-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmtimesx.c
executable file
·1481 lines (1371 loc) · 54.3 KB
/
mtimesx.c
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
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*************************************************************************************
*
* MATLAB (R) is a trademark of The Mathworks (R) Corporation
*
* Function: mtimesx
* Filename: mtimesx.c
* Programmer: James Tursa
* Version: 1.40
* Date: October 4, 2010
* Copyright: (c) 2009, 2010 by James Tursa, All Rights Reserved
*
* This code uses the BSD License:
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the distribution
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*
* Requires: mtimesx_RealTimesReal.c
*
* mtimesx is a multiply function that utilizes BLAS calls and custom code for
* matrix-matrix, matrix-vector, vector-vector, scalar-vector, scalar-matrix,
* and scalar-array calculations. Full support is given for transpose, conjugate
* transpose, and conjugate operations. Operands can be single, double, or
* sparse double. Sparse double matrices have direct support for scalar-matrix
* operations ... all other computations involving sparse matrices are done
* with calls to the MATLAB intrinsic functions.
*
* Building:
*
* mtimesx is typically self building. That is, the first time you call mtimesx,
* the mtimesx.m file recognizes that the mex routine needs to be compiled and
* then the compilation and linkage to the BLAS library will happen automatically.
* If the automatic compilation does not work, then you can do it manually as
* follows:
* - Put all of the files in one directory that is on the MATLAB path (but don't
* put them in a toolbox directory)
* - Make that directory the current directory.
* - Create a string variable with the pathname and filename of the BLAS library
* file to use. e.g. for microsoft it may be something like the following:
*
* - Issue the following command at the MATLAB prompt.
* >> mex('mtimesx.c',lib_blas)
*
* And if you are on a linux machine then you may need to add the string
* '-DEFINEUNIX' to the mex command argument list.
*
* The usage is as follows (arguments in brackets [ ] are optional):
*
* Syntax
*
* M = mtimesx( [mode] )
* C = mtimesx(A [,transa] ,B [,transb] [,mode])
*
* Description
*
* mtimesx performs the matrix calculation C = op(A) * op(B), where:
* A = A single or double or sparse scalar, matrix, or array.
* B = A single or double or sparse scalar, matrix, or array.
* transa = A character indicating a pre-operation on A (optional)
* transb = A character indicating a pre-operation on B (optional)
* The pre-operation can be any of:
* 'N' or 'n' = No operation (the default if trans_ is missing)
* 'T' or 't' = Transpose
* 'C' or 'c' = Conjugate Transpose
* 'G' or 'g' = Conjugate (no transpose)
* mode = 'MATLAB' or 'SPEED' (sets mode for current and future calculations,
* case insensitive, optional)
* M is a string indicating the current calculation mode, before setting the new one.
* C is the result of the matrix multiply operation.
*
* mtimesx uses a combination of BLAS library routine calls and custom code
* to optimize the matrix or scalar multiplies with respect to speed
* without sacrificing accuracy. As a result, some calculations are done in
* a different way than the equivalent MATLAB calculation, resulting in
* slightly different answers. Although slightly different, the mtimesx
* results are just as accurate as the MATLAB results. The general matrix
* multiply calculation uses the same BLAS calls as MATLAB, so there is no
* difference in speed or results for this case. For the scalar * matrix,
* scalar * vector, scalar * array, and matrix * vector cases there
* generally *will* be a difference between mtimesx and the MATLAB built in
* mtimes function. Usually, mtimesx will be faster that the equivalent
* MATLAB function, but some of these cases are slightly slower.
*
* M = mtimesx returns a string with the current calculation mode. The string
* will either be 'MATLAB' or 'SPEED' or one of the OMP modes.
*
* M = mtimesx(mode) sets the calculation mode to mode. The mode variable
* must be either the string 'MATLAB' or the string 'SPEED'. The return
* variable M is the previous mode setting prior to setting the new mode.
* The mode is case insensitive (lower or upper case is OK). You can also
* set one of the OMP modes if you have compiled with an OpenMP compiler.
*
* Mode:
*
* The 'MATLAB' mode uses the same sequence of BLAS calls that MATLAB does,
* and uses scalar multiply code that generates the same results as MATLAB.
* The purpose of this mode is to reproduce the exact same results as the
* same MATLAB operation. As such, there will often be no speed improvement
* by using mtimesx vs using the MATLAB intrinsic mtimes.
*
* The 'SPEED' mode uses different sequences of BLAS calls than MATLAB does
* whenever a speed improvement might be realized. Custom code for dot product
* calculations is used, and certain matrix-matrix and matrix-vector operations
* involving conjugates and/or transposes uses a series of dot product type
* calculations instead of calling BLAS routines if a speed improvement can
* be realized. This mode produces results that will sometimes differ slightly
* from the same MATLAB operation, but the results will still be accurate.
*
* The 'OMP' modes are basically the same as 'SPEED' mode except that some
* of the calculations are multi-threaded using OpenMP directives.
*
* Examples:
*
* C = mtimesx(A,B) % performs the calculation C = A * B
* C = mtimesx(A,'T',B) % performs the calculation C = A.' * B
* C = mtimesx(A,B,'G') % performs the calculation C = A * conj(B)
* C = mtimesx(A,'C',B,'C') % performs the calculation C = A' * B'
* mtimesx % returns the current calculation mode
* mtimesx('MATLAB') % sets calculation mode to match MATLAB
*
* Note: You cannot combine double sparse and single inputs, since MATLAB does not
* support a single sparse result. You also cannot combine sparse inputs with full
* nD (n > 2) inputs, since MATLAB does not support a sparse nD result. The only
* exception is a sparse scalar times an nD full array. In that special case,
* mtimesx will treat the sparse scalar as a full scalar and return a full nD result.
*
* Note: The ‘N’, ‘T’, and ‘C’ have the same meanings as the direct inputs to the BLAS
* routines. The ‘G’ input has no direct BLAS counterpart, but was relatively easy to
* implement in mtimesx and saves time (as opposed to computing conj(A) or conj(B)
* explicitly before calling mtimesx).
*
* mtimesx supports nD inputs. For these cases, the first two dimensions specify the
* matrix multiply involved. The remaining dimensions are duplicated and specify the
* number of individual matrix multiplies to perform for the result. i.e., mtimesx
* treats these cases as arrays of 2D matrices and performs the operation on the
* associated parings. For example:
*
* If A is (2,3,4,5) and B is (3,6,4,5), then
* mtimesx(A,B) would result in C(2,6,4,5)
* where C(:,:,i,j) = A(:,:,i,j) * B(:,:,i,j), i=1:4, j=1:5
*
* which would be equivalent to the MATLAB m-code:
* C = zeros(2,6,4,5);
* for m=1:4
* for n=1:5
* C(:,:,m,n) = A(:,:,m,n) * B(:,:,m,n);
* end
* end
*
* The first two dimensions must conform using the standard matrix multiply rules
* taking the transa and transb pre-operations into account, and dimensions 3:end
* must match exactly or be singleton (equal to 1). If a dimension is singleton
* then it is virtually expanded to the required size (i.e., equivalent to a
* repmat operation to get it to a conforming size but without the actual data
* copy). For example:
*
* If A is (2,3,4,5) and B is (3,6,1,5), then
* mtimesx(A,B) would result in C(2,6,4,5)
* where C(:,:,i,j) = A(:,:,i,j) * B(:,:,1,j), i=1:4, j=1:5
*
* which would be equivalent to the MATLAB m-code:
* C = zeros(2,6,4,5);
* for m=1:4
* for n=1:5
* C(:,:,m,n) = A(:,:,m,n) * B(:,:,1,n);
* end
* end
*
* When a transpose (or conjugate transpose) is involved, the first two dimensions
* are transposed in the multiply as you would expect. For example:
*
* If A is (3,2,4,5) and B is (3,6,4,5), then
* mtimesx(A,'C',B,'G') would result in C(2,6,4,5)
* where C(:,:,i,j) = A(:,:,i,j)' * conj( B(:,:,i,j) ), i=1:4, j=1:5
*
* which would be equivalent to the MATLAB m-code:
* C = zeros(2,6,4,5);
* for m=1:4
* for n=1:5
* C(:,:,m,n) = A(:,:,m,n)' * conj( B(:,:,m,n) );
* end
* end
*
* If A is a scalar (1,1) and B is (3,6,4,5), then
* mtimesx(A,'G',B,'C') would result in C(6,3,4,5)
* where C(:,:,i,j) = conj(A) * B(:,:,i,j)', i=1:4, j=1:5
*
* which would be equivalent to the MATLAB m-code:
* C = zeros(6,3,4,5);
* for m=1:4
* for n=1:5
* C(:,:,m,n) = conj(A) * B(:,:,m,n)';
* end
* end
*
* Change Log:
* 2009/Sep/27 --> 1.00, Initial Release
* 2009/Dec/10 --> 1.11, Fixed bug for empty transa & transb inputs
* 2010/Feb/23 --> 1.20, Fixed bug for dgemv and sgemv calls
* 2010/Aug/02 --> 1.30, Added (nD scalar) * (nD array) capability
* Replaced buggy mxRealloc with custom routine
* 2010/Oct/04 --> 1.40, Added OpenMP support for custom code
* Expanded sparse * single and sparse * nD support
* Fixed (nD complex scalar)C * (nD array) bug
*
****************************************************************************/
/* OpenMP ------------------------------------------------------------- */
#ifdef _OPENMP
#include <omp.h>
#endif
/* Includes ----------------------------------------------------------- */
#include <string.h>
#include <stddef.h>
#include <ctype.h>
#include <math.h>
#include <time.h>
#include "mex.h"
/* Macros ------------------------------------------------------------- */
#ifndef MWSIZE_MAX
#define mwIndex int
#define mwSignedIndex int
#define mwSize int
#define mwSize_t size_t
#else
#define mwSize_t mwSize
#endif
#define METHOD_BLAS 0
#define METHOD_LOOPS 1
#define METHOD_LOOPS_OMP 2
#define MTIMESX_NOT_SET 0
#define MTIMESX_BLAS 1
#define MTIMESX_LOOPS 2
#define MTIMESX_LOOPS_OMP 3
#define MTIMESX_MATLAB 4
#define MTIMESX_SPEED 5
#define MTIMESX_SPEED_OMP 6
#define STRINGIZE(name) #name
#define TOKENSTRING(name) STRINGIZE(name)
#define DIRECTIVE_MAX 1000
#ifndef COMPILER
#define COMPILER (unknown)
#endif
/* Prototypes --------------------------------------------------------- */
mxArray *DoubleTimesDouble(mxArray *, char, mxArray *, char);
mxArray *FloatTimesFloat(mxArray *, char, mxArray *, char);
mxArray *mxCreateSharedDataCopy(const mxArray *pr);
char mxArrayToTrans(const mxArray *mx);
void *myRealloc(void *vp, mwSize_t n);
void mtimesx_logo(void);
mxArray *modestring(int);
/* Global Variables --------------------------------------------------- */
int mtimesx_mode = MTIMESX_MATLAB;
int max_threads = 0;
int threads_used = 0;
int debug = 0;
int debug_message = 0;
/* Functions ect for OpenMP implementations ------- */
#ifdef _OPENMP
#define OPENMP_ENABLED 1.0
/* Functions etc for non OpenMP implementations ----------------- */
/* The omp_get_num_procs() function is courtesy of Dirk-Jan Kroon */
/* and is based on his FEX submission maxNumCompThreads --------- */
#else
#define OPENMP_ENABLED 0.0
#define omp_get_wtick() 0.0
#define omp_get_wtime() ((double)clock()/((double)CLOCKS_PER_SEC))
#if defined(_WIN32) || defined(_WIN64)
#include <windows.h>
int omp_get_num_procs( ) {
SYSTEM_INFO sysinfo;
GetSystemInfo(&sysinfo);
return sysinfo.dwNumberOfProcessors;
}
#elif MACOS
#include <sys/param.h>
#include <sys/sysctl.h>
int omp_get_num_procs( ) {
int nm[2];
size_t len = 4;
uint32_t count;
nm[0] = CTL_HW; nm[1] = HW_AVAILCPU;
sysctl(nm, 2, &count, &len, NULL, 0);
if(count < 1) {
nm[1] = HW_NCPU;
sysctl(nm, 2, &count, &len, NULL, 0);
if(count < 1) { count = 1; }
}
return count;
}
#else
#include <unistd.h>
int omp_get_num_procs( ) {
return sysconf(_SC_NPROCESSORS_ONLN);
}
#endif
#endif
/*------------------------------------------------------------------------ */
/*------------------------------------------------------------------------ */
/*------------------------------------------------------------------------ */
/*------------------------------------------------------------------------ */
void mexFunction(int nlhs, mxArray *plhs[],int nrhs, const mxArray *prhs[])
{
mxArray *A, *B, *C, *Araw, *Braw;
mxArray *rhs[4];
char transa, transb;
char *directive, *cp;
char transstring[2] = "_";
int unsupported = 0;
int i, j, k, got_directive, nrhs_old = nrhs;
int mtimesx_mode_new, max_threads_new,
already_set_mode, already_set_debug, already_set_threads;
mxArray *ans;
double threads;
/*-------------------------------------------------------------------------
* Check for proper number of inputs and outputs
*------------------------------------------------------------------------- */
if( nlhs > 1 ) {
mexErrMsgTxt("Must have at most 1 output.");
}
/*-------------------------------------------------------------------------
* If no inputs, just return the current mode
*------------------------------------------------------------------------- */
if( nrhs == 0 ) {
plhs[0] = modestring(mtimesx_mode);
return;
}
/*-------------------------------------------------------------------------
* Find out if any input is a directive
*------------------------------------------------------------------------- */
i = 0;
mtimesx_mode_new = MTIMESX_NOT_SET;
max_threads_new = 0;
already_set_mode = 0;
already_set_debug = 0;
already_set_threads = 0;
while( i < nrhs ) {
if( mxIsChar(prhs[i]) ) {
if( mxGetNumberOfElements(prhs[i]) > DIRECTIVE_MAX ) {
mexErrMsgTxt("Unknown directive.");
} else if( cp = directive = mxArrayToString(prhs[i]) ) {
j = 0;
while( *cp ) {
if( *cp == '(' ) j++;
if( *cp == ')' ) j--;
if( j == 0 ) {
*cp = toupper( *cp );
}
cp++;
}
k = cp - directive;
got_directive = 1;
if( strcmp(directive,"MATLAB") == 0 ) {
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
already_set_mode = 1;
mtimesx_mode_new = MTIMESX_MATLAB;
} else if( strcmp(directive,"SPEED") == 0 ) {
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
already_set_mode = 1;
mtimesx_mode_new = MTIMESX_SPEED;
} else if( strcmp(directive,"BLAS") == 0 ) {
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
already_set_mode = 1;
mtimesx_mode_new = MTIMESX_BLAS;
} else if( strcmp(directive,"LOOPS") == 0 ) {
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
already_set_mode = 1;
mtimesx_mode_new = MTIMESX_LOOPS;
} else if( strcmp(directive,"LOGO") == 0 ) {
if( nrhs_old > 1 ) {
mexErrMsgTxt("Cannot combine LOGO directive with other arguments.");
}
mtimesx_logo();
return;
} else if( strcmp(directive,"HELP") == 0 ) {
if( nrhs_old > 1 ) {
mexErrMsgTxt("Cannot combine HELP directive with other arguments.");
}
mexEvalString("help mtimesx;");
return;
} else if( strcmp(directive,"DEBUG") == 0 ) {
if( already_set_debug ) {
mexErrMsgTxt("Cannot set DEBUG/NODEBUG twice in same call.");
}
already_set_debug = 1;
debug = 1;
} else if( strcmp(directive,"NODEBUG") == 0 ) {
if( already_set_debug ) {
mexErrMsgTxt("Cannot set DEBUG/NODEBUG twice in same call.");
}
already_set_debug = 1;
debug = 0;
} else if( strcmp(directive,"OMP_GET_NUM_PROCS") == 0 ||
strcmp(directive,"OMP_GET_NUM_PROCS()") == 0 ) {
if( nrhs > 1 ) {
mexErrMsgTxt("Cannot combine OMP_GET_NUM_PROCS directive with other arguments.");
}
plhs[0] = mxCreateDoubleScalar(omp_get_num_procs());
return;
} else if( strcmp(directive,"OMP_GET_MAX_THREADS") == 0 ||
strcmp(directive,"OMP_GET_MAX_THREADS()") == 0 ) {
if( nrhs > 1 ) {
mexErrMsgTxt("Cannot combine OMP_GET_MAX_THREADS directive with other arguments.");
}
plhs[0] = mxCreateDoubleScalar(max_threads);
return;
} else if( strcmp(directive,"OMP_GET_NUM_THREADS") == 0 ||
strcmp(directive,"OMP_GET_NUM_THREADS") == 0 ) {
if( nrhs > 1 ) {
mexErrMsgTxt("Cannot combine OMP_GET_NUM_THREADS directive with other arguments.");
}
plhs[0] = mxCreateDoubleScalar(threads_used);
return;
} else if( strcmp(directive,"COMPILER") == 0 ) {
if( nrhs > 1 ) {
mexErrMsgTxt("Cannot combine COMPILER directive with other arguments.");
}
plhs[0] = mxCreateString(TOKENSTRING(COMPILER));
return;
} else if( strcmp(directive,"OPENMP") == 0 ) {
if( nrhs > 1 ) {
mexErrMsgTxt("Cannot combine OPENMP directive with other arguments.");
}
plhs[0] = mxCreateDoubleScalar(OPENMP_ENABLED);
return;
} else if( strcmp(directive,"SPEEDOMP") == 0 ) {
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
already_set_mode = 1;
#ifdef _OPENMP
mtimesx_mode_new = MTIMESX_SPEED_OMP;
if( max_threads == 0 ) {
max_threads_new = omp_get_num_procs();
}
#else
mtimesx_mode_new = MTIMESX_SPEED;
#endif
} else if( strcmp(directive,"LOOPSOMP") == 0 ) {
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
already_set_mode = 1;
#ifdef _OPENMP
mtimesx_mode_new = MTIMESX_LOOPS_OMP;
if( max_threads == 0 ) {
max_threads_new = omp_get_num_procs();
}
#else
mtimesx_mode_new = MTIMESX_LOOPS;
#endif
} else if( strcmp(directive,"OMP_GET_WTICK") == 0 ||
strcmp(directive,"OMP_GET_WTICK()") == 0 ) {
if( nrhs > 1 ) {
mexErrMsgTxt("Cannot combine OMP_GET_WTICK directive with other arguments.");
}
plhs[0] = mxCreateDoubleScalar(omp_get_wtick());
return;
} else if( strcmp(directive,"OMP_GET_WTIME") == 0 ||
strcmp(directive,"OMP_GET_WTIME()") == 0 ) {
if( nrhs > 1 ) {
mexErrMsgTxt("Cannot combine OMP_GET_WTIME directive with other arguments.");
}
plhs[0] = mxCreateDoubleScalar(omp_get_wtime());
return;
} else if( strcmp(directive,"OMP") == 0 ) {
#ifdef _OPENMP
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
if( already_set_threads ) {
mexErrMsgTxt("Cannot set threads twice in same call.");
}
already_set_mode = 1;
already_set_threads = 1;
max_threads_new = omp_get_num_procs();
mtimesx_mode_new = MTIMESX_SPEED_OMP;
#else
if( already_set_mode ) {
mexErrMsgTxt("Cannot set mode twice in same call.");
}
already_set_mode = 1;
mtimesx_mode_new = MTIMESX_SPEED;
#endif
} else if( strcmp(directive,"OMP_SET_NUM_THREADS") == 0 ||
strcmp(directive,"OMP_SET_MAX_THREADS") == 0 ) {
if( already_set_threads ) {
mexErrMsgTxt("Cannot set threads twice in same call.");
}
already_set_threads = 1;
#ifdef _OPENMP
max_threads_new = omp_get_num_procs();
#else
max_threads_new = -1;
#endif
} else if( k > 20 && directive[19] == '(' ) {
directive[19] = '\0';
if( strcmp(directive,"OMP_SET_NUM_THREADS") == 0 ||
strcmp(directive,"OMP_SET_MAX_THREADS") == 0 ) {
if( already_set_threads ) {
mexErrMsgTxt("Cannot set threads twice in same call.");
}
already_set_threads = 1;
directive[ 8] = 'N';
directive[ 9] = 'U';
directive[10] = 'M';
directive[19] = '=';
cp = directive + 20;
j = 1;
while( *cp ) {
if( *cp == '(' ) {
j++;
} else if( *cp == ')' ) {
j--;
if( !j ) {
*cp = ';';
break;
}
}
cp++;
}
if( j ) {
mexErrMsgTxt("Expression or statement is incorrect--possibly unbalanced (");
}
if( mexEvalString(directive) ) {
mexErrMsgTxt("Unable to evaluate expression for number of threads");
}
ans = mexGetVariablePtr("caller","OMP_SET_NUM_THREADS");
if( ans == NULL ) {
mexErrMsgTxt("Unable to evaluate expression for number of threads");
}
j = threads = mxGetScalar(ans);
mexEvalString("clear OMP_SET_NUM_THREADS;");
if( j < 1 || j != threads ) {
mexErrMsgTxt("Number of threads must be a positive integer.");
}
if( j > omp_get_num_procs() ) j = omp_get_num_procs();
#ifdef _OPENMP
max_threads_new = j;
#else
max_threads_new = -1;
#endif
} else {
got_directive = 0;
}
} else {
got_directive = 0;
}
mxFree(directive);
} else {
mexErrMsgTxt("Error allocating memory for directive string");
}
if( got_directive ) {
for( j=i; j<nrhs-1; j++ ) {
prhs[j] = prhs[j+1];
}
nrhs--;
} else {
i++;
}
} else {
i++;
}
}
if( mtimesx_mode_new > 0 ) {
mtimesx_mode = mtimesx_mode_new;
}
if( max_threads_new > 0 ) {
max_threads = max_threads_new;
}
if( nrhs == 0 ) {
if( mtimesx_mode_new || !max_threads_new ) {
plhs[0] = modestring(mtimesx_mode);
} else {
plhs[0] = mxCreateDoubleScalar(max_threads);
}
return;
}
if( nrhs < 2 || nrhs > 4 ) {
mexErrMsgTxt("Must have 2 - 4 non-directive inputs for multiply function");
}
/*-------------------------------------------------------------------------
* Pick off the transpose character inputs. If they are missing or empty
* or blank, then use 'N' by default.
*------------------------------------------------------------------------- */
Araw = (mxArray *) prhs[0];
if( nrhs == 2 ) {
transa = 'N';
Braw = (mxArray *) prhs[1];
transb = 'N';
} else if( nrhs == 3 ) {
if( mxIsChar(prhs[1]) ) {
transa = mxArrayToTrans(prhs[1]);
Braw = (mxArray *) prhs[2];
transb = 'N';
} else if( mxIsChar(prhs[2]) ) {
transa = 'N';
Braw = (mxArray *) prhs[1];
transb = mxArrayToTrans(prhs[2]);
} else {
mexErrMsgTxt("2nd or 3rd input must be char.");
}
} else { /* nrhs == 4 */
if( mxIsChar(prhs[1]) && mxIsChar(prhs[3]) ) {
transa = mxArrayToTrans(prhs[1]);
Braw = (mxArray *) prhs[2];
transb = mxArrayToTrans(prhs[3]);
} else {
mexErrMsgTxt("2nd and 4th inputs must be char.");
}
}
/*-----------------------------------------------------------------------------
* Check for valid TRANS characters
*----------------------------------------------------------------------------- */
if( transa == 'n' ) transa = 'N';
if( transa == 't' ) transa = 'T';
if( transa == 'c' ) transa = 'C';
if( transa == 'g' ) transa = 'G';
if( transb == 'n' ) transb = 'N';
if( transb == 't' ) transb = 'T';
if( transb == 'c' ) transb = 'C';
if( transb == 'g' ) transb = 'G';
if( (transa != 'N' && transa != 'T' && transa != 'C' && transa != 'G') ||
(transb != 'N' && transb != 'T' && transb != 'C' && transb != 'G') ) {
mexErrMsgTxt("Invalid TRANS character. Expected N, T, C, or G.");
}
/*-----------------------------------------------------------------------------
* Check for proper input type and call the appropriate multiply routine.
* To be similar to MATLAB for mixed single-double operations, convert
* single inputs to double, do the calc, then convert back to single.
*----------------------------------------------------------------------------- */
/* Future Enhancement: Put in special code for scalar multiplies, do calc in double */
if( mxIsDouble(Araw) ) {
if( mxIsDouble(Braw) ) {
/*
if( mxIsSparse(Araw) && !mxIsSparse(Braw) &&
(mxGetNumberOfElements(Araw) != 1 || mxGetNumberOfDimensions(Braw) == 2) ) {
k = mexCallMATLAB(1, &B, 1, &Braw, "sparse");
plhs[0] = DoubleTimesDouble(Araw, transa, B, transb);
mxDestroyArray(B);
} else if( !mxIsSparse(Araw) && mxIsSparse(Braw) &&
(mxGetNumberOfElements(Braw) != 1 || mxGetNumberOfDimensions(Araw) == 2) ) {
k = mexCallMATLAB(1, &A, 1, &Araw, "sparse");
plhs[0] = DoubleTimesDouble(A, transa, Braw, transb);
mxDestroyArray(A);
} else {
plhs[0] = DoubleTimesDouble(Araw, transa, Braw, transb);
}
*/
if( mxIsSparse(Araw) && !mxIsSparse(Braw) &&
mxGetNumberOfElements(Araw) == 1 && mxGetNumberOfDimensions(Braw) == 2 ) {
k = mexCallMATLAB(1, &B, 1, &Braw, "sparse");
plhs[0] = DoubleTimesDouble(Araw, transa, B, transb);
mxDestroyArray(B);
} else if( !mxIsSparse(Araw) && mxIsSparse(Braw) &&
mxGetNumberOfElements(Braw) == 1 && mxGetNumberOfDimensions(Araw) == 2 ) {
k = mexCallMATLAB(1, &A, 1, &Araw, "sparse");
plhs[0] = DoubleTimesDouble(A, transa, Braw, transb);
mxDestroyArray(A);
} else {
plhs[0] = DoubleTimesDouble(Araw, transa, Braw, transb);
}
} else if( mxIsSingle(Braw) ) {
if( mxIsSparse(Araw) ) {
if( mxGetNumberOfElements(Araw) == 1 ) {
k = mexCallMATLAB(1, &C, 1, &Araw, "full");
k = mexCallMATLAB(1, &A, 1, &C, "single");
mxDestroyArray(C);
plhs[0] = FloatTimesFloat(A, transa, Braw, transb);
mxDestroyArray(A);
} else {
k = mexCallMATLAB(1, &B, 1, &Braw, "double");
plhs[0] = DoubleTimesDouble(Araw, transa, B, transb);
mxDestroyArray(B);
}
} else if( mxGetNumberOfElements(Araw) == 1 || mxGetNumberOfElements(Braw) == 1 ) {
k = mexCallMATLAB(1, &B, 1, &Braw, "double");
C = DoubleTimesDouble(Araw, transa, B, transb);
mxDestroyArray(B);
k = mexCallMATLAB(1, plhs, 1, &C, "single");
mxDestroyArray(C);
} else {
k = mexCallMATLAB(1, &A, 1, &Araw, "single");
plhs[0] = FloatTimesFloat(A, transa, Braw, transb);
mxDestroyArray(A);
}
} else {
unsupported = 1;
}
} else if( mxIsSingle(Araw) ) {
if( mxIsDouble(Braw) ) {
if( mxIsSparse(Braw) ) {
if( mxGetNumberOfElements(Braw) == 1 ) {
k = mexCallMATLAB(1, &C, 1, &Braw, "full");
k = mexCallMATLAB(1, &B, 1, &C, "single");
mxDestroyArray(C);
plhs[0] = FloatTimesFloat(Araw, transa, B, transb);
mxDestroyArray(B);
} else {
k = mexCallMATLAB(1, &A, 1, &Araw, "double");
plhs[0] = DoubleTimesDouble(A, transa, Braw, transb);
mxDestroyArray(A);
}
} else if( mxGetNumberOfElements(Araw) == 1 || mxGetNumberOfElements(Braw) == 1 ) {
k = mexCallMATLAB(1, &A, 1, &Araw, "double");
C = DoubleTimesDouble(A, transa, Braw, transb);
mxDestroyArray(A);
k = mexCallMATLAB(1, plhs, 1, &C, "single");
mxDestroyArray(C);
} else {
k = mexCallMATLAB(1, &B, 1, &Braw, "single");
plhs[0] = FloatTimesFloat(Araw, transa, B, transb);
mxDestroyArray(B);
}
} else if( mxIsSingle(Braw) ) {
plhs[0] = FloatTimesFloat(Araw, transa, Braw, transb);
} else {
unsupported = 1;
}
} else {
unsupported = 1;
}
if( unsupported ) {
if( debug ) {
mexPrintf("MTIMESX: Unsupported types ... calling MATLAB intrinsic function mtimes\n");
}
rhs[0] = Araw;
transstring[0] = transa;
rhs[1] = mxCreateString(transstring);
rhs[2] = Braw;
transstring[0] = transb;
rhs[3] = mxCreateString(transstring);
k = mexCallMATLAB(1, plhs, 4, rhs, "mtimesx_sparse");
mxDestroyArray(rhs[3]);
mxDestroyArray(rhs[1]);
threads_used = 0;
}
return;
}
/*-------------------------------------------------------------------------------
*
* Convert mxArray char input into trans character.
*
*------------------------------------------------------------------------------- */
char mxArrayToTrans(const mxArray *mx)
{
mwSize n;
char *cp;
char trans;
if( cp = mxArrayToString(mx) ) {
n = mxGetNumberOfElements(mx);
if( n == 0 ) {
trans = 'N'; /* Treat empty char strings the same as 'N' */
} else {
trans = *cp; /* Pick off only the first character */
}
mxFree(cp);
} else {
mexErrMsgTxt("Error allocating memory for trans string");
}
return trans;
}
/*-------------------------------------------------------------------------------
*
* Clean a real sparse matrix of zeros. The code is a bit ugly because it was
* highly optimized for speed.
*
*------------------------------------------------------------------------------- */
mwIndex spcleanreal(mxArray *mx)
{
mwSize n, nrow;
mwIndex *ir, *jc, *iz;
mwIndex j, x, y, diff;
double *pr, *pz;
n = mxGetN(mx);
pz = pr = mxGetData(mx);
ir = mxGetIr(mx);
jc = mxGetJc(mx);
diff = 0;
for( y=0; y<n; y++ ) {
nrow = jc[y+1] - jc[y];
for( x=0; x<nrow; x++ ) {
if( *pr == 0.0 ) {
iz = (ir += (pr - pz));
pz = pr;
j = y + 1;
goto copydata;
}
pr++;
}
}
return jc[n];
for( y=0; y<n; y++ ) {
j = y + 1;
nrow = (jc[j] - diff) - jc[y];
for( x=0; x<nrow; x++ ) {
if( *pr != 0.0 ) {
*pz++ = *pr;
*iz++ = *ir;
} else {
copydata:
diff++;
}
pr++;
ir++;
}
jc[j] -= diff;
}
return jc[n];
}
/*-------------------------------------------------------------------------------
*
* Clean a complex sparse matrix of zeros. The code is a bit ugly because it was
* highly optimized for speed.
*
*------------------------------------------------------------------------------- */
mwIndex spcleancomplex(mxArray *mx)
{
mwSize n, nrow;
mwIndex *ir, *jc, *iz;
mwIndex j, x, y, diff;
double *pr, *pz;
double *pi, *py;
n = mxGetN(mx);
pz = pr = mxGetData(mx);
py = pi = mxGetImagData(mx);
ir = mxGetIr(mx);
jc = mxGetJc(mx);
diff = 0;
for( y=0; y<n; y++ ) {
nrow = jc[y+1] - jc[y];
for( x=0; x<nrow; x++ ) {
if( *pr == 0.0 && *pi == 0.0 ) {
iz = (ir += (pr - pz));
pz = pr;
py = pi;
j = y + 1;
goto copydata;
}
pr++;
pi++;
}
}
return jc[n];
for( y=0; y<n; y++ ) {
j = y + 1;
nrow = (jc[j] - diff) - jc[y];
for( x=0; x<nrow; x++ ) {
if( *pr != 0.0 || *pi != 0.0 ) {
*pz++ = *pr;
*py++ = *pi;
*iz++ = *ir;
} else {
copydata:
diff++;
}
pr++;
pi++;
ir++;
}
jc[j] -= diff;
}
return jc[n];
}
/*------------------------------------------------------------------------------- */
mwIndex spclean(mxArray *mx)
{
if( mxIsComplex(mx) ) {
return spcleancomplex(mx);
} else {
return spcleanreal(mx);
}
}
/*-------------------------------------------------------------------------------
* myRealloc for use only in cases of decreasing a block size. Avoids using the
* buggy mxRealloc function for this case.
*------------------------------------------------------------------------------- */
void *myRealloc(void *vp, mwSize_t n)
{
void *xp;
if( xp = mxMalloc(n) ) { /* not really a necessary test for a mex routine */
memcpy(xp, vp, n);
mxFree(vp);
return xp;
} else {
return vp;
}
}
/*-------------------------------------------------------------------------------
* Logo.
*------------------------------------------------------------------------------- */
void mtimesx_logo(void)
{
mxArray *rhs[9];
double *x, *y, *z;
double r;
mwSize i, j;
/* fh = figure */
mexCallMATLAB(1, rhs, 0, NULL,"figure");
/* set(fh, 'Color', [1 1 1]) */
rhs[1] = mxCreateString("Color");
rhs[2] = mxCreateDoubleMatrix(1, 3, mxREAL);
x = mxGetPr(rhs[2]);
x[0] = x[1] = x[2] = 1.0;