File: client.cpp

package info (click to toggle)
hipsolver 6.4.1-1
  • links: PTS, VCS
  • area: main
  • in suites: sid
  • size: 11,096 kB
  • sloc: cpp: 72,703; f90: 8,280; sh: 573; python: 531; ansic: 84; makefile: 51; xml: 10
file content (501 lines) | stat: -rw-r--r-- 20,251 bytes parent folder | download
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
/* ************************************************************************
 * Copyright (C) 2020-2023 Advanced Micro Devices, Inc. All rights reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell cop-
 * ies of the Software, and to permit persons to whom the Software is furnished
 * to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in all
 * copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IM-
 * PLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
 * FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
 * COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
 * IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNE-
 * CTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 *
 *
 * ************************************************************************ */

#include "../include/hipsolver_dispatcher.hpp"
#include "../rocblascommon/program_options.hpp"

using rocblas_int    = int;
using rocblas_stride = ptrdiff_t;
using namespace roc;

// clang-format off
const char* help_str = R"HELP_STR(
hipSOLVER benchmark client help.

Usage: ./hipsolver-bench <options>

In addition to some common general options, the following list of options corresponds to all the parameters
that might be needed to test a given hipSOLVER function. The parameters are named as in the API user guide.
The arrays are initialized internally by the program with random values.

Note: When a required parameter/option is not provided, it will take the default value as listed below.
If no default value is defined, the program will try to calculate a suitable value depending on the context
of the problem and the tested function; if this is not possible, the program will abort with an error.
Functions that accept multiple size parameters can generally be provided a single size parameter (typically,
m) and a square-size matrix will be assumed.

Example: ./hipsolver-bench -f getrf -m 30 --lda 75
This will test getrf with a random 30x30 matrix.

Options:
)HELP_STR";
// clang-format on

int main(int argc, char* argv[])
try
{
    Arguments argus;

    // disable unit_check in client benchmark, it is only
    // used in gtest unit test
    argus.unit_check = 0;

    // enable timing check,otherwise no performance data collected
    argus.timing = 1;

    std::string function;
    char        precision;
    rocblas_int device_id;

    // take arguments and set default values
    // clang-format off
    options_description desc("rocsolver client command line options");
    desc.add_options()("help,h", "Produces this help message.")

        // test options
        ("batch_count",
         value<rocblas_int>(&argus.batch_count)->default_value(1),
            "Number of matrices or problem instances in the batch.\n"
            "                           Only applicable to batch routines.\n"
            "                           ")

        ("device",
         value<rocblas_int>(&device_id)->default_value(0),
            "Set the default device to be used for subsequent program runs.\n"
            "                           ")

        ("function,f",
         value<std::string>(&function)->default_value("getrf"),
            "The LAPACK function to test.\n"
            "                           Options are: getrf, getrs, potrf, potrf_batched, etc.\n"
            "                           ")

        ("iters,i",
         value<rocblas_int>(&argus.iters)->default_value(10),
            "Iterations to run inside the GPU timing loop.\n"
            "                           Reported time will be the average.\n"
            "                           ")

        ("mem_query",
         value<rocblas_int>(&argus.mem_query)->default_value(0),
            "Calculate the required amount of device workspace memory? 0 = No, 1 = Yes.\n"
            "                           This forces the client to print only the amount of device memory required by\n"
            "                           the function, in bytes.\n"
            "                           ")

        ("perf",
         value<rocblas_int>(&argus.perf)->default_value(0),
            "Ignore CPU timing results? 0 = No, 1 = Yes.\n"
            "                           This forces the client to print only the GPU time and the error if requested.\n"
            "                           ")

        ("precision,r",
         value<char>(&precision)->default_value('s'),
            "Precision to be used in the tests.\n"
            "                           Options are: s, d, c, z.\n"
            "                           ")

        // ("singular",
        //  value<rocblas_int>(&argus.singular)->default_value(0),
        //     "Test with degenerate matrices? 0 = No, 1 = Yes\n"
        //     "                           This will produce matrices that are singular, non positive-definite, etc.\n"
        //     "                           ")

        ("verify,v",
         value<rocblas_int>(&argus.norm_check)->default_value(0),
            "Validate GPU results with CPU? 0 = No, 1 = Yes.\n"
            "                           This will additionally print the relative error of the computations.\n"
            "                           ")

        // size options
        ("k",
         value<rocblas_int>(),
            "Matrix/vector size parameter.\n"
            "                           Represents a sub-dimension of a problem.\n"
            "                           For example, the number of Householder reflections in a transformation.\n"
            "                           ")

        ("m",
         value<rocblas_int>(),
            "Matrix/vector size parameter.\n"
            "                           Typically, the number of rows of a matrix.\n"
            "                           ")

        ("n",
         value<rocblas_int>(),
            "Matrix/vector size parameter.\n"
            "                           Typically, the number of columns of a matrix,\n"
            "                           or the order of a system or transformation.\n"
            "                           ")

        ("nrhs",
         value<rocblas_int>(),
            "Matrix/vector size parameter.\n"
            "                           Typically, the number of columns of a matrix on the right-hand side of a problem.\n"
            "                           ")

        // leading dimension options
        ("lda",
         value<rocblas_int>(),
            "Matrix size parameter.\n"
            "                           Leading dimension of matrices A.\n"
            "                           ")

        ("ldb",
         value<rocblas_int>(),
            "Matrix size parameter.\n"
            "                           Leading dimension of matrices B.\n"
            "                           ")

        ("ldc",
         value<rocblas_int>(),
            "Matrix size parameter.\n"
            "                           Leading dimension of matrices C.\n"
            "                           ")

        // ("ldt",
        //  value<rocblas_int>(),
        //     "Matrix size parameter.\n"
        //     "                           Leading dimension of matrices T.\n"
        //     "                           ")

        ("ldu",
         value<rocblas_int>(),
            "Matrix size parameter.\n"
            "                           Leading dimension of matrices U.\n"
            "                           ")

        ("ldv",
         value<rocblas_int>(),
            "Matrix size parameter.\n"
            "                           Leading dimension of matrices V.\n"
            "                           ")

        // ("ldw",
        //  value<rocblas_int>(),
        //     "Matrix size parameter.\n"
        //     "                           Leading dimension of matrices W.\n"
        //     "                           ")

        ("ldx",
         value<rocblas_int>(),
            "Matrix size parameter.\n"
            "                           Leading dimension of matrices X.\n"
            "                           ")

        // ("ldy",
        //  value<rocblas_int>(),
        //     "Matrix size parameter.\n"
        //     "                           Leading dimension of matrices Y.\n"
        //     "                           ")

        // stride options
        ("strideA",
         value<rocblas_stride>(),
            "Matrix/vector stride parameter.\n"
            "                           Stride for matrices/vectors A.\n"
            "                           ")

        // ("strideB",
        //  value<rocblas_stride>(),
        //     "Matrix/vector stride parameter.\n"
        //     "                           Stride for matrices/vectors B.\n"
        //     "                           ")

        // ("strideD",
        //  value<rocblas_stride>(),
        //     "Matrix/vector stride parameter.\n"
        //     "                           Stride for matrices/vectors D.\n"
        //     "                           ")

        // ("strideE",
        //  value<rocblas_stride>(),
        //     "Matrix/vector stride parameter.\n"
        //     "                           Stride for matrices/vectors E.\n"
        //     "                           ")

        // ("strideQ",
        //  value<rocblas_stride>(),
        //     "Matrix/vector stride parameter.\n"
        //     "                           Stride for vectors tauq.\n"
        //     "                           ")

        // ("strideP",
        //  value<rocblas_stride>(),
        //     "Matrix/vector stride parameter.\n"
        //     "                           Stride for vectors tau, taup, and ipiv.\n"
        //     "                           ")

        ("strideS",
         value<rocblas_stride>(),
            "Matrix/vector stride parameter.\n"
            "                           Stride for matrices/vectors S.\n"
            "                           ")

        ("strideU",
         value<rocblas_stride>(),
            "Matrix/vector stride parameter.\n"
            "                           Stride for matrices/vectors U.\n"
            "                           ")

        ("strideV",
         value<rocblas_stride>(),
            "Matrix/vector stride parameter.\n"
            "                           Stride for matrices/vectors V.\n"
            "                           ")

        // bdsqr options
        // ("nc",
        //  value<rocblas_int>()->default_value(0),
        //     "The number of columns of matrix C.\n"
        //     "                           Only applicable to bdsqr.\n"
        //     "                           ")

        // ("nu",
        //  value<rocblas_int>(),
        //     "The number of columns of matrix U.\n"
        //     "                           Only applicable to bdsqr.\n"
        //     "                           ")

        // ("nv",
        //  value<rocblas_int>()->default_value(0),
        //     "The number of columns of matrix V.\n"
        //     "                           Only applicable to bdsqr.\n"
        //     "                           ")

        // laswp options
        // ("k1",
        //  value<rocblas_int>(),
        //     "First index for row interchange.\n"
        //     "                           Only applicable to laswp.\n"
        //     "                           ")

        // ("k2",
        //  value<rocblas_int>(),
        //     "Last index for row interchange.\n"
        //     "                           Only applicable to laswp.\n"
        //     "                           ")

        // gesvd options
        ("jobu",
         value<char>()->default_value('N'),
            "N = none, A = the entire orthogonal matrix is computed,\n"
            "                           S = the singular vectors are computed,\n"
            "                           O = the singular vectors overwrite the original matrix.\n"
            "                           Indicates how the left singular vectors are to be calculated and stored.\n"
            "                           ")

        ("jobv",
         value<char>()->default_value('N'),
            "N = none, A = the entire orthogonal matrix is computed,\n"
            "                           S = the singular vectors are computed,\n"
            "                           O = the singular vectors overwrite the original matrix.\n"
            "                           Indicates how the right singular vectors are to be calculated and stored.\n"
            "                           ")

        // partial eigenvalue/singular value decomposition options
        ("il",
         value<rocblas_int>(),
            "Lower index in ordered subset of eigenvalues.\n"
            "                           Used in partial eigenvalue decomposition functions.\n"
            "                           ")

        ("iu",
         value<rocblas_int>(),
            "Upper index in ordered subset of eigenvalues.\n"
            "                           Used in partial eigenvalue decomposition functions.\n"
            "                           ")

        ("range",
         value<char>()->default_value('A'),
            "A = all eigenvalues, V = in (vl, vu], I = from the il-th to the iu-th.\n"
            "                           For partial eigenvalue decompositions, it indicates the type of interval in which\n"
            "                           the eigenvalues will be found.\n"
            "                           ")

        ("rank",
         value<rocblas_int>(),
            "The number of singular values to be computed.\n"
            "                           Used in partial SVD functions.\n"
            "                           ")

        ("vl",
         value<double>(),
            "Lower bound of half-open interval (vl, vu].\n"
            "                           Used in partial eigenvalue decomposition functions.\n"
            "                           Note: the used random input matrices have all eigenvalues in [-20, 20].\n"
            "                           ")

        ("vu",
         value<double>(),
            "Upper bound of half-open interval (vl, vu].\n"
            "                           Used in partial eigenvalue decomposition functions.\n"
            "                           Note: the used random input matrices have all eigenvalues in [-20, 20].\n"
            "                           ")

        // iterative Jacobi options
        ("econ",
         value<rocblas_int>()->default_value(0),
            "Enable economy size for singular vector matrices? 0 = No, 1 = Yes.\n"
            "                           Only applicable to gesvdj.\n"
            "                           ")

        ("max_sweeps",
         value<rocblas_int>()->default_value(100),
            "Maximum number of sweeps/iterations.\n"
            "                           Used in iterative Jacobi functions.\n"
            "                           ")

        ("tolerance",
         value<double>(),
            "Absolute tolerance at which convergence is accepted.\n"
            "                           Used in iterative Jacobi functions.\n"
            "                           ")

        ("sort_eig",
         value<rocblas_int>()->default_value(1),
            "0 = no sorting, 1 = ascending order.\n"
            "                           Indicates whether the computed eigenvalues are sorted in ascending order.\n"
            "                           Used in iterative Jacobi functions.\n"
            "                           ")

        // sparse routine options
        ("base1",
         value<rocblas_int>(),
            "0 = use base zero indices, 1 = use base one indices.\n"
            "                           ")

        ("nnzA",
         value<rocblas_int>(),
            "Matrix size parameter.\n"
            "                           Maximum number of non-zero entries of matrices A.\n"
            "                           ")

        ("reorder",
         value<rocblas_int>(),
            "0 = no reordering, 1 = RCM reordering,\n"
            "                           2 = AMD reordering, 3 = METIS reordering.\n"
            "                           ")

        // other options
        // ("direct",
        //  value<char>()->default_value('F'),
        //     "F = forward, B = backward.\n"
        //     "                           The order in which a series of transformations are applied.\n"
        //     "                           ")

        // ("fast_alg",
        //  value<char>()->default_value('O'),
        //     "O = out-of-place, I = in-place.\n"
        //     "                           Enables out-of-place computations.\n"
        //     "                           ")

        // ("incx",
        //  value<rocblas_int>()->default_value(1),
        //     "Increment between values in vector x.\n"
        //     "                           ")

        ("itype",
         value<char>()->default_value('1'),
            "1 = Ax, 2 = ABx, 3 = BAx.\n"
            "                           Problem type for generalized eigenproblems.\n"
            "                           ")

        ("jobz",
         value<char>()->default_value('N'),
            "N = none, V = compute eigenvectors/singular vectors of the matrix,\n"
            "                           Indicates how the eigenvectors/singular vectors are to be calculated and stored.\n"
            "                           ")

        ("side",
         value<char>(),
            "L = left, R = right.\n"
            "                           The side from which a matrix should be multiplied.\n"
            "                           ")

        // ("storev",
        //  value<char>(),
        //     "C = column-wise, R = row-wise.\n"
        //     "                           Indicates whether data is stored column-wise or row-wise.\n"
        //     "                           ")

        ("trans",
         value<char>()->default_value('N'),
            "N = no transpose, T = transpose, C = conjugate transpose.\n"
            "                           Indicates if a matrix should be transposed.\n"
            "                           ")

        ("uplo",
         value<char>()->default_value('U'),
            "U = upper, L = lower.\n"
            "                           Indicates where the data for a triangular or symmetric/hermitian matrix is stored.\n"
            "                           ");
    // clang-format on

    variables_map vm;
    store(parse_command_line(argc, argv, desc), vm);
    notify(vm);

    // print help message
    if(vm.count("help"))
    {
        std::cout << help_str << desc << std::endl;
        return 0;
    }

    argus.populate(vm);

    // set device ID
    if(!argus.perf)
    {
        rocblas_int device_count = query_device_property();
        if(device_count <= device_id)
            throw std::invalid_argument("Invalid Device ID");
    }
    set_device(device_id);

    // catch invalid arguments
    argus.validate_precision("precision");
    argus.validate_operation("trans");
    argus.validate_side("side");
    argus.validate_fill("uplo");
    // argus.validate_direct("direct");
    // argus.validate_storev("storev");
    argus.validate_svect("jobu");
    argus.validate_svect("jobv");
    // argus.validate_workmode("fast_alg");
    argus.validate_itype("itype");
    argus.validate_evect("jobz");
    argus.validate_erange("range");

    // select and dispatch function test/benchmark
    hipsolver_dispatcher::invoke(function, precision, argus);

    return 0;
}

catch(const std::invalid_argument& exp)
{
    std::cerr << exp.what() << std::endl;
    return -1;
}