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
|
/*
-- MAGMA (version 2.9.0) --
Univ. of Tennessee, Knoxville
Univ. of California, Berkeley
Univ. of Colorado, Denver
@date January 2025
@author Stan Tomov
@author Mark Gates
@author Azzam Haidar
@author Ahmad Abdelfattah
*/
#include "magma_internal.h"
// for every size [1:32], how many 1D configs can a warp hold?
#define NTCOL_1D_DEFAULT 32, 16, 10, 8, 6, 5, 4, 4, 3, 3, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
const magma_int_t ntcol_1d_default[] = {NTCOL_1D_DEFAULT};
// =============================================================================
// GEMM
// =============================================================================
// Kepler (or older)
const magma_int_t sgemm_batched_ntcol_300[] = {64,32,32,32,10,8,10,4,3,5,2,2,3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t dgemm_batched_ntcol_300[] = {64,32,32,8,5,8,10,8,6,5,4,2,3,1,2,2,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t cgemm_batched_ntcol_300[] = {64,32,32,64,5,8,10,10,8,5,4,2,3,1,2,2,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t zgemm_batched_ntcol_300[] = {64,16,8,6,4,4,12,10,8,4,1,3,2,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
// Pascal (used also for maxwell)
const magma_int_t sgemm_batched_ntcol_600[] = {64,64,64,32,14,13,9,5,7,3,5,3,3,2,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t dgemm_batched_ntcol_600[] = {64,64,32,16,16,8,10,1,8,8,2,2,3,5,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t cgemm_batched_ntcol_600[] = {64,32,64,32,32,8,13,1,3,8,2,2,3,2,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t zgemm_batched_ntcol_600[] = {64,14,14,14,32,8,16,2,4,2,4,2,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
// Volta
const magma_int_t sgemm_batched_ntcol_700[] = {64,64,32,8,10,8,10,8,8,10,4,2,3,5,2,1,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t dgemm_batched_ntcol_700[] = {64,15,32,15,15,8,10,8,8,10,8,2,3,5,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t cgemm_batched_ntcol_700[] = {64,64,16,32,8,8,10,8,8,5,2,2,3,5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
const magma_int_t zgemm_batched_ntcol_700[] = {64,32,32,15,8,6,4,4,3,5,1,2,1,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1};
// =============================================================================
// LU
// =============================================================================
// Kepler (or older)
const magma_int_t sgetrf_batched_ntcol_300[] = {32, 16, 8, 8, 8, 8, 8, 32, 8, 16, 8, 8, 8, 8, 8, 8, 4, 4, 4, 8, 4, 4, 4, 4, 4, 8, 8, 8, 4, 4, 4, 4};
const magma_int_t dgetrf_batched_ntcol_300[] = {32, 16, 8, 16, 8, 4, 4, 32, 4, 4, 8, 8, 8, 8, 8, 8, 4, 4, 16, 16, 16, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4};
const magma_int_t cgetrf_batched_ntcol_300[] = {NTCOL_1D_DEFAULT};
const magma_int_t zgetrf_batched_ntcol_300[] = {NTCOL_1D_DEFAULT};
// Pascal (used also for maxwell)
const magma_int_t sgetrf_batched_ntcol_600[] = {8, 5, 3, 2, 3, 3, 3, 3, 32, 32, 32, 32, 2, 15, 16, 16, 16, 13, 12, 10, 11, 10, 10, 8, 9, 8, 2, 6, 6, 6, 1, 1};
const magma_int_t dgetrf_batched_ntcol_600[] = {8, 4, 2, 3, 2, 2, 2, 5, 3, 3, 5, 12, 3, 3, 9, 8, 6, 6, 6, 5, 6, 6, 6, 8, 8, 8, 7, 4, 4, 4, 4, 4};
const magma_int_t cgetrf_batched_ntcol_600[] = {NTCOL_1D_DEFAULT};
const magma_int_t zgetrf_batched_ntcol_600[] = {NTCOL_1D_DEFAULT};
// Volta
const magma_int_t sgetrf_batched_ntcol_700[] = {8, 12, 3, 4, 4, 4, 2, 2, 1, 1, 1, 1, 1, 1, 1, 8, 12, 12, 12, 10, 10, 10, 10, 8, 8, 8, 6, 6, 4, 6, 4, 1};
const magma_int_t dgetrf_batched_ntcol_700[] = {7, 4, 2, 2, 1, 1, 1, 2, 1, 1, 1, 3, 1, 1, 1, 9, 12, 8, 8, 10, 10, 10, 10, 8, 8, 8, 7, 6, 6, 6, 4, 4};
const magma_int_t cgetrf_batched_ntcol_700[] = {NTCOL_1D_DEFAULT};
const magma_int_t zgetrf_batched_ntcol_700[] = {NTCOL_1D_DEFAULT};
// =============================================================================
// QR
// =============================================================================
// Kepler (or older)
const magma_int_t sgeqrf_batched_ntcol_300[] = {NTCOL_1D_DEFAULT};
const magma_int_t dgeqrf_batched_ntcol_300[] = {NTCOL_1D_DEFAULT};
const magma_int_t cgeqrf_batched_ntcol_300[] = {NTCOL_1D_DEFAULT};
const magma_int_t zgeqrf_batched_ntcol_300[] = {NTCOL_1D_DEFAULT};
// Pascal (used also for maxwell)
const magma_int_t sgeqrf_batched_ntcol_600[] = {16, 4, 3, 4, 3, 3, 6, 14, 32, 32, 32, 32, 16, 14, 16, 10, 9, 10, 9, 9, 6, 8, 6, 6, 8, 8, 2, 6, 6, 2, 4, 2};
const magma_int_t dgeqrf_batched_ntcol_600[] = {16, 4, 3, 3, 5, 5, 5, 10, 14, 10, 10, 10, 12, 12, 10, 4, 6, 6, 4, 8, 7, 7, 4, 4, 6, 2, 2, 1, 2, 2, 1, 1};
const magma_int_t cgeqrf_batched_ntcol_600[] = {NTCOL_1D_DEFAULT};
const magma_int_t zgeqrf_batched_ntcol_600[] = {NTCOL_1D_DEFAULT};
// Volta
const magma_int_t sgeqrf_batched_ntcol_700[] = {32, 4, 3, 6, 5, 3, 2, 2, 3, 3, 3, 3, 3, 11, 3, 16, 12, 12, 12, 12, 10, 10, 10, 8, 8, 8, 1, 1, 1, 1, 1, 1};
const magma_int_t dgeqrf_batched_ntcol_700[] = {32, 4, 2, 2, 1, 1, 1, 5, 9, 16, 11, 8, 3, 3, 15, 16, 8, 8, 8, 4, 12, 4, 4, 6, 6, 4, 6, 4, 6, 4, 4, 4};
const magma_int_t cgeqrf_batched_ntcol_700[] = {NTCOL_1D_DEFAULT};
const magma_int_t zgeqrf_batched_ntcol_700[] = {NTCOL_1D_DEFAULT};
#ifdef __cplusplus
extern "C" {
#endif
// Definition of blocking sizes for NVIDIA cards
#if defined(MAGMA_HAVE_CUDA) || defined(MAGMA_HAVE_HIP)
// =============================================================================
/// @addtogroup magma_tuning
/// Tuning of the batched kernels that are invoked on extremely small sizes
/// @{
/***************************************************************************//**
@return the ntcol value for very small xgemm_batched ( m = n = k)
*******************************************************************************/
magma_int_t magma_get_zgemm_batched_ntcol(magma_int_t m)
{
magma_int_t* ntcol_array;
if(m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)zgemm_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)zgemm_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)zgemm_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgemm_batched_ntcol
magma_int_t magma_get_cgemm_batched_ntcol(magma_int_t m)
{
magma_int_t* ntcol_array;
if(m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)cgemm_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)cgemm_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)cgemm_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgemm_batched_ntcol
magma_int_t magma_get_dgemm_batched_ntcol(magma_int_t m)
{
magma_int_t* ntcol_array;
if(m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)dgemm_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)dgemm_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)dgemm_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgemm_batched_ntcol
magma_int_t magma_get_sgemm_batched_ntcol(magma_int_t m)
{
magma_int_t* ntcol_array;
if(m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)sgemm_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)sgemm_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)sgemm_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/***************************************************************************//**
@return the ntcol value for very small xgetrf_batched ( m = n )
*******************************************************************************/
magma_int_t magma_get_zgetrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)zgetrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)zgetrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)zgetrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgetrf_batched_ntcol
magma_int_t magma_get_cgetrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)cgetrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)cgetrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)cgetrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgetrf_batched_ntcol
magma_int_t magma_get_dgetrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)dgetrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)dgetrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)dgetrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgetrf_batched_ntcol
magma_int_t magma_get_sgetrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)sgetrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)sgetrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)sgetrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/***************************************************************************//**
@return the ntcol value for very small xgeqrf_batched ( m = n )
*******************************************************************************/
magma_int_t magma_get_zgeqrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)zgeqrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)zgeqrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)zgeqrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgeqrf_batched_ntcol
magma_int_t magma_get_cgeqrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)cgeqrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)cgeqrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)cgeqrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgeqrf_batched_ntcol
magma_int_t magma_get_dgeqrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)dgeqrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)dgeqrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)dgeqrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
/// @see magma_get_zgeqrf_batched_ntcol
magma_int_t magma_get_sgeqrf_batched_ntcol(magma_int_t m, magma_int_t n)
{
magma_int_t* ntcol_array;
if(m != n || m < 0 || m > 32) return 1;
magma_int_t arch = magma_getdevice_arch();
if (arch <= 300) ntcol_array = (magma_int_t*)sgeqrf_batched_ntcol_300;
else if (arch <= 600) ntcol_array = (magma_int_t*)sgeqrf_batched_ntcol_600;
else if (arch <= 700) ntcol_array = (magma_int_t*)sgeqrf_batched_ntcol_700;
else ntcol_array = (magma_int_t*)ntcol_1d_default;
return ntcol_array[m-1];
}
// =============================================================================
/// @}
// end group magma_tuning
#endif // MAGMA_HAVE_CUDA
#ifdef __cplusplus
} // extern "C"
#endif
|
