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
|
/* ************************************************************************
* Copyright 2013 Advanced Micro Devices, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
* ************************************************************************/
// TRSV Column-Major Upper Kernel
//#include <TRSV.h>
const char * trsv_CU_SolveTriangle_kernel = "
#ifdef DOUBLE_PRECISION
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#else
#pragma OPENCL EXTENSION cl_amd_fp64 : enable
#endif
#endif
#ifdef PACKED
#define A( row, col) (*( A + (((col)*((col)+1))/2 + (row))))
#elif defined(BANDED)
#define A( row, col) A[ (row) * lda + (col)]
#else
#define A( row, col) A[ (row) + (col) * lda]
#endif
// Only one workgroup of threads launched
__kernel void %PREFIXtrsv_CU_SolveTriangle_kernel( __global %TYPE const * restrict _A, __global %TYPE* _xnew, uint N, int incx, int isUnity,
uint lda, int doConj, int startRow, int startCol, uint offa, uint offx
#ifdef BANDED
, uint KU
#endif
)
{
__global %TYPE* xnew;
__global %TYPE const * restrict A = _A + offa;
if ( incx < 0 ) // Goto end of vector
{
xnew = _xnew + offx + ( N - 1) * abs(incx);
}
else
{
xnew = _xnew + offx;
}
__local %TYPE xShared; // To share solved x value with other threads..
size_t gIdx = get_global_id(0);
size_t bIdx = get_group_id(0);
size_t threadIdx= get_local_id(0);
%TYPE sum = %MAKEVEC(0.0);
%TYPE xVal = %MAKEVEC(0.0);
%TYPE loadedA = %MAKEVEC(0.0);
int targetCol = startCol;
int targetRow = startRow + threadIdx;
int loops = (startCol - startRow) + 1;
#ifdef BANDED
int bandCol = (loops - 1) - threadIdx;
#endif
for( int i=0; i < loops; i++)
{
if ( targetRow == targetCol)
{
xVal = xnew[ targetRow * incx];
%SUB(sum, xVal, sum);
if( isUnity)
{
xShared = sum;
}
else // Handle diagonal element
{
#ifdef BANDED
loadedA = A((targetRow), (bandCol));
#else
loadedA = A((targetRow), (targetCol));
#endif
%CONJUGATE(doConj, loadedA);
%DIV(xShared, sum, loadedA);
}
xnew[ targetRow * incx ] = xShared;
}
// Sync so that xShared it available to all threads
barrier(CLK_LOCAL_MEM_FENCE);
if ( targetRow < targetCol)
{
#ifdef BANDED
loadedA = A((targetRow), (bandCol));
#else
loadedA = A((targetRow), (targetCol));
#endif
%CONJUGATE(doConj, loadedA);
%MAD(sum, loadedA, xShared);
}
// Avoid Race...
barrier(CLK_LOCAL_MEM_FENCE);
targetCol--;
#ifdef BANDED
bandCol--;
#endif
}
}";
const char * trsv_CL_SolveTriangle_kernel = "
#ifdef DOUBLE_PRECISION
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#else
#pragma OPENCL EXTENSION cl_amd_fp64 : enable
#endif
#endif
#ifdef PACKED
#define A(row, col) (*( A + ((( (col) *((2*N) + 1 - (col))) / 2) + ((row) - (col)))))
#elif defined(BANDED)
#define A(row, col) A[ (row) * lda + (col)]
#else
#define A(row, col) A[ (row) + (col) * lda]
#endif
#pragma OPENCL EXTENSION cl_amd_printf : enable
// Only one block of threads launched
__kernel void %PREFIXtrsv_CL_SolveTriangle_kernel( __global const %TYPE* _A, __global %TYPE* _xnew, uint N, int incx, int isUnity,
uint lda, int doConj, int startCol, int endRow, uint offa, uint offx
#ifdef BANDED
, uint KL
#endif
)
{
__global %TYPE* xnew;
__global %TYPE* A = _A + offa;
if ( incx < 0 ) // Goto end of vector
{
xnew = _xnew + offx + ( N - 1) * abs(incx);
}
else
{
xnew = _xnew + offx;
}
__local %TYPE xShared; // To share solved x value with other threads..
size_t gIdx = get_global_id(0);
size_t bIdx = get_group_id(0);
size_t threadIdx= get_local_id(0);
%TYPE sum = %MAKEVEC(0.0);
%TYPE xVal = %MAKEVEC(0.0);
%TYPE loadedA = %MAKEVEC(0.0);
int targetCol = startCol;
int targetRow = endRow - threadIdx;
int loops = (endRow - startCol) + 1;
#ifdef BANDED
int bandCol = (KL + 1) - loops + threadIdx;
#endif
// printf(\"%u : bandCol %d targetCol %d targetRow %d loops %d KL %d\\n\", threadIdx, bandCol, targetCol, targetRow, loops, KL);
for( int i=0; i < loops; i++)
{
if ( targetRow == targetCol)
{
xVal = xnew[ targetRow * incx];
//printf(\"Before1 %u : xShared %f, sum %f\\n\", threadIdx, xShared, sum);
%SUB(sum, xVal, sum);
//printf(\"Before2 %u : xShared %f, sum %f XvAL %f, targetRow %d\\n\", threadIdx, xShared, sum, xVal, targetRow);
if( isUnity)
{
xShared = sum;
}
else // Handle diagonal element
{
#ifndef BANDED
loadedA = A((targetRow), (targetCol));
#else
loadedA = A((targetRow), (bandCol));
#endif
%CONJUGATE(doConj, loadedA);
%DIV(xShared, sum, loadedA);
}
//printf(\"After %u : xShared %f, sum %f\\n\", threadIdx, xShared, sum);
xnew[ targetRow * incx ] = xShared;
}
// Sync so that xShared it available to all threads
barrier(CLK_LOCAL_MEM_FENCE);
if ( targetRow > targetCol)
{
#ifndef BANDED
loadedA = A((targetRow), (targetCol));
#else
loadedA = A((targetRow), (bandCol));
#endif
%CONJUGATE(doConj, loadedA);
//printf(\"%u : xShared %f, sum %f loadedA %f\\n\", threadIdx, xShared, sum, loadedA);
%MAD(sum, loadedA, xShared);
}
// Avoid Race...
barrier(CLK_LOCAL_MEM_FENCE);
targetCol++;
#ifdef BANDED
bandCol++;
#endif
}
}
";
const char * trsv_CUT_SolveTriangle_kernel = "
#ifdef DOUBLE_PRECISION
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#else
#pragma OPENCL EXTENSION cl_amd_fp64 : enable
#endif
#endif
#ifdef PACKED
#define A( row, col) (*( A + (((col)*((col)+1))/2 + (row))))
#elif defined(BANDED)
#define A(row, col) A[ (row) * lda + (col)]
#else
#define A( row, col) A[ (row) + (col) * lda]
#endif
#pragma OPENCL EXTENSION cl_amd_printf : enable
__kernel void %PREFIXtrsv_CUT_SolveTriangle_kernel( __global const %TYPE* _A,
__global %TYPE* _xnew,
uint N,
int incx,
int isUnity,
uint lda,
int doConj,
int startRow, int endRow, uint offa, uint offx
#ifdef BANDED
, uint KU
#endif
)
{
__global %TYPE* xnew;
__global const %TYPE* A = _A + offa;
if ( incx < 0 ) // Goto end of vector
{
xnew = _xnew + offx + ( N - 1) * abs(incx);
}
else
{
xnew = _xnew + offx;
}
int blockSize = get_local_size(0);
int threadID = get_local_id(0);
int targetRow;
#ifdef BANDED
int bandRow = startRow;
int bandCol = threadID;
// printf(\"threadID %d, bandRow %d bandCol %d\\n\",threadID, bandRow, bandCol);
#endif
__local volatile %TYPE saccShared[%TRIANGLE_HEIGHT];
targetRow = startRow + threadID;
//#pragma unroll
for( int idx = threadID; (idx < %TRIANGLE_HEIGHT) && ((startRow + idx) < endRow); idx += blockSize)
{
saccShared[idx] = xnew[ (startRow + idx) * incx];
}
barrier(CLK_LOCAL_MEM_FENCE);
%TYPE diagA = %INIT(0.0);
if(targetRow < endRow)
{
#ifndef BANDED
diagA = A((targetRow), (targetRow));
#else
diagA = A((startRow + threadID), (0));
#endif
%CONJUGATE(doConj, diagA);
}
%TYPE tempA, tempS;
for(int i = 0; i < %TRIANGLE_HEIGHT; i++)
{
if((i <= threadID) && (i > 0) && (targetRow < endRow))
{
#ifndef BANDED
tempA = A((startRow + i - 1), (targetRow));
#else
tempA = A((bandRow - 1), (bandCol + 1));
// printf(\"threadID %d, bandRow %d bandCol %d A %f\\n\",threadID, bandRow, bandCol, tempA);
#endif
%CONJUGATE(doConj, tempA);
%MUL(tempS, tempA, saccShared[i-1]);
%SUB(saccShared[threadID], saccShared[threadID], tempS);
}
if((i == threadID) && (targetRow < endRow) && (!isUnity))
{
tempS = saccShared[threadID];
// printf(\"threadID %d, saccShared %f, diagA %f\\n\", threadID, tempS, diagA);
%DIV(saccShared[threadID], tempS, diagA);
}
barrier(CLK_LOCAL_MEM_FENCE);
#ifdef BANDED
bandRow++; bandCol--;
#endif
}
barrier(CLK_LOCAL_MEM_FENCE);
if(targetRow < endRow)
{
xnew[(targetRow * incx)] = saccShared[threadID];
}
}
";
const char * trsv_CLT_SolveTriangle_kernel = "
#ifdef DOUBLE_PRECISION
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#else
#pragma OPENCL EXTENSION cl_amd_fp64 : enable
#endif
#endif
#ifdef PACKED
#define A(row, col) (*( A + ((( (col) *((2*N) + 1 - (col))) / 2) + ((row) - (col)))))
#elif defined(BANDED)
#define A(row, col) A[ (row) * lda + (col)]
#else
#define A(row, col) A[ (row) + (col) * lda]
#endif
#pragma OPENCL EXTENSION cl_amd_printf : enable
// Column-Major Lower Non-Unity case
// StartRow points to actual Row to start from( absolute Column number)
// endRow points to actual Row to stop + 1( absolute Column number)
__kernel void %PREFIXtrsv_CLT_SolveTriangle_kernel( __global const %TYPE* _A,
__global %TYPE* _xnew,
uint N,
int incx,
int isUnity,
uint lda,
int doConj,
int startRow, int endRow, uint offa, uint offx
#ifdef BANDED
,uint KL
#endif
)
{
__global %TYPE* xnew;
__global const %TYPE *A = _A + offa;
if ( incx < 0 ) // Goto end of vector
{
xnew = _xnew + offx + ( N - 1) * abs(incx);
}
else
{
xnew = _xnew + offx;
}
int blockSize = get_local_size(0);
int threadID = get_local_id(0);
__local volatile %TYPE saccShared[%TRIANGLE_HEIGHT];
int targetRow;
targetRow = (endRow - 1) - threadID;
#ifdef BANDED
int bandRow = (endRow - 1);
int bandCol = (KL) - threadID;
#endif
//#pragma unroll
for( int idx = threadID; (idx < %TRIANGLE_HEIGHT) && (((endRow - 1) - idx) >= startRow); idx += blockSize)
{
saccShared[idx] = xnew[((endRow - 1) - idx) * incx];
}
barrier(CLK_LOCAL_MEM_FENCE);
%TYPE diagA = %INIT(0.0);
if(targetRow >= startRow)
{
#ifndef BANDED
diagA = A((targetRow), (targetRow));
#else
diagA = A((bandRow - threadID), (KL));
// printf(\"ThreadID %d, bandRow %d bandCol %d\\n\", threadID, bandRow, bandCol);
#endif
%CONJUGATE(doConj, diagA);
}
%TYPE tempA, tempS;
for( int i = (endRow - 1); i >= startRow; i--)
{
if((targetRow == i) && (!isUnity))
{
tempS = saccShared[threadID];
%DIV(saccShared[threadID], tempS, diagA);
}
barrier(CLK_LOCAL_MEM_FENCE);
if((targetRow < i) && (targetRow >= startRow))
{
#ifndef BANDED
tempA = A((i), (targetRow));
#else
tempA = A((bandRow), (bandCol));
#endif
%CONJUGATE(doConj, tempA);
%MUL(tempS, tempA, saccShared[(endRow - 1) - i]);
%SUB(saccShared[threadID], saccShared[threadID], tempS);
}
barrier(CLK_LOCAL_MEM_FENCE);
#ifdef BANDED
bandRow--; bandCol++;
#endif
}
barrier(CLK_LOCAL_MEM_FENCE);
if(targetRow >= startRow)
{
xnew[(targetRow * incx)] = saccShared[threadID];
}
}
";
|
