-
Notifications
You must be signed in to change notification settings - Fork 351
/
Copy pathcm_kernel_rt.cpp
6202 lines (5472 loc) · 238 KB
/
cm_kernel_rt.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
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* Copyright (c) 2007-2017, Intel Corporation
*
* 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 copies 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 IMPLIED, 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 CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/
//!
//! \file cm_kernel_rt.cpp
//! \brief Contains CmKernelRT definitions.
//!
#include "cm_kernel_rt.h"
#include "cm_program.h"
#include "cm_device_rt.h"
#include "cm_surface_manager.h"
#include "cm_surface_2d_up_rt.h"
#include "cm_surface_3d_rt.h"
#include "cm_buffer_rt.h"
#include "cm_mov_inst.h"
#include "cm_kernel_data.h"
#include "cm_thread_space_rt.h"
#include "cm_state_buffer.h"
#include "cm_surface_vme.h"
#include "cm_debug.h"
#include "cm_surface_sampler8x8.h"
#include "cm_surface_sampler.h"
#include "cm_group_space.h"
#include "cm_surface_2d_rt.h"
#include "cm_sampler8x8_state_rt.h"
#include "cm_visa.h"
#include "cm_extension_creator.h"
#include "cm_execution_adv.h"
#define GENERATE_GLOBAL_SURFACE_INDEX
#define READ_FIELD_FROM_BUF( dst, type ) \
dst = *((type *) &buf[bytePosition]); \
bytePosition += sizeof(type);
#define PER_ARG_SIZE_IN_DWORD 3
#define KERNEL_INFO_SIZE_IN_DWORD 4
#define DW_ALIGNMENT( byte_address ) \
if( byte_address % 4 ) \
byte_address = ( byte_address / 4 + 1 ) * 4;
#define GRF_ALIGNMENT( byte_address ) \
if( byte_address % 32 ) \
byte_address = ( byte_address / 32 + 1 ) * 32;
// To check if surface type nType is equal to the surface type list in argument ...
#define CHECK_SURFACE_TYPE( nType, ... ) ( _CheckSurfaceType( nType, __VA_ARGS__, -1 ) )
#define IsKernelArg(arg) ((arg).unitCount == 1)
// Warning : x must be uint32_t
#define SET_MEMORY_OBJECT_CONTROL(x, memCtl) \
x = ((uint16_t)(memCtl.mem_ctrl<< 8 | memCtl.mem_type << 4 | memCtl.age)) << 16 | (x);
#define ADD_INTO_VME_INDEX_ARRAY(value) \
vmeIndexArray[vmeIndexArrayPosition] = value ; \
vmeIndexArrayPosition ++;
#define ADD_INTO_VME_CM_INDEX_ARRAY(value) ; \
vmeCmIndexArray[vmeCmIndexArrayPosition] = value ; \
vmeCmIndexArrayPosition ++;
typedef CM_ARG* PCM_ARG;
#define CM_KERNEL_DATA_CLEAN 0 // kernel data clean
#define CM_KERNEL_DATA_KERNEL_ARG_DIRTY 1 // per kernel arg dirty
#define CM_KERNEL_DATA_THREAD_ARG_DIRTY (1 << 1) // per thread arg dirty
#define CM_KERNEL_DATA_PAYLOAD_DATA_DIRTY (1 << 2) // indirect payload data dirty
#define CM_KERNEL_DATA_PAYLOAD_DATA_SIZE_DIRTY (1 << 3) // indirect payload data size changes
#define CM_KERNEL_DATA_GLOBAL_SURFACE_DIRTY (1 << 4) // global surface dirty
#define CM_KERNEL_DATA_THREAD_COUNT_DIRTY (1 << 5) // thread count dirty, reset() be called
#define cMKERNELDATASAMPLERBTIDIRTY (1 << 6) // sampler bti dirty
#define CM_KERNEL_DATA_THREAD_GROUP_SPACE_DIRTY (1 << 7) // threadgroupspace dirty
int32_t Partition( PCM_ARG* args, int32_t p, int32_t r )
{
uint16_t x = args[p]->unitOffsetInPayload;
int32_t i = p - 1;
int32_t j = r + 1;
while( 1 )
{
do {
j --;
} while( args[j]->unitOffsetInPayload > x );
do {
i ++;
} while( args[i]->unitOffsetInPayload < x );
if( i < j )
{
PCM_ARG tmpP = args[i];
args[i] = args[j];
args[j] = tmpP;
}
else
{
return j;
}
}
}
// Cannot be called directly! use macro CHECK_SURFACE_TYPE!
bool _CheckSurfaceType( int nType, ... )
{
bool match = false;
va_list ap;
va_start( ap, nType );
int type = 0;
while ( ( type = va_arg( ap, int ) ) >= 0 )
{
if( type == nType )
{
match = true;
break;
}
}
va_end(ap);
return match;
}
void QuickSort( PCM_ARG* args, int32_t p, int32_t r )
{
if( p < r )
{
int32_t q = Partition( args, p, r );
QuickSort( args, p, q );
QuickSort( args, q + 1, r );
}
}
namespace CMRT_UMD
{
static bool bCmMovInstRegistered = CmExtensionCreator<CmMovInstConstructor>::RegisterClass<CmMovInstConstructor>();
//*-----------------------------------------------------------------------------
//| Purpose: Create object for mov instructions
//| instructions will be copied into DstMem
//*-----------------------------------------------------------------------------
uint32_t CmMovInstConstructor::ConstructObjMovs(uint32_t dstOffset, uint32_t srcOffset, uint32_t size, CmDynamicArray &movInsts, uint32_t index, bool isBdw, bool isHwDebug)
{
return MovInst_RT::CreateMoves(dstOffset, srcOffset, size, movInsts, index, isBdw, isHwDebug);
}
//*-----------------------------------------------------------------------------
//| Purpose: Create CM Kernel
//| Arguments :
//| device [in] Pointer to device
//| program [in] Pointer to cm Program
//| kernelName [in] Name of kernel
//| kernelId [in] Kernel's ID
//| kernel [in/out] Reference Pointer to CM Kernel
//| options [in] jitter, or non-jitter
//| Returns: Result of the operation.
//*-----------------------------------------------------------------------------
int32_t CmKernelRT::Create(CmDeviceRT *device,
CmProgramRT *program,
const char *kernelName,
uint32_t kernelIndex,
uint32_t kernelSeqNum,
CmKernelRT* &kernel,
const char *options)
{
int32_t result = CM_SUCCESS;
CM_HAL_STATE * state = device ? ((PCM_CONTEXT_DATA)device->GetAccelData())->cmHalState : nullptr;
if (device)
{
if (state && state->advExecutor)
{
kernel = state->advExecutor->CreateKernelRT(device, program, kernelIndex, kernelSeqNum);
}
else
{
kernel = new (std::nothrow) CmKernelRT(device, program, kernelIndex, kernelSeqNum);
}
}
if( kernel )
{
if (device)
{
device->m_memObjectCount.kernelCount++;
}
kernel->Acquire();
result = kernel->Initialize( kernelName, options );
if( result != CM_SUCCESS )
{
CmKernelRT::Destroy( kernel, program);
return result;
}
}
else
{
CM_ASSERTMESSAGE("Error: Failed to create CmKernel due to out of system memory.");
return CM_OUT_OF_HOST_MEMORY;
}
if (options)
{
if (strcmp(options, "PredefinedGPUCopyKernel") == 0)
{
kernel->m_blCreatingGPUCopyKernel = true;
}
else
{
kernel->m_blCreatingGPUCopyKernel = false;
}
}
#if USE_EXTENSION_CODE
if (device)
result = kernel->InitForGTPin(device, program, kernel);
#endif
return result;
}
//*-----------------------------------------------------------------------------
//| Purpose: Destory Kernel
//| Returns: Result of the operation.
//*-----------------------------------------------------------------------------
int32_t CmKernelRT::Destroy( CmKernelRT* &kernel, CmProgramRT *&program )
{
uint32_t refCount = kernel->SafeRelease();
if (refCount == 0)
{
kernel = nullptr;
}
refCount = program->SafeRelease();
if (refCount == 0)
{
program = nullptr;
}
return CM_SUCCESS;
}
//*-----------------------------------------------------------------------------
//| Purpose: Acuqire Kernel: increment refcount
//| Returns: Result of the operation.
//*-----------------------------------------------------------------------------
int32_t CmKernelRT::Acquire( void)
{
m_refcount ++;
return m_refcount;
}
//*-----------------------------------------------------------------------------
//| Purpose: SafeRelease Kernel: Delete the instance
//| Returns: Result of the operation.
//*-----------------------------------------------------------------------------
int32_t CmKernelRT::SafeRelease( void)
{
--m_refcount;
if (m_refcount == 0)
{
m_device->m_memObjectCount.kernelCount--;
PCM_CONTEXT_DATA cmData = (PCM_CONTEXT_DATA)m_device->GetAccelData();
PCM_HAL_STATE state = cmData->cmHalState;
if (state->dshEnabled)
{
state->pfnDSHUnregisterKernel(state, m_id);
}
delete this;
return 0;
}
return m_refcount;
}
//*-----------------------------------------------------------------------------
//| Purpose: Kernel constructor
//| Returns: Result of the operation.
//*-----------------------------------------------------------------------------
CmKernelRT::CmKernelRT(CmDeviceRT *device,
CmProgramRT *program,
uint32_t kernelIndex,
uint32_t kernelSeqNum):
m_device( device ),
m_surfaceMgr( nullptr ),
m_program( program ),
m_options( nullptr ),
m_binary( nullptr ),
m_binaryOrig(nullptr),
m_binarySize(0),
m_binarySizeOrig(0),
m_threadCount( 0 ),
m_lastThreadCount( 0 ),
m_sizeInCurbe( 0 ),
m_sizeInPayload( 0 ),
m_argCount( 0 ),
m_args( nullptr ),
m_kernelInfo(nullptr),
m_kernelIndexInProgram( CM_INVALID_KERNEL_INDEX ),
m_curbeEnabled( true ),
m_nonstallingScoreboardEnabled(false),
m_dirty( CM_KERNEL_DATA_CLEAN ),
m_lastKernelData( nullptr ),
m_lastKernelDataSize( 0 ),
m_indexInTask(0),
m_threadSpaceAssociated(false),
m_perThreadArgExists(false),
m_perKernelArgExists( false ),
m_threadSpace( nullptr ),
m_adjustScoreboardY( 0 ),
m_lastAdjustScoreboardY( 0 ),
m_blCreatingGPUCopyKernel( false),
m_usKernelPayloadDataSize( 0 ),
m_kernelPayloadData( nullptr ),
m_usKernelPayloadSurfaceCount( 0 ),
m_samplerBtiCount( 0 ),
m_refcount(0),
m_halMaxValues( nullptr ),
m_halMaxValuesEx( nullptr ),
m_surfaceArray(nullptr),
m_threadGroupSpace( nullptr ),
m_vmeSurfaceCount( 0 ),
m_maxSurfaceIndexAllocated(0),
m_barrierMode(CM_LOCAL_BARRIER),
m_isClonedKernel(false),
m_cloneKernelID(0),
m_hasClones( false ),
m_stateBufferBounded( CM_STATE_BUFFER_NONE ),
m_movInstConstructor(nullptr)
{
program->Acquire();
m_program = program;
device->GetSurfaceManager(m_surfaceMgr);
m_id = kernelSeqNum; // Unique number for each kernel. This ID is used in Batch buffer.
m_id <<= 32;
m_kernelIndex = kernelIndex;
for (int i = 0; i < CM_GLOBAL_SURFACE_NUMBER; i++)
{
m_globalSurfaces[i] = nullptr;
m_globalCmIndex[i] = 0;
}
m_blhwDebugEnable = program->IsHwDebugEnabled();
CmSafeMemSet(m_pKernelPayloadSurfaceArray, 0, sizeof(m_pKernelPayloadSurfaceArray));
CmSafeMemSet(m_IndirectSurfaceInfoArray, 0, sizeof(m_IndirectSurfaceInfoArray));
CmSafeMemSet( m_samplerBtiEntry, 0, sizeof( m_samplerBtiEntry ) );
if (m_samplerBtiCount > 0)
{
CmSafeMemSet(m_samplerBtiEntry, 0, sizeof(m_samplerBtiEntry));
m_samplerBtiCount = 0;
}
ResetKernelSurfaces();
}
//*-----------------------------------------------------------------------------
//| Purpose: Destructor of Class CmKernel
//| Returns: None.
//*-----------------------------------------------------------------------------
CmKernelRT::~CmKernelRT( void )
{
MosSafeDeleteArray(m_options);
DestroyArgs();
if(m_lastKernelData)
{
CmKernelData::Destroy( m_lastKernelData );
}
if( m_device->CheckGTPinEnabled() && !m_blCreatingGPUCopyKernel)
{
MosSafeDeleteArray(m_binary);
}
if( CM_INVALID_KERNEL_INDEX != m_kernelIndexInProgram )
{
m_program->ReleaseKernelInfo(m_kernelIndexInProgram);
}
for(int i=0; i< CM_GLOBAL_SURFACE_NUMBER; i++)
{
SurfaceIndex *surfIndex = m_globalSurfaces[i];
MosSafeDelete(surfIndex);
}
MosSafeDeleteArray(m_kernelPayloadData);
MosSafeDeleteArray(m_surfaceArray);
MosSafeDelete(m_movInstConstructor);
}
//*-----------------------------------------------------------------------------
//| Purpose: Initialize CM kernel
//| Returns: Result of the operation.
//*-----------------------------------------------------------------------------
int32_t CmKernelRT::Initialize( const char* kernelName, const char* options )
{
if( kernelName == nullptr )
{
CM_ASSERTMESSAGE("Error: Kernel name is null.");
return CM_NULL_POINTER;
}
size_t length = strnlen( kernelName, CM_MAX_KERNEL_NAME_SIZE_IN_BYTE );
if( length >= CM_MAX_KERNEL_NAME_SIZE_IN_BYTE )
{
CM_ASSERTMESSAGE("Error: Kernel name size is too long.");
return CM_FAILURE;
}
uint32_t kernelCount = 0;
m_program->GetKernelCount( kernelCount );
CM_KERNEL_INFO* kernelInfo = nullptr;
uint32_t i = 0;
for( i = 0; i < kernelCount; i ++ )
{
m_program->GetKernelInfo( i, kernelInfo );
if( !kernelInfo )
{
CM_ASSERTMESSAGE("Error: Invalid kernel info.");
return CM_NULL_POINTER;
}
if( strcmp( kernelInfo->kernelName, kernelName ) == 0 )
{
break;
}
}
if( i == kernelCount )
{
CM_ASSERTMESSAGE("Error: Invalid kernel count.");
return CM_FAILURE;
}
m_device->GetHalMaxValues( m_halMaxValues, m_halMaxValuesEx);
m_program->AcquireKernelInfo(i);
m_kernelInfo = kernelInfo;
m_kernelIndexInProgram = i;
if( options )
{
size_t length = strnlen( options, CM_MAX_OPTION_SIZE_IN_BYTE );
if(length >= CM_MAX_OPTION_SIZE_IN_BYTE)
{
CM_ASSERTMESSAGE("Error: Option string is too long.");
return CM_INVALID_ARG_VALUE;
}
else
{
m_options = MOS_NewArray(char, (length+1));
if( !m_options )
{
CM_ASSERTMESSAGE("Error: Out of system memory.");
return CM_OUT_OF_HOST_MEMORY;
}
CmSafeMemCopy( m_options, options, length);
m_options[ length ] = '\0';
char* tmp = strstr( m_options, "nocurbe" );
if( tmp )
{
m_curbeEnabled = false;
}
}
}
m_nonstallingScoreboardEnabled = true;
void* commonISACode = nullptr;
uint32_t commonISACodeSize = 0;
m_program->GetCommonISACode(commonISACode, commonISACodeSize);
if ((commonISACode == nullptr) || (commonISACodeSize <= 0))
{
CM_ASSERTMESSAGE("Error: Invalid VISA.");
return CM_INVALID_COMMON_ISA;
}
bool useVisaApi = true;
vISA::ISAfile *isaFile = nullptr;
vISA::KernelBody *kernelBody = nullptr;
auto getVersionAsInt = [](int major, int minor) { return major * 100 + minor; };
if (getVersionAsInt(m_program->m_cisaMajorVersion, m_program->m_cisaMinorVersion) < getVersionAsInt(3, 2))
{
useVisaApi = false;
}
else
{
isaFile = m_program->getISAfile();
if (!isaFile)
{
CM_ASSERTMESSAGE("Error: Invalid VISA.");
return CM_INVALID_COMMON_ISA;
}
kernelBody = isaFile->getKernelsData().at(m_kernelIndexInProgram);
}
uint8_t *buf = (uint8_t*)commonISACode;
uint32_t bytePosition = m_kernelInfo->kernelIsaOffset;
uint32_t kernelInfoRefCount = 0;
m_program->GetKernelInfoRefCount(m_kernelIndexInProgram, kernelInfoRefCount);
if (kernelInfoRefCount <= 2) //Only read for 1st time Kernel creation, later we reuse them
{
if (useVisaApi)
{
m_kernelInfo->globalStringCount = kernelBody->getStringCount();
}
{
READ_FIELD_FROM_BUF(m_kernelInfo->globalStringCount, unsigned short);
}
m_kernelInfo->globalStrings = (const char**) malloc( m_kernelInfo->globalStringCount * sizeof(char*) );
if(m_kernelInfo->globalStrings == nullptr)
{
CM_ASSERTMESSAGE("Error: Out of system memory.");
return CM_OUT_OF_HOST_MEMORY;
}
CmSafeMemSet(m_kernelInfo->globalStrings, 0, m_kernelInfo->globalStringCount * sizeof(char*) );
if (useVisaApi)
{
int i = 0;
for (vISA::StringPool *globalString : kernelBody->getStringPool())
{
size_t stringLength = std::strlen(globalString->getString());
char *string = (char*)malloc(stringLength + 1);
if (string == nullptr)
{
CM_ASSERTMESSAGE("Error: Out of system memory.");
return CM_OUT_OF_HOST_MEMORY;
}
CmSafeMemCopy(string, globalString->getString(), stringLength);
string[stringLength] = '\0';
m_kernelInfo->globalStrings[i] = string;
i++;
}
}
else
{
for (int i = 0; i < (int)m_kernelInfo->globalStringCount; i++)
{
char* string = (char*)malloc(CM_MAX_KERNEL_STRING_IN_BYTE + 1);
if (string == nullptr)
{
CM_ASSERTMESSAGE("Error: Out of system memory.");
return CM_OUT_OF_HOST_MEMORY;
}
int j = 0;
while (buf[bytePosition] != '\0' && j < CM_MAX_KERNEL_STRING_IN_BYTE) {
string[j++] = buf[bytePosition++];
}
string[j] = '\0';
bytePosition++;
m_kernelInfo->globalStrings[i] = string;
}
}
}
uint32_t count = 0;
if (useVisaApi)
{
count = kernelBody->getNumInputs();
}
else
{
bytePosition = m_kernelInfo->inputCountOffset;
uint8_t countTemp = 0;
READ_FIELD_FROM_BUF(countTemp, uint8_t);
count = countTemp;
}
if( count > m_halMaxValues->maxArgsPerKernel )
{
CM_ASSERTMESSAGE("Error: Invalid kernel arg count.");
return CM_EXCEED_KERNEL_ARG_AMOUNT;
}
m_args = MOS_NewArray(CM_ARG, count);
if( (!m_args) && (count != 0) )
{
CM_ASSERTMESSAGE("Error: Out of system memory.");
MosSafeDeleteArray(m_options);
return CM_OUT_OF_HOST_MEMORY;
}
CmSafeMemSet(m_args, 0, sizeof(CM_ARG) * count);
m_argCount = count;
uint32_t preDefinedSurfNum;
if ( (m_program->m_cisaMajorVersion > 3) || ((m_program->m_cisaMajorVersion == 3) && (m_program->m_cisaMinorVersion >=1)) ) //CISA 3.1 +
{
preDefinedSurfNum = COMMON_ISA_NUM_PREDEFINED_SURF_VER_3_1;
}
else if ((m_program->m_cisaMajorVersion == 3) && (m_program->m_cisaMinorVersion == 0))
{
preDefinedSurfNum = COMMON_ISA_NUM_PREDEFINED_SURF_VER_2_1;
}
else //CISA 2.0
{
preDefinedSurfNum = COMMON_ISA_NUM_PREDEFINED_SURF_VER_2;
}
uint32_t argSize = 0;
for (uint32_t i = 0; i < m_argCount; i++)
{
vISA::InputInfo *inputInfo = nullptr;
uint8_t kind = 0;
if (useVisaApi)
{
inputInfo = kernelBody->getInputInfo()[i];
kind = inputInfo->getKind();
}
else
{
READ_FIELD_FROM_BUF(kind, uint8_t);
}
if (kind == 0x2) // compiler value for surface
{
kind = ARG_KIND_SURFACE;
// runtime value for surface. surface will be further classified to 1D/2D/3D
}
else if (kind == 0x3) // compiler value for vme index
{
kind = ARG_KIND_VME_INDEX;
}
else if (kind == 0x8)
{
kind = ARG_KIND_IMPLICT_LOCALSIZE;
m_args[i].isSet = true;
m_args[i].unitCount = 1;
}
else if (kind == 0x10) {
kind = ARG_KIND_IMPLICT_GROUPSIZE;
m_args[i].isSet = true;
m_args[i].unitCount = 1;
}
else if (kind == 0x18) {
kind = ARG_KIND_IMPLICIT_LOCALID;
m_args[i].isSet = true;
m_args[i].unitCount = 1;
m_perKernelArgExists = true; //only VISA3.3+, can come here, so, no matter it is there any explicit arg, implicit arg exits
}
else if (kind == 0x2A) {
kind = ARG_KIND_SURFACE_2D_SCOREBOARD;
}
else if (kind == 0x20) {
kind = ARG_KIND_GENERAL_DEPVEC;
}
else if (kind == 0x30) {
kind = ARG_KIND_GENERAL_DEPCNT;
}
else if (kind == 0x80) {
// IMP_PSEUDO_INPUT = 0x80 is pseudo input. All inputs after this
// will be ignored by CMRT without checking and payload copied.
// This resizes the argument count to achieve this.
m_argCount = i;
break;
}
m_args[i].unitKind = kind;
m_args[i].unitKindOrig = kind;
if (kind == ARG_KIND_SURFACE && m_kernelInfo->surfaceCount)
{
m_args[i].surfaceKind = DATA_PORT_SURF;
}
if (useVisaApi)
{
m_args[i].unitOffsetInPayload = inputInfo->getOffset();
m_args[i].unitOffsetInPayloadOrig = inputInfo->getOffset();
m_args[i].unitSize = inputInfo->getSize();
m_args[i].unitSizeOrig = inputInfo->getSize();
}
else
{
uint32_t varID;
READ_FIELD_FROM_BUF(varID, uint16_t);
uint16_t tmpW;
READ_FIELD_FROM_BUF(tmpW, uint16_t);
m_args[i].unitOffsetInPayload = tmpW;
m_args[i].unitOffsetInPayloadOrig = tmpW;
READ_FIELD_FROM_BUF(tmpW, uint16_t);
m_args[i].unitSize = tmpW;
m_args[i].unitSizeOrig = tmpW;
}
argSize += m_args[i].unitSize;
}
//////////////////////////////////////////////////////////////////////////
if (kernelInfoRefCount <= 2) //Only read for 1st time Kernel creation, later we reuse them
{
uint16_t attributeCount = 0;
if (useVisaApi)
{
attributeCount = kernelBody->getAttributeCount();
}
else
{
/////////////////////////////////////////////////////////////////////////
// Get pre-defined kernel attributes, Start
//skipping size and entry
bytePosition += 8;
READ_FIELD_FROM_BUF(attributeCount, uint16_t);
}
for (int i = 0; i < attributeCount; i++)
{
vISA::AttributeInfo *attribute = nullptr;
uint32_t nameIndex = 0;
uint8_t size = 0;
if (useVisaApi)
{
attribute = kernelBody->getAttributeInfo()[i];
nameIndex = attribute->getName();
size = attribute->getSize();
}
else
{
READ_FIELD_FROM_BUF(nameIndex, uint16_t);
READ_FIELD_FROM_BUF(size, uint8_t);
}
if( strcmp( m_kernelInfo->globalStrings[nameIndex], "AsmName" ) == 0 )
{
if (useVisaApi)
{
CmSafeMemCopy(m_kernelInfo->kernelASMName, attribute->getValue(), size);
}
else
{
CmSafeMemCopy(m_kernelInfo->kernelASMName, &buf[bytePosition], size);
bytePosition += size;
}
}
else if (strcmp( m_kernelInfo->globalStrings[nameIndex], "SLMSize" ) == 0)
{
if (useVisaApi)
{
m_kernelInfo->kernelSLMSize = attribute->getValue()[0];
}
else
{
READ_FIELD_FROM_BUF(m_kernelInfo->kernelSLMSize, uint8_t);
}
/* Notes by Stony@2014-04-09
* <=CISA3.1: the size is number of 4KB
* > CISA3.1: the size is number of 1KB
* Here convert it to the number of 1KB if <=CISA 3.1
*/
if ((m_program->m_cisaMajorVersion == 3) && (m_program->m_cisaMinorVersion <= 1))
{
m_kernelInfo->kernelSLMSize = m_kernelInfo->kernelSLMSize * 4;
}
// align to power of 2
uint32_t v = m_kernelInfo->kernelSLMSize;
v--;
v |= v >> 1;
v |= v >> 2;
v |= v >> 4;
v |= v >> 8;
v |= v >> 16;
v++;
m_kernelInfo->kernelSLMSize = ( uint8_t )v;
}
else if (strcmp(m_kernelInfo->globalStrings[nameIndex], "NoBarrier") == 0)
{
m_kernelInfo->blNoBarrier = true;
if (!useVisaApi)
{
bytePosition += size;
}
}
else
{
if (!useVisaApi)
{
bytePosition += size;
}
}
}
if (m_kernelInfo->blNoBarrier && m_options && strstr(m_options, "-hasBarrier"))
{
m_kernelInfo->blNoBarrier = false;
}
}
if(argSize > m_halMaxValues->maxArgByteSizePerKernel)
{
CM_ASSERTMESSAGE("Error: Invalid kernel arg size.");
return CM_EXCEED_KERNEL_ARG_SIZE_IN_BYTE;
}
buf = (uint8_t*)commonISACode;
if(m_program->IsJitterEnabled())
{
//m_JitterEnable = true;
char *programOptions;
m_program->GetKernelOptions(programOptions);
//if no options or same options, copy load program's binary. else re-jitter
{
m_binary = (char *)m_kernelInfo->jitBinaryCode;
m_binarySize = m_kernelInfo->jitBinarySize;
m_kernelInfo->origBinary = m_kernelInfo->jitBinaryCode;
m_kernelInfo->origBinarySize = m_kernelInfo->jitBinarySize;
}
}
else
{
char* binary = (char*)(buf + m_kernelInfo->genxBinaryOffset );
//No copy, point to the binary offset in CISA code.
m_binary = binary;
m_binarySize = m_kernelInfo->genxBinarySize;
m_kernelInfo->origBinary = binary;
m_kernelInfo->origBinarySize = m_kernelInfo->genxBinarySize;
}
if (m_kernelInfo->blNoBarrier)
{
m_barrierMode = CM_NO_BARRIER;
}
m_movInstConstructor = CmExtensionCreator<CmMovInstConstructor>::CreateClass();
if (m_movInstConstructor == nullptr)
{
CM_ASSERTMESSAGE("Error: Failed to allocate movInstConstructor due to out of system memory.");
return CM_OUT_OF_HOST_MEMORY;
}
CmNotifierGroup *notifiers = m_device->GetNotifiers();
if (notifiers != nullptr)
{
notifiers->NotifyKernelCreated(this);
}
return CM_SUCCESS;
}
//*-----------------------------------------------------------------------------
//! A CmKernel can run in multiple threads concurrently. This
//! fucntion is to set the number of threads.
//! INPUT:
//! number of threads
//! OUTPUT:
//! CM_SUCCESS or
//! CM_INVALID_ARG_VALUE if the number is larger than CmKernel's capacity
//*-----------------------------------------------------------------------------
CM_RT_API int32_t CmKernelRT::SetThreadCount(uint32_t count )
{
INSERT_API_CALL_LOG(GetHalState());
// Check per kernel, per task check will be at enqueue time
if ((int)count <= 0)
return CM_INVALID_ARG_VALUE;
if (m_threadSpace == nullptr)
{
if (m_threadCount)
{
// Setting threadCount twice with different values will cause reset of kernels
if (m_threadCount != count)
{
Reset();
m_threadCount = count;
m_dirty |= CM_KERNEL_DATA_THREAD_COUNT_DIRTY;
}
}
else // first time
{
m_threadCount = count;
}
}
return CM_SUCCESS;
}
int32_t CmKernelRT::GetThreadCount(uint32_t& count ) const
{
count = m_threadCount;
return CM_SUCCESS;
}
int32_t CmKernelRT::GetKernelSurfaces(bool *&surfArray) const
{
surfArray = m_surfaceArray;
return CM_SUCCESS;
}
int32_t CmKernelRT::ResetKernelSurfaces()
{
uint32_t surfacePoolSize = m_surfaceMgr->GetSurfacePoolSize();
if (!m_surfaceArray)
{
m_surfaceArray = MOS_NewArray(bool, surfacePoolSize);
if (!m_surfaceArray)
{
CM_ASSERTMESSAGE("Error: Failed to rest kernel surfaces due to out of system memory.");
return CM_OUT_OF_HOST_MEMORY;
}
}
CmSafeMemSet( m_surfaceArray, 0, surfacePoolSize * sizeof( bool ) );
return CM_SUCCESS;
}
//*-----------------------------------------------------------------------------
//| Purpose: Get CmSurface from surface manager.
//| Use "value + indexSurfaceArray" to locate its surfaceIndex
//| Returns: CmSurface. Null if not found
//*-----------------------------------------------------------------------------
CmSurface* CmKernelRT::GetSurfaceFromSurfaceArray( SurfaceIndex* value, uint32_t indexSurfaceArray) const
{
int32_t hr = CM_SUCCESS;
CmSurface *surface = nullptr;
SurfaceIndex* surfaceIndex = nullptr;
surfaceIndex = value + indexSurfaceArray;
CM_CHK_NULL_GOTOFINISH_CMERROR(surfaceIndex);
if (surfaceIndex->get_data() == CM_NULL_SURFACE
|| surfaceIndex->get_data() == 0)
{
surface = (CmSurface *)CM_NULL_SURFACE;
goto finish;
}
m_surfaceMgr->GetSurface(surfaceIndex->get_data(), surface);
finish:
return surface;
}
//*-----------------------------------------------------------------------------
//| Purpose: Set kernel arg for single vme surface or multiple vme surfaces
//| in surface array. So far, don't support vme surface array in thread arg.
//| Returns: Result of the operation.
//*-----------------------------------------------------------------------------
int32_t CmKernelRT::SetArgsVme(CM_KERNEL_INTERNAL_ARG_TYPE nArgType, uint32_t argIndex, const void *value, uint32_t nThreadID)
{
uint32_t elementNum = 0;
CM_ARG& arg = m_args[ argIndex ];
uint32_t totalVmeArgValueSize = 0;
uint32_t totalSurfacesInVme = 0;
uint32_t tempVmeArgValueSize = 0;
uint32_t vmeArgValueOffset = 0;
uint32_t lastVmeSurfCount = 0;
CmSurfaceVme* surfVme = nullptr;
uint8_t *vmeArgValueArray = nullptr;
uint16_t *vmeCmIndexArray = nullptr;
int32_t hr = CM_SUCCESS;
//Get Number of elements in surface array
if (arg.unitVmeArraySize == 0)
{ //First Time
elementNum = arg.unitSize / sizeof(uint32_t);
}
else
{
elementNum = arg.unitVmeArraySize;
}
//Get Size of vmeIndexArray and vmeCmIndexArray.
for(uint32_t i=0; i< elementNum; i++)