| 1 |
/* |
| 2 |
* Copyright (c) 2018-2019, Intel Corporation |
| 3 |
* |
| 4 |
* Permission is hereby granted, free of charge, to any person obtaining a |
| 5 |
* copy of this software and associated documentation files (the "Software"), |
| 6 |
* to deal in the Software without restriction, including without limitation |
| 7 |
* the rights to use, copy, modify, merge, publish, distribute, sublicense, |
| 8 |
* and/or sell copies of the Software, and to permit persons to whom the |
| 9 |
* Software is furnished to do so, subject to the following conditions: |
| 10 |
* |
| 11 |
* The above copyright notice and this permission notice shall be included |
| 12 |
* in all copies or substantial portions of the Software. |
| 13 |
* |
| 14 |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
| 15 |
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 16 |
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
| 17 |
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR |
| 18 |
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, |
| 19 |
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
| 20 |
* OTHER DEALINGS IN THE SOFTWARE. |
| 21 |
*/ |
| 22 |
//! |
| 23 |
//! \file cm_kernel_ex.h |
| 24 |
//! \brief Contains Class CmKernelEx definitions |
| 25 |
//! |
| 26 |
#pragma once |
| 27 |
|
| 28 |
#include "cm_kernel_rt.h" |
| 29 |
#include "cm_group_space.h" |
| 30 |
|
| 31 |
class CmSurfaceState; |
| 32 |
class CmSSH; |
| 33 |
class CmISHBase; |
| 34 |
class CmMediaState; |
| 35 |
namespace CMRT_UMD |
| 36 |
{ |
| 37 |
class CmThreadSpaceRT; |
| 38 |
class CmThreadGroupSpace; |
| 39 |
}; |
| 40 |
class CmKernelEx : public CMRT_UMD::CmKernelRT |
| 41 |
{ |
| 42 |
public: |
| 43 |
CmKernelEx(CMRT_UMD::CmDeviceRT *device, |
| 44 |
CMRT_UMD::CmProgramRT *program, |
| 45 |
uint32_t kernelIndex, |
| 46 |
uint32_t kernelSeqNum): |
| 47 |
CMRT_UMD::CmKernelRT(device, program, kernelIndex, kernelSeqNum), |
| 48 |
m_indexMap(nullptr), |
| 49 |
m_flatArgs(nullptr), |
| 50 |
m_flatArgCount(0), |
| 51 |
m_data(nullptr), |
| 52 |
m_surfaceInArg(nullptr), |
| 53 |
m_curbe(nullptr), |
| 54 |
m_explicitCurbeSize(0), |
| 55 |
m_curbeSize(0), |
| 56 |
m_curbeSizePerThread(0), |
| 57 |
m_curbeSizeCrossThread(0), |
| 58 |
m_propertyIndexes(nullptr), |
| 59 |
m_cmSurfIndexes(nullptr), |
| 60 |
m_hashValue(0), |
| 61 |
m_dummyThreadSpace(nullptr), |
| 62 |
m_dummyThreadGroupSpace(nullptr), |
| 63 |
m_ish(nullptr), |
| 64 |
m_indexInIsh(-1), |
| 65 |
m_offsetInIsh(0) |
| 66 |
{ |
| 67 |
} |
| 68 |
|
| 69 |
virtual ~CmKernelEx(); |
| 70 |
|
| 71 |
int32_t SetKernelArg(uint32_t index, size_t size, const void * value); |
| 72 |
|
| 73 |
int32_t SetSurfaceBTI(CMRT_UMD::SurfaceIndex *surfIndex, uint32_t bti); |
| 74 |
|
| 75 |
int32_t SetSamplerBTI(CMRT_UMD::SamplerIndex* sampler, uint32_t nIndex); |
| 76 |
|
| 77 |
int32_t SetStaticBuffer(uint32_t index, const void *value); |
| 78 |
|
| 79 |
inline uint32_t GetCurbeSize() {return m_curbeSize; } |
| 80 |
|
| 81 |
inline uint32_t GetCurbeSizePerThread() {return m_curbeSizePerThread; } |
| 82 |
|
| 83 |
inline uint32_t GetCurbeSizeCrossThread() {return m_curbeSizeCrossThread; } |
| 84 |
|
| 85 |
inline uint8_t* GetCurbe() {return m_curbe; } |
| 86 |
|
| 87 |
uint32_t GetMaxBteNum(); |
| 88 |
|
| 89 |
MOS_STATUS AllocateCurbe(); // for media walker |
| 90 |
|
| 91 |
MOS_STATUS AllocateCurbeAndFillImplicitArgs(CMRT_UMD::CmThreadGroupSpace *globalGroupSpace); // for gpgpu/compute walker |
| 92 |
|
| 93 |
MOS_STATUS UpdateCurbe(CmSSH *ssh, CmMediaState *mediaState, uint32_t kernelIdx); |
| 94 |
|
| 95 |
MOS_STATUS UpdateSWSBArgs(CMRT_UMD::CmThreadSpaceRT *threadSpace); |
| 96 |
|
| 97 |
MOS_STATUS UpdateFastTracker(uint32_t trackerIndex, uint32_t tracker); |
| 98 |
|
| 99 |
MOS_STATUS LoadReservedSurfaces(CmSSH *ssh); |
| 100 |
|
| 101 |
MOS_STATUS LoadReservedSamplers(CmMediaState *mediaState, uint32_t kernelIdx); |
| 102 |
|
| 103 |
MOS_STATUS GetSamplerCount(uint32_t *count3D, uint32_t *countAVS); |
| 104 |
|
| 105 |
CMRT_UMD::CmThreadSpaceRT* GetThreadSpaceEx(); |
| 106 |
|
| 107 |
CMRT_UMD::CmThreadGroupSpace* GetThreadGroupSpaceEx(); |
| 108 |
|
| 109 |
void SurfaceDumpEx(uint32_t kernelNumber, int32_t taskId); |
| 110 |
|
| 111 |
bool IsFastPathSupported(); |
| 112 |
|
| 113 |
// interfaces with ISH |
| 114 |
inline void Recorded(CmISHBase *ish, int index, uint32_t offset) |
| 115 |
{ |
| 116 |
m_ish = ish; |
| 117 |
m_indexInIsh = index; |
| 118 |
m_offsetInIsh = offset; |
| 119 |
} |
| 120 |
|
| 121 |
inline void GetRecordedInfo(CmISHBase **ish, int *index) |
| 122 |
{ |
| 123 |
*ish = m_ish; |
| 124 |
*index = m_indexInIsh; |
| 125 |
} |
| 126 |
|
| 127 |
inline uint32_t GetOffsetInIsh() {return m_offsetInIsh; } |
| 128 |
|
| 129 |
inline void GetNativeKernel(uint8_t **kernel, uint32_t *size) |
| 130 |
{ |
| 131 |
*kernel = (uint8_t *)m_binary; |
| 132 |
*size = m_binarySize; |
| 133 |
} |
| 134 |
|
| 135 |
inline uint32_t GetBarrierMode() {return m_barrierMode; } |
| 136 |
|
| 137 |
inline std::map<int, CmSurfaceState *>& GetReservedSurfaceBteIndex() {return m_reservedSurfaceBteIndexes; } |
| 138 |
|
| 139 |
inline std::map<int, void *>& GetReservedSamplerBteIndex() {return m_reservedSamplerBteIndexes; } |
| 140 |
|
| 141 |
inline uint64_t GetHashValue() {return m_hashValue; } |
| 142 |
|
| 143 |
protected: |
| 144 |
int32_t Initialize(const char *kernelName, const char *options); |
| 145 |
|
| 146 |
bool IsSurface(uint16_t kind); |
| 147 |
|
| 148 |
CM_ARG_KIND ToArgKind(CMRT_UMD::CmSurface *surface); |
| 149 |
|
| 150 |
inline bool ArgArraySupported(uint16_t kind) {return IsSurface(kind) || kind == ARG_KIND_SAMPLER; } |
| 151 |
|
| 152 |
CmSurfaceState* GetSurfaceState(CMRT_UMD::CmSurface *surface, uint32_t index); |
| 153 |
|
| 154 |
void* GetSamplerParam(uint32_t index); |
| 155 |
|
| 156 |
struct _CmArg |
| 157 |
{ |
| 158 |
uint32_t offset; |
| 159 |
uint16_t kind; |
| 160 |
uint16_t isaKind; |
| 161 |
uint16_t unitSize; |
| 162 |
uint16_t payloadOffset; |
| 163 |
uint16_t sizeInCurbe; |
| 164 |
bool isSet; |
| 165 |
}; |
| 166 |
|
| 167 |
uint32_t *m_indexMap; |
| 168 |
_CmArg *m_flatArgs; |
| 169 |
uint32_t m_flatArgCount; |
| 170 |
uint8_t *m_data; |
| 171 |
uint8_t *m_surfaceInArg; |
| 172 |
|
| 173 |
uint8_t *m_curbe; |
| 174 |
uint32_t m_explicitCurbeSize; |
| 175 |
uint32_t m_curbeSize; |
| 176 |
uint32_t m_curbeSizePerThread; |
| 177 |
uint32_t m_curbeSizeCrossThread; |
| 178 |
|
| 179 |
uint8_t *m_propertyIndexes; |
| 180 |
uint32_t *m_cmSurfIndexes; |
| 181 |
|
| 182 |
uint64_t m_hashValue; |
| 183 |
|
| 184 |
// reserved an internal space for setThreadCount |
| 185 |
CMRT_UMD::CmThreadSpace *m_dummyThreadSpace; |
| 186 |
CMRT_UMD::CmThreadGroupSpace *m_dummyThreadGroupSpace; |
| 187 |
|
| 188 |
//recorded info in ish |
| 189 |
CmISHBase *m_ish; |
| 190 |
int m_indexInIsh; |
| 191 |
uint32_t m_offsetInIsh; |
| 192 |
|
| 193 |
// reserved surface bte index |
| 194 |
std::map<int, CmSurfaceState *>m_reservedSurfaceBteIndexes; |
| 195 |
|
| 196 |
// reserved sampler bte index |
| 197 |
std::map<int, void *>m_reservedSamplerBteIndexes; |
| 198 |
}; |