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
|
/*
BLIS
An object-based framework for developing high-performance BLAS-like
libraries.
Copyright (C) 2014, The University of Texas at Austin
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.
- Neither the name(s) of the copyright holder(s) nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
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
HOLDER 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.
*/
#include "blis.h"
void bli_sgemm_opt_d2x4(
dim_t k,
float* restrict alpha,
float* restrict a,
float* restrict b,
float* restrict beta,
float* restrict c, inc_t rs_c, inc_t cs_c
)
{
bli_check_error_code( BLIS_NOT_YET_IMPLEMENTED );
}
void bli_dgemm_opt_d2x4(
dim_t k,
double* restrict alpha,
double* restrict a,
double* restrict b,
double* restrict beta,
double* restrict c, inc_t rs_c, inc_t cs_c
)
{
dim_t k_iter;
dim_t k_left;
k_iter = k / 8;
k_left = k % 8;
__asm__ volatile
(
" \n\t"
"movl %6, %%ecx \n\t" // load address of c
" \n\t"
"movl %8, %%edi \n\t" // load cs_c
"sall $3, %%edi \n\t" // cs_c *= sizeof(double)
" \n\t"
"leal (%%ecx,%%edi,2), %%edx \n\t" // load address of c + 2*cs_c
" \n\t"
"prefetcht0 (%%ecx) \n\t" // give a T0 prefetch hint for c00.
"prefetcht0 (%%ecx,%%edi) \n\t" // give a T0 prefetch hint for c01.
"prefetcht0 (%%edx) \n\t" // give a T0 prefetch hint for c02.
"prefetcht0 (%%edx,%%edi) \n\t" // give a T0 prefetch hint for c03.
" \n\t"
"movl %2, %%eax \n\t" // load address of a.
"movl %3, %%ebx \n\t" // load address of b.
" \n\t"
"addl $8 * 16, %%eax \n\t" // increment pointers to allow byte
"addl $8 * 16, %%ebx \n\t" // offsets in the unrolled iterations.
" \n\t"
"movapd -8 * 16(%%eax), %%xmm0 \n\t" // initialize loop by pre-loading elements
"movapd -8 * 16(%%ebx), %%xmm1 \n\t" // of a and b.
" \n\t"
"pxor %%xmm2, %%xmm2 \n\t"
"pxor %%xmm3, %%xmm3 \n\t"
" \n\t"
"pxor %%xmm4, %%xmm4 \n\t"
"pxor %%xmm5, %%xmm5 \n\t"
"pxor %%xmm6, %%xmm6 \n\t"
"pxor %%xmm7, %%xmm7 \n\t"
" \n\t"
" \n\t"
" \n\t"
"movl %0, %%esi \n\t" // i = k_iter;
"testl %%esi, %%esi \n\t" // check i via logical AND.
"je .CONSIDERKLEFT \n\t" // if i == 0, jump to code that
" \n\t" // contains the k_left loop.
" \n\t"
" \n\t"
".LOOPKITER: \n\t" // MAIN LOOP
" \n\t"
"prefetcht0 (8*21+4)*8(%%eax) \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 0
"movapd -7 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd -6 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -7 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 1
"movapd -5 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd -4 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -6 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 2
"movapd -3 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd -2 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -5 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 3
"movapd -1 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd 0 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -4 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
"prefetcht0 (8*21+12)*8(%%eax) \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 4
"movapd 1 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd 2 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -3 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 5
"movapd 3 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd 4 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -2 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 6
"movapd 5 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd 6 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -1 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration 7
"movapd 7 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd 8 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd 0 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
"addl $256, %%ebx \n\t" // b += 8*2*2 (unroll x nr x ndup)
"addl $128, %%eax \n\t" // a += 8*4 (unroll x mr)
" \n\t"
" \n\t"
"decl %%esi \n\t" // i -= 1;
"jne .LOOPKITER \n\t" // iterate again if i != 0.
" \n\t"
" \n\t"
" \n\t"
".CONSIDERKLEFT: \n\t"
" \n\t"
"movl %1, %%esi \n\t" // i = k_left;
"testl %%esi, %%esi \n\t" // check i via logical AND.
"je .POSTACCUM \n\t" // if i == 0, we're done; jump to end.
" \n\t" // else, we prepare to enter k_left loop.
" \n\t"
" \n\t"
".LOOPKLEFT: \n\t" // EDGE LOOP
" \n\t"
"addpd %%xmm3, %%xmm7 \n\t" // iteration i
"movapd -7 * 16(%%ebx), %%xmm3 \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"pshufd $0x4e, %%xmm1, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm1 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
" \n\t"
"addpd %%xmm1, %%xmm5 \n\t"
"movapd -6 * 16(%%ebx), %%xmm1 \n\t"
"addpd %%xmm2, %%xmm4 \n\t"
"pshufd $0x4e, %%xmm3, %%xmm2 \n\t"
"mulpd %%xmm0, %%xmm3 \n\t"
"mulpd %%xmm0, %%xmm2 \n\t"
"movapd -7 * 16(%%eax), %%xmm0 \n\t"
" \n\t"
" \n\t"
" \n\t"
" \n\t"
"addl $32, %%ebx \n\t" // b += 4 (1 x mr)
"addl $16, %%eax \n\t" // a += 2*2 (1 x nr x ndup)
" \n\t"
"decl %%esi \n\t" // i -= 1;
"jne .LOOPKLEFT \n\t" // iterate again if i != 0.
" \n\t"
" \n\t"
" \n\t"
".POSTACCUM: \n\t"
" \n\t"
"addpd %%xmm2, %%xmm6 \n\t"
"addpd %%xmm3, %%xmm7 \n\t"
" \n\t"
" \n\t"
"movl %4, %%eax \n\t" // load address of alpha
"movl %5, %%ebx \n\t" // load address of beta
"movddup (%%eax), %%xmm2 \n\t" // load alpha and duplicate
"movddup (%%ebx), %%xmm3 \n\t" // load beta and duplicate
" \n\t"
" \n\t"
" \n\t"
"movl %7, %%esi \n\t" // load rs_c
"sall $3, %%esi \n\t" // rs_c *= sizeof(double)
" \n\t"
" \n\t"
"movapd %%xmm4, %%xmm0 \n\t"
"movsd %%xmm5, %%xmm4 \n\t"
"movsd %%xmm0, %%xmm5 \n\t"
" \n\t"
"movapd %%xmm6, %%xmm0 \n\t"
"movsd %%xmm7, %%xmm6 \n\t"
"movsd %%xmm0, %%xmm7 \n\t"
" \n\t"
" \n\t"
" \n\t"
"movlpd (%%ecx), %%xmm0 \n\t" // load c00 and c10,
"movhpd (%%ecx,%%esi), %%xmm0 \n\t"
"mulpd %%xmm2, %%xmm4 \n\t" // scale by alpha,
"mulpd %%xmm3, %%xmm0 \n\t" // scale by beta,
"addpd %%xmm4, %%xmm0 \n\t" // add the gemm result,
"movlpd %%xmm0, (%%ecx) \n\t" // and store back to memory.
"movhpd %%xmm0, (%%ecx,%%esi) \n\t"
"addl %%edi, %%ecx \n\t"
" \n\t"
"movlpd (%%edx), %%xmm1 \n\t" // load c02 and c12,
"movhpd (%%edx,%%esi), %%xmm1 \n\t"
"mulpd %%xmm2, %%xmm6 \n\t" // scale by alpha,
"mulpd %%xmm3, %%xmm1 \n\t" // scale by beta,
"addpd %%xmm6, %%xmm1 \n\t" // add the gemm result,
"movlpd %%xmm1, (%%edx) \n\t" // and store back to memory.
"movhpd %%xmm1, (%%edx,%%esi) \n\t"
"addl %%edi, %%edx \n\t"
" \n\t"
"movlpd (%%ecx), %%xmm0 \n\t" // load c01 and c11,
"movhpd (%%ecx,%%esi), %%xmm0 \n\t"
"mulpd %%xmm2, %%xmm5 \n\t" // scale by alpha,
"mulpd %%xmm3, %%xmm0 \n\t" // scale by beta,
"addpd %%xmm5, %%xmm0 \n\t" // add the gemm result,
"movlpd %%xmm0, (%%ecx) \n\t" // and store back to memory.
"movhpd %%xmm0, (%%ecx,%%esi) \n\t"
" \n\t"
"movlpd (%%edx), %%xmm1 \n\t" // load c03 and c13,
"movhpd (%%edx,%%esi), %%xmm1 \n\t"
"mulpd %%xmm2, %%xmm7 \n\t" // scale by alpha,
"mulpd %%xmm3, %%xmm1 \n\t" // scale by beta,
"addpd %%xmm7, %%xmm1 \n\t" // add the gemm result,
"movlpd %%xmm1, (%%edx) \n\t" // and store back to memory.
"movhpd %%xmm1, (%%edx,%%esi) \n\t"
" \n\t"
" \n\t"
" \n\t"
: // output operands (none)
: // input operands
"m" (k_iter),
"m" (k_left),
"m" (a),
"m" (b),
"m" (alpha),
"m" (beta),
"m" (c),
"m" (rs_c),
"m" (cs_c)
: // register clobber list
"eax", "ebx", "ecx", "edx", "esi", "edi",
"xmm0", "xmm1", "xmm2", "xmm3",
"xmm4", "xmm5", "xmm6", "xmm7",
"memory"
);
}
void bli_cgemm_opt_d2x4(
dim_t k,
scomplex* restrict alpha,
scomplex* restrict a,
scomplex* restrict b,
scomplex* restrict beta,
scomplex* restrict c, inc_t rs_c, inc_t cs_c
)
{
bli_check_error_code( BLIS_NOT_YET_IMPLEMENTED );
}
void bli_zgemm_opt_d2x4(
dim_t k,
dcomplex* restrict alpha,
dcomplex* restrict a,
dcomplex* restrict b,
dcomplex* restrict beta,
dcomplex* restrict c, inc_t rs_c, inc_t cs_c
)
{
bli_check_error_code( BLIS_NOT_YET_IMPLEMENTED );
}
|
