This repository has been archived by the owner on Jul 26, 2024. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathtest_kronmult6_batched.cpp
574 lines (461 loc) · 18.6 KB
/
test_kronmult6_batched.cpp
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
#include <iostream>
#include <cassert>
#include <chrono>
#include <unistd.h>
#include "kroncommon.hpp"
#include "kronmult6_batched.hpp"
#include "kronmult5_batched.hpp"
#include "kronmult4_batched.hpp"
#include "kronmult3_batched.hpp"
#include "kronmult2_batched.hpp"
#include "kronmult1_batched.hpp"
#ifdef USE_GPU
#include <cuda_runtime.h>
#else
#include <stdlib.h>
#include <string.h>
#endif
static inline
void host2gpu( void *dest, void *src, size_t nbytes )
{
#ifdef USE_GPU
cudaError_t istat = cudaMemcpy( dest,
src,
nbytes,
cudaMemcpyHostToDevice );
expect( istat == cudaSuccess );
#else
memcpy( dest, src, nbytes );
#endif
}
static inline
void gpu2host( void *dest, void *src, size_t nbytes )
{
#ifdef USE_GPU
cudaError_t istat = cudaMemcpy( dest,
src,
nbytes,
cudaMemcpyDeviceToHost);
expect( istat == cudaSuccess );
#else
memcpy( dest, src, nbytes );
#endif
}
static inline
void *myalloc( size_t nbytes ) {
void *devPtr = nullptr;
#ifdef USE_GPU
cudaError_t istat = cudaMalloc( &devPtr, nbytes );
expect( istat == cudaSuccess );
#else
devPtr = malloc( nbytes );
#endif
expect( devPtr != nullptr );
return(devPtr);
}
static inline
void myfree( void * devPtr ) {
#ifdef USE_GPU
cudaError_t istat = cudaFree( devPtr);
expect( istat == cudaSuccess );
#else
free( devPtr );
#endif
}
template<typename T>
T test_kronmult_batched( int const idim,
int const n, int const batchCount,
int const idebug = 0,
bool const do_check = true )
{
// -------------------------
// Aarray is (n,n,idim,batchCount)
// Xarray is (n^idim by batchCount)
// Yarray is (n^idim by batchCount)
// Zarray is (n^idim by batchCount)
// Warray is (n^idim by batchCount)
// ----------------------------
int const Xsize = std::pow(n,idim);
T *Aarray_ = (T *) malloc( sizeof(T)*n*n*idim*batchCount);
T *Xarray_ = (T *) malloc( sizeof(T)*Xsize * batchCount);
T *Yarray_ = (T *) malloc( sizeof(T)*Xsize * batchCount);
T *Zarray_ = (T *) malloc( sizeof(T)*Xsize * batchCount);
T *Warray_ = (T *) malloc( sizeof(T)*Xsize * batchCount);
expect( Aarray_ != nullptr );
expect( Xarray_ != nullptr );
expect( Yarray_ != nullptr );
expect( Zarray_ != nullptr );
expect( Warray_ != nullptr );
T *dAarray_ = (T *) myalloc( sizeof(T)*n*n*idim*batchCount);
T *dXarray_ = (T *) myalloc( sizeof(T)*Xsize * batchCount );
T *dZarray_ = (T *) myalloc( sizeof(T)*Xsize * batchCount );
T *dYarray_ = (T *) myalloc( sizeof(T)*Xsize * batchCount );
T *dWarray_ = (T *) myalloc( sizeof(T)*Xsize * batchCount );
auto Aarray = [&] (int const i,
int const j,
int const k,
int const ibatch ) -> T& {
return( Aarray_[ indx4f(i,j,k,ibatch, n,n,idim) ] );
};
auto Xarray = [&] (int const i,
int const ibatch) -> T& {
return( Xarray_[ indx2f(i,ibatch,Xsize) ] );
};
auto Yarray = [&] (int const i,
int const ibatch) -> T& {
return( Yarray_[ indx2f(i,ibatch,Xsize) ] );
};
auto Zarray = [&] (int const i,
int const ibatch) -> T& {
return( Zarray_[ indx2f(i,ibatch,Xsize) ] );
};
auto Warray = [&] (int const i,
int const ibatch) -> T& {
return( Warray_[ indx2f(i,ibatch,Xsize) ] );
};
// ---------------------
// initialize the arrays
// save a copy of Xarray in Z
// ---------------------
#pragma omp parallel for
for(int ibatch=1; ibatch <= batchCount; ibatch++) {
for(int i=1; i <= Xsize; i++) {
T const r1 = (i + (ibatch-1)*Xsize );
T const r2 = Xsize*batchCount;
// --------------------------------
// note Zarray is a copy of Xarray
// --------------------------------
Xarray(i,ibatch) = r1/r2;
Zarray(i,ibatch) = Xarray(i,ibatch);
Yarray(i,ibatch) = 0;
Warray(i,ibatch) = 0;
};
};
#pragma omp parallel for
for(int ibatch=1; ibatch <= batchCount; ibatch++) {
for(int k=1; k <= idim; k++) {
for(int j=1; j <= n; j++) {
for(int i=1; i <= n; i++) {
T const r1 = i + (j-1)*n + (k-1)*n*n + (ibatch-1)*batchCount;
T const r2 = n*n*idim*batchCount;
Aarray(i,j,k, ibatch) = r1/r2;
};
};
};
};
// ---------------------
// copy from host to GPU
// interface is host2gpu( dest, src, nbytes )
// ---------------------
host2gpu( dAarray_, Aarray_, sizeof(T)*n*n*idim*batchCount );
host2gpu( dXarray_, Xarray_, sizeof(T)*Xsize*batchCount );
host2gpu( dYarray_, Yarray_, sizeof(T)*Xsize*batchCount );
host2gpu( dZarray_, Zarray_, sizeof(T)*Xsize*batchCount );
host2gpu( dWarray_, Warray_, sizeof(T)*Xsize*batchCount );
auto time_start = std::chrono::steady_clock::now();
#ifdef USE_GPU
{
int constexpr warpsize = 32;
int const nwarps = min(32,n);
int const nthreads = nwarps * warpsize;
// --------------------------------------------
// note the input Zarray will be over-written
// --------------------------------------------
switch(idim) {
case 1: kronmult1_batched<T><<< batchCount, nthreads >>>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 2: kronmult2_batched<T><<< batchCount, nthreads >>>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 3: kronmult3_batched<T><<< batchCount, nthreads >>>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 4: kronmult4_batched<T><<< batchCount, nthreads >>>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 5: kronmult5_batched<T><<< batchCount, nthreads >>>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 6: kronmult6_batched<T><<< batchCount, nthreads >>>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
default:
expect( false );
};
// -------------------------------------------
// note important to wait for kernel to finish
// -------------------------------------------
cudaError_t istat = cudaDeviceSynchronize();
expect( istat == cudaSuccess );
}
#else
{
// --------------------------------------------
// note the input Zarray will be over-written
// --------------------------------------------
switch(idim) {
case 1: kronmult1_batched<T>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 2: kronmult2_batched<T>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 3: kronmult3_batched<T>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 4: kronmult4_batched<T>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 5: kronmult5_batched<T>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
case 6: kronmult6_batched<T>( n,
dAarray_,
dZarray_,
dYarray_,
dWarray_,
batchCount );
break;
default:
expect( false );
};
}
#endif
auto time_end = std::chrono::steady_clock::now();
auto elapsed_time_us = std::chrono::duration_cast<std::chrono::microseconds>(time_end - time_start).count();
auto elapsed_time_sec = elapsed_time_us * 0.001 * 0.001;
// ------------------------------------------
// copy from gpu to host
// interface is gpu2host( dest, src, nbytes )
// ------------------------------------------
gpu2host( Yarray_, dYarray_, sizeof(T)*Xsize*batchCount);
{
double const giga = 1000.0*1000.0*1000.0;
double const flops = 12.0*(std::pow(n,(idim+1))) * batchCount;
double const gflops = flops/giga;
double const gflops_per_sec = gflops /elapsed_time_sec;
if (flops > 0.01 * giga) {
std::cout << " idim = " << idim
<< " n = " << n
<< " batchCount = " << batchCount
<< " elapsed_time = " << elapsed_time_sec << " seconds "
<< " Gflops/sec = " << gflops_per_sec
<< "\n";
};
};
T max_abserr = 0;
if (do_check) {
// -------------
// check results
// -------------
for(int ibatch=1; ibatch <= batchCount; ibatch++) {
T const * const A1_ = &(Aarray(1,1,1,ibatch));
T const * const A2_ = &(Aarray(1,1,2,ibatch));
T const * const A3_ = &(Aarray(1,1,3,ibatch));
T const * const A4_ = &(Aarray(1,1,4,ibatch));
T const * const A5_ = &(Aarray(1,1,5,ibatch));
T const * const A6_ = &(Aarray(1,1,6,ibatch));
T const * const X_ = &(Xarray(1,ibatch));
T * const Y_ = &(Yarray(1,ibatch));
auto X = [&] (int const i) -> T const & {
return( X_[ (i)-1 ]);
};
auto Y = [&] (int const i) -> T& {
return( Y_[ (i)-1 ]);
};
auto A1 = [&](int const i,
int const j) -> T const & {
return( A1_[ indx2f(i,j,n) ] );
};
auto A2 = [&](int const i,
int const j) -> T const & {
return( A2_[ indx2f(i,j,n) ] );
};
auto A3 = [&](int const i,
int const j) -> T const & {
return( A3_[ indx2f(i,j,n) ] );
};
auto A4 = [&](int const i,
int const j) -> T const & {
return( A4_[ indx2f(i,j,n) ] );
};
auto A5 = [&](int const i,
int const j) -> T const & {
return( A5_[ indx2f(i,j,n) ] );
};
auto A6 = [&](int const i,
int const j) -> T const & {
return( A6_[ indx2f(i,j,n) ] );
};
int const max_i1 = (idim >= 1) ? n : 1;
int const max_i2 = (idim >= 2) ? n : 1;
int const max_i3 = (idim >= 3) ? n : 1;
int const max_i4 = (idim >= 4) ? n : 1;
int const max_i5 = (idim >= 5) ? n : 1;
int const max_i6 = (idim >= 6) ? n : 1;
int const max_j1 = (idim >= 1) ? n : 1;
int const max_j2 = (idim >= 2) ? n : 1;
int const max_j3 = (idim >= 3) ? n : 1;
int const max_j4 = (idim >= 4) ? n : 1;
int const max_j5 = (idim >= 5) ? n : 1;
int const max_j6 = (idim >= 6) ? n : 1;
#pragma omp parallel for collapse(6) reduction(max:max_abserr)
for(int i1=1; i1 <= max_i1; i1++)
for(int i2=1; i2 <= max_i2; i2++)
for(int i3=1; i3 <= max_i3; i3++)
for(int i4=1; i4 <= max_i4; i4++)
for(int i5=1; i5 <= max_i5; i5++)
for(int i6=1; i6 <= max_i6; i6++) {
int const ic = 1+indx6f( i6,i5,i4,i3,i2,i1,
max_i6, max_i5, max_i4,
max_i3, max_i2 );
T Y_ic = 0;
for(int j1=1; j1 <= max_j1; j1++) {
for(int j2=1; j2 <= max_j2; j2++) {
for(int j3=1; j3 <= max_j3; j3++) {
for(int j4=1; j4 <= max_j4; j4++) {
for(int j5=1; j5 <= max_j5; j5++) {
for(int j6=1; j6 <= max_j6; j6++) {
// -------------------------------
// note last index i6 goes fastest
// -------------------------------
int const jc = 1+indx6f( j6,j5,j4,j3,j2,j1,
max_j6, max_j5, max_j4,
max_j3, max_j2 );
T C_ic_jc = 1;
C_ic_jc *= (idim >= 1) ? A1(i1,j1) : 1;
C_ic_jc *= (idim >= 2) ? A2(i2,j2) : 1;
C_ic_jc *= (idim >= 3) ? A3(i3,j3) : 1;
C_ic_jc *= (idim >= 4) ? A4(i4,j4) : 1;
C_ic_jc *= (idim >= 5) ? A5(i5,j5) : 1;
C_ic_jc *= (idim >= 6) ? A6(i6,j6) : 1;
T const X_jc = X(jc);
Y_ic += C_ic_jc * X_jc;
};
};
};
};
};
};
T const abs_err = std::abs( Y_ic - Y(ic) );
max_abserr = std::max( max_abserr, abs_err );
if (idebug >= 1) {
T const tol = 1.0/(1000.0 * 1000.0);
if (abs_err > tol ) {
std::cout << " idim = " << idim
<< " ic = " << ic
<< " Y_ic = " << Y_ic
<< " Y(ic) = " << Y(ic)
<< " abs_err = " << abs_err << "\n";
};
};
};
}; // end for ibatch
};
// -------
// cleanup
// -------
myfree( dAarray_ ); dAarray_ = nullptr;
myfree( dXarray_ ); dXarray_ = nullptr;
myfree( dYarray_ ); dYarray_ = nullptr;
myfree( dZarray_ ); dZarray_ = nullptr;
myfree( dWarray_ ); dWarray_ = nullptr;
free( Aarray_ ); Aarray_ = nullptr;
free( Xarray_ ); Xarray_ = nullptr;
free( Yarray_ ); Yarray_ = nullptr;
free( Zarray_ ); Zarray_ = nullptr;
free( Warray_ ); Warray_ = nullptr;
return(max_abserr);
}
int main() {
int idebug = 0;
int batch_table[] = {1,16,128};
int const size_batch_table = sizeof(batch_table)/sizeof(batch_table[0]);
int n_table[] = {1, 2,3, 4 };
int const size_n_table = sizeof(n_table)/sizeof(n_table[0]);
int nerrors = 0;
for (int idim =1; idim <= 6; idim++) {
for (int ibatch_table=0; ibatch_table < size_batch_table; ibatch_table++) {
for (int in_table = 0; in_table < size_n_table; in_table++) {
int const n = n_table[in_table];
int const batchCount = batch_table[ibatch_table];
double const max_abserr = test_kronmult_batched<double>( idim, n, batchCount, idebug );
double const tol = 1.0/(1000.0 * 1000.0);
bool const isok = (max_abserr <= tol);
if (!isok) {
nerrors += 1;
};
if ((idebug >= 1) || (!isok)) {
std::cout << " idim = " << idim
<< " n = " << n
<< " batchCount = " << batchCount
<< " max_abserr= " << max_abserr << "\n";
};
};
};
};
if (nerrors == 0) {
std::cout << "ALL PASSED" << "\n";
}
else {
std::cout << "There are " << nerrors << " errors" << "\n";
};
if (nerrors == 0) {
// ---------------------
// try performance test
// ---------------------
int const batchCount = 256;
bool const do_check = 0;
idebug = 0;
int const idim = 6;
for(int n=4; n <= 8; n++) {
test_kronmult_batched<double>(idim,n, batchCount, idebug, do_check );
};
};
return(0);
}