summaryrefslogtreecommitdiff
path: root/linear.c
blob: 030fe3b17c869bf39c05a830a63add22f177dac4 (plain)
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
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <math.h>
#include <limits.h>
#include <time.h>

#include "nmglobal.h"
#include "randnum.h"


/***********************
**  LU DECOMPOSITION  **
** (Linear Equations) **
************************
** These routines come from "Numerical Recipes in Pascal".
** Note that, as in the assignment algorithm, though we
** separately define LUARRAYROWS and LUARRAYCOLS, the two
** must be the same value (this routine depends on a square
** matrix).
*/

/*
** DEFINES
*/

#define LUARRAYROWS 101L
#define LUARRAYCOLS 101L

/*
** TYPEDEFS
*/
typedef struct
{       union
	{       double *p;
		double (*ap)[][LUARRAYCOLS];
	} ptrs;
} LUdblptr;

/*
** GLOBALS
*/
double *LUtempvv;

/*
** PROTOTYPES
*/
static clock_t DoLUIteration(double *a, double *b,
	double *abase, double *bbase,
	unsigned long numarrays);
static void build_problem( double a[][LUARRAYCOLS],
	int n, double b[LUARRAYROWS]);
static int ludcmp(double a[][LUARRAYCOLS],
	int n, int indx[], int *d);
static void lubksb(double a[][LUARRAYCOLS],
	int n, int indx[LUARRAYROWS],
	double b[LUARRAYROWS]);
static int lusolve(double a[][LUARRAYCOLS],
	int n, double b[LUARRAYROWS]);

/*********
** DoLU **
**********
** Perform the LU decomposition benchmark.
*/
void
DoLU(void)
{
        const char*     context = "FPU:LU";
        LUStruct*       loclustruct = &global_lustruct;
        clock_t         total_time = 0;
        int             iterations = 0;
        double*         a = NULL;
        double*         b = NULL;
        double*         abase = NULL;
        double*         bbase = NULL;
        LUdblptr        ptra;
        int             n;
        int             i;

        /*
        ** Our first step is to build a "solvable" problem.  This
        ** will become the "seed" set that all others will be
        ** derived from. (I.E., we'll simply copy these arrays
        ** into the others.
        */
        a = malloc(sizeof(double) * LUARRAYCOLS * LUARRAYROWS);
        b = malloc(sizeof(double) * LUARRAYROWS);
        n=LUARRAYROWS;

        /*
        ** We need to allocate a temp vector that is used by the LU
        ** algorithm.  This removes the allocation routine from the
        ** timing.
        */
        LUtempvv = malloc(sizeof(double)*LUARRAYROWS);

        /*
        ** Build a problem to be solved.
        */
        ptra.ptrs.p=a;                  /* Gotta coerce linear array to 2D array */
        build_problem(*ptra.ptrs.ap,n,b);

        /*
        ** Now that we have a problem built, see if we need to do
        ** auto-adjust.  If so, repeatedly call the DoLUIteration routine,
        ** increasing the number of solutions per iteration as you go.
        */
        if (loclustruct->adjust == FALSE) {
                loclustruct->adjust = TRUE;

                loclustruct->numarrays=0;
                for(i=1;i<=MAXLUARRAYS;i++)
                {
                        abase = realloc(abase, sizeof(double) * LUARRAYCOLS*LUARRAYROWS*(i+1));
                        if (!abase) {
                                fprintf(stderr, "Error in %s, could not allocate memory. Exitting...\n", context);
                                free(a);
                                free(b);
                                free(LUtempvv);
                                exit(1);
                        }

                        bbase = realloc(bbase, sizeof(double) * LUARRAYROWS*(i+1));
                        if (!bbase) {
                                fprintf(stderr, "Error in %s, could not allocate memory. Exitting...\n", context);
                                free(a);
                                free(b);
                                free(abase);
                                free(LUtempvv);
                                exit(1);
                        }
                        if(DoLUIteration(a,b,abase,bbase,i)>global_min_ticks)
                        {       loclustruct->numarrays=i;
                                break;
                        }
                }
                /*
                ** Were we able to do it?
                */
                if (loclustruct->numarrays==0) {
                        fprintf(stderr, "FPU:LU -- Array limit reached\n");
                        free(a);
                        free(b);
                        free(abase);
                        free(bbase);
                        free(LUtempvv);
                        exit(1);
                }
        } else {
                /*
                ** Don't need to adjust -- just allocate the proper
                ** number of arrays and proceed.
                */
                abase = malloc(sizeof(double) *	LUARRAYCOLS*LUARRAYROWS*loclustruct->numarrays);
                if (!abase) {
                        fprintf(stderr, "Error in %s, could not allocate memory. Exitting...\n", context);
                        free(a);
                        free(b);
                        free(LUtempvv);
                        exit(1);
                }

                bbase = malloc(sizeof(double) *	LUARRAYROWS*loclustruct->numarrays);
                if (!bbase) {
                        fprintf(stderr, "Error in %s, could not allocate memory. Exitting...\n", context);
                        free(a);
                        free(b);
                        free(abase);
                        free(LUtempvv);
                        exit(1);
                }
        }

        do {
                total_time += DoLUIteration(a, b, abase, bbase, loclustruct->numarrays);
                iterations += loclustruct->numarrays;
        } while(total_time < loclustruct->request_secs * CLOCKS_PER_SEC);

        free(a);
        free(b);
        free(abase);
        free(bbase);
        free(LUtempvv);

        loclustruct->iterspersec = (double)(iterations * CLOCKS_PER_SEC) / (double)total_time;
}

/******************
** DoLUIteration **
*******************
** Perform an iteration of the LU decomposition benchmark.
** An iteration refers to the repeated solution of several
** identical matrices.
*/
static clock_t
DoLUIteration(double *a,double *b, double *abase, double *bbase, unsigned long numarrays)
{
        clock_t start, stop;
double *locabase;
double *locbbase;
LUdblptr ptra;  /* For converting ptr to 2D array */
unsigned long j,i;              /* Indexes */


/*
** Move the seed arrays (a & b) into the destination
** arrays;
*/
for(j=0;j<numarrays;j++)
{       locabase=abase+j*LUARRAYROWS*LUARRAYCOLS;
	locbbase=bbase+j*LUARRAYROWS;
	for(i=0;i<LUARRAYROWS*LUARRAYCOLS;i++)
		*(locabase+i)=*(a+i);
	for(i=0;i<LUARRAYROWS;i++)
		*(locbbase+i)=*(b+i);
}

        start = clock();

for(i=0;i<numarrays;i++)
{       locabase=abase+i*LUARRAYROWS*LUARRAYCOLS;
	locbbase=bbase+i*LUARRAYROWS;
	ptra.ptrs.p=locabase;
	lusolve(*ptra.ptrs.ap,LUARRAYROWS,locbbase);
}

        stop = clock();

        return stop - start;
}

/******************
** build_problem **
*******************
** Constructs a solvable set of linear equations.  It does this by
** creating an identity matrix, then loading the solution vector
** with random numbers.  After that, the identity matrix and
** solution vector are randomly "scrambled".  Scrambling is
** done by (a) randomly selecting a row and multiplying that
** row by a random number and (b) adding one randomly-selected
** row to another.
*/
static void build_problem(double a[][LUARRAYCOLS],
		int n,
		double b[LUARRAYROWS])
{
long i,j,k,k1;  /* Indexes */
double rcon;     /* Random constant */

/*
** Reset random number generator
*/
/* randnum(13L); */
randnum((int32_t)13);

/*
** Build an identity matrix.
** We'll also use this as a chance to load the solution
** vector.
*/
for(i=0;i<n;i++)
{       /* b[i]=(double)(abs_randwc(100L)+1L); */
	b[i]=(double)(abs_randwc((int32_t)100)+(int32_t)1);
	for(j=0;j<n;j++)
		if(i==j)
		        /* a[i][j]=(double)(abs_randwc(1000L)+1L); */
			a[i][j]=(double)(abs_randwc((int32_t)1000)+(int32_t)1);
		else
			a[i][j]=(double)0.0;
}

#ifdef DEBUG
printf("Problem:\n");
for(i=0;i<n;i++)
{
/*
	for(j=0;j<n;j++)
		printf("%6.2f ",a[i][j]);
*/
	printf("%.0f/%.0f=%.2f\t",b[i],a[i][i],b[i]/a[i][i]);
/*
        printf("\n");
*/
}
#endif

/*
** Scramble.  Do this 8n times.  See comment above for
** a description of the scrambling process.
*/

for(i=0;i<8*n;i++)
{
	/*
	** Pick a row and a random constant.  Multiply
	** all elements in the row by the constant.
	*/
 /*       k=abs_randwc((long)n);
	rcon=(double)(abs_randwc(20L)+1L);
	for(j=0;j<n;j++)
		a[k][j]=a[k][j]*rcon;
	b[k]=b[k]*rcon;
*/
	/*
	** Pick two random rows and add second to
	** first.  Note that we also occasionally multiply
	** by minus 1 so that we get a subtraction operation.
	*/
        /* k=abs_randwc((long)n); */
        /* k1=abs_randwc((long)n); */
	k=abs_randwc((int32_t)n);
	k1=abs_randwc((int32_t)n);
	if(k!=k1)
	{
		if(k<k1) rcon=(double)1.0;
			else rcon=(double)-1.0;
		for(j=0;j<n;j++)
			a[k][j]+=a[k1][j]*rcon;;
		b[k]+=b[k1]*rcon;
	}
}

return;
}


/***********
** ludcmp **
************
** From the procedure of the same name in "Numerical Recipes in Pascal",
** by Press, Flannery, Tukolsky, and Vetterling.
** Given an nxn matrix a[], this routine replaces it by the LU
** decomposition of a rowwise permutation of itself.  a[] and n
** are input.  a[] is output, modified as follows:
**   --                       --
**  |  b(1,1) b(1,2) b(1,3)...  |
**  |  a(2,1) b(2,2) b(2,3)...  |
**  |  a(3,1) a(3,2) b(3,3)...  |
**  |  a(4,1) a(4,2) a(4,3)...  |
**  |  ...                      |
**   --                        --
**
** Where the b(i,j) elements form the upper triangular matrix of the
** LU decomposition, and the a(i,j) elements form the lower triangular
** elements.  The LU decomposition is calculated so that we don't
** need to store the a(i,i) elements (which would have laid along the
** diagonal and would have all been 1).
**
** indx[] is an output vector that records the row permutation
** effected by the partial pivoting; d is output as +/-1 depending
** on whether the number of row interchanges was even or odd,
** respectively.
** Returns 0 if matrix singular, else returns 1.
*/
static int ludcmp(double a[][LUARRAYCOLS],
		int n,
		int indx[],
		int *d)
{

double big;     /* Holds largest element value */
double sum;
double dum;     /* Holds dummy value */
int i,j,k;      /* Indexes */
int imax=0;     /* Holds max index value */
double tiny;    /* A really small number */

tiny=(double)1.0e-20;

*d=1;           /* No interchanges yet */

for(i=0;i<n;i++)
{       big=(double)0.0;
	for(j=0;j<n;j++)
		if((double)fabs(a[i][j]) > big)
			big=fabs(a[i][j]);
	/* Bail out on singular matrix */
	if(big==(double)0.0) return(0);
	LUtempvv[i]=1.0/big;
}

/*
** Crout's algorithm...loop over columns.
*/
for(j=0;j<n;j++)
{       if(j!=0)
		for(i=0;i<j;i++)
		{       sum=a[i][j];
			if(i!=0)
				for(k=0;k<i;k++)
					sum-=(a[i][k]*a[k][j]);
			a[i][j]=sum;
		}
	big=(double)0.0;
	for(i=j;i<n;i++)
	{       sum=a[i][j];
		if(j!=0)
			for(k=0;k<j;k++)
				sum-=a[i][k]*a[k][j];
		a[i][j]=sum;
		dum=LUtempvv[i]*fabs(sum);
		if(dum>=big)
		{       big=dum;
			imax=i;
		}
	}
	if(j!=imax)             /* Interchange rows if necessary */
	{       for(k=0;k<n;k++)
		{       dum=a[imax][k];
			a[imax][k]=a[j][k];
			a[j][k]=dum;
		}
		*d=-*d;         /* Change parity of d */
		dum=LUtempvv[imax];
		LUtempvv[imax]=LUtempvv[j]; /* Don't forget scale factor */
		LUtempvv[j]=dum;
	}
	indx[j]=imax;
	/*
	** If the pivot element is zero, the matrix is singular
	** (at least as far as the precision of the machine
	** is concerned.)  We'll take the original author's
	** recommendation and replace 0.0 with "tiny".
	*/
	if(a[j][j]==(double)0.0)
		a[j][j]=tiny;

	if(j!=(n-1))
	{       dum=1.0/a[j][j];
		for(i=j+1;i<n;i++)
			a[i][j]=a[i][j]*dum;
	}
}

return(1);
}

/***********
** lubksb **
************
** Also from "Numerical Recipes in Pascal".
** This routine solves the set of n linear equations A X = B.
** Here, a[][] is input, not as the matrix A, but as its
** LU decomposition, created by the routine ludcmp().
** Indx[] is input as the permutation vector returned by ludcmp().
**  b[] is input as the right-hand side an returns the
** solution vector X.
** a[], n, and indx are not modified by this routine and
** can be left in place for different values of b[].
** This routine takes into account the possibility that b will
** begin with many zero elements, so it is efficient for use in
** matrix inversion.
*/
static void lubksb( double a[][LUARRAYCOLS],
		int n,
		int indx[LUARRAYROWS],
		double b[LUARRAYROWS])
{

int i,j;        /* Indexes */
int ip;         /* "pointer" into indx */
int ii;
double sum;

/*
** When ii is set to a positive value, it will become
** the index of the first nonvanishing element of b[].
** We now do the forward substitution. The only wrinkle
** is to unscramble the permutation as we go.
*/
ii=-1;
for(i=0;i<n;i++)
{       ip=indx[i];
	sum=b[ip];
	b[ip]=b[i];
	if(ii!=-1)
		for(j=ii;j<i;j++)
			sum=sum-a[i][j]*b[j];
	else
		/*
		** If a nonzero element is encountered, we have
		** to do the sums in the loop above.
		*/
		if(sum!=(double)0.0)
			ii=i;
	b[i]=sum;
}
/*
** Do backsubstitution
*/
for(i=(n-1);i>=0;i--)
{
	sum=b[i];
	if(i!=(n-1))
		for(j=(i+1);j<n;j++)
			sum=sum-a[i][j]*b[j];
	b[i]=sum/a[i][i];
}
return;
}

/************
** lusolve **
*************
** Solve a linear set of equations: A x = b
** Original matrix A will be destroyed by this operation.
** Returns 0 if matrix is singular, 1 otherwise.
*/
static int lusolve(double a[][LUARRAYCOLS],
		int n,
		double b[LUARRAYROWS])
{
int indx[LUARRAYROWS];
int d;
#ifdef DEBUG
int i,j;
#endif

if(ludcmp(a,n,indx,&d)==0) return(0);

/* Matrix not singular -- proceed */
lubksb(a,n,indx,b);

#ifdef DEBUG
printf("Solution:\n");
for(i=0;i<n;i++)
{
  for(j=0;j<n;j++){
  /*
    printf("%6.2f ",a[i][j]);
  */
  }
  printf("%6.2f\t",b[i]);
  /*
    printf("\n");
  */
}
printf("\n");
#endif

return(1);
}