ ], add_op, plus_pair, A, B, dtn_sax) ;
GB_spec_compare (X1, X2) ;
if (track_coverage)
c = sum (GraphBLAS_grbcov > 0) ;
d = c - clast ;
if (d > 0)
fprintf ('{%2d, %d, %d},', ...
k, A_sparsity, B_sparsity) ;
fprintf (' ... %% (%3d, %3d)\n', d, c-cfirst) ;
end
clast = c ;
else
fprintf ('.') ;
end
end
fprintf ('\n') ;
fprintf ('test251: all tests passed\n') ;
|
