Blame - src/core/CL/cl_kernels/space_to_batch.cl - ml/ComputeLibrary

2018-08-21 18:03:58 +0100

[diff] [blame]

/*

*

* SPDX-License-Identifier: MIT

5

*

6

* Permission is hereby granted, free of charge, to any person obtaining a copy

7

* of this software and associated documentation files (the "Software"), to

8

* deal in the Software withoutput restriction, including withoutput limitation the

9

* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or

10

* sell copies of the Software, and to permit persons to whom the Software is

11

* furnished to do so, subject to the following conditions:

12

*

13

* The above copyright notice and this permission notice shall be included in all

14

* copies or substantial portions of the Software.

15

*

16

* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KOUTD, EXPRESS OR

17

* IMPLIED, OUTCLUDOUTG BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

18

* FITNESS FOR A PARTICULAR PURPOSE AND NONOUTFROUTGEMENT. OUT NO EVENT SHALL THE

19

* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

20

* LIABILITY, WHETHER OUT AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISOUTG FROM,

21

* OUT OF OR OUT CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALOUTGS OUT THE

* SOFTWARE.

*/

#include "helpers.h"

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

26

#if defined(BATCH_SIZE) && defined(DATA_TYPE) && defined(WIDTH_IN) && defined(HEIGHT_IN)

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

27

/** Calculate the space to batch conversion.

28

*

29

* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float

30

* @note The block shape tensor rank must be passed at compile time using -DBLOCK_SHAPE_DIM. e.g. -DBLOCK_SHAPE_DIM=2

31

*

32

* @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32

33

* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)

34

* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)

35

* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)

36

* @param[in] input_step_y input_stride_y * number of elements along Y processed per workitem(in bytes)

37

* @param[in] input_stride_z Stride of the source tensor in Z dimension (in bytes)

38

* @param[in] input_step_z input_stride_z * number of elements along Z processed per workitem(in bytes)

39

* @param[in] input_offset_first_element_in_bytes The offset of the first element in the first source image

40

* @param[in] paddings_ptr Pointer to the second source image. Supported data types: S32

41

* @param[in] paddings_stride_x Stride of the paddinds tensor in X dimension (in bytes)

42

* @param[in] paddings_step_x paddings_stride_x * number of elements along X processed per workitem(in bytes)

43

* @param[in] paddings_stride_y Stride of the paddinds tensor in Y dimension (in bytes)

44

* @param[in] paddings_step_y paddings_stride_y * number of elements along Y processed per workitem(in bytes)

45

* @param[in] paddingse_offset_first_element_in_bytes The offset of the first element in the second source image

46

* @param[in] block_shape_ptr Pointer to the block shape tensor. Supported data types: S32

47

* @param[in] block_shape_stride_x Stride of the block shape tensor in X dimension (in bytes)

48

* @param[in] block_shape_step_x block_shape_stride_x * number of elements along X processed per workitem(in bytes)

49

* @param[in] block_shape_stride_y Stride of the block shape tensor in Y dimension (in bytes)

50

* @param[in] block_shape_step_y block_shape_stride_y * number of elements along Y processed per workitem(in bytes)

51

* @param[in] block_shape_offset_first_element_in_bytes The offset of the first element in the block shapetensor

52

* @param[in] batch_id The output tensor batch id

53

* @param[out] output_ptr Pointer to the destination tensor. Supported data types: same as @p input_ptr

54

* @param[in] output_stride_x Stride of the destination tensor in X dimension (in bytes)

55

* @param[in] output_step_x output_stride_x * number of elements along X processed per workitem(in bytes)

56

* @param[in] output_stride_y Stride of the destination tensor in Y dimension (in bytes)

57

* @param[in] output_step_y output_stride_y * number of elements along Y processed per workitem(in bytes)

58

* @param[in] output_stride_z Stride of the destination tensor in Z dimension (in bytes)

59

* @param[in] output_step_z output_stride_z * number of elements along Z processed per workitem(in bytes)

60

* @param[in] output_offset_first_element_in_bytes The offset of the first element in the destination image

61

*/

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

62

__kernel void space_to_batch_nchw(

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

63

TENSOR4D_DECLARATION(input),

64

IMAGE_DECLARATION(paddings),

65

VECTOR_DECLARATION(block_shape),

66

const int batch_id,

67

TENSOR3D_DECLARATION(output))

68

{

69

Tensor4D in = CONVERT_TO_TENSOR4D_STRUCT_NO_STEP(input, 0);

70

Image pad = CONVERT_TO_IMAGE_STRUCT_NO_STEP(paddings);

71

Vector block = CONVERT_TO_VECTOR_STRUCT_NO_STEP(block_shape);

72

Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(output);

73

Michalis Spyrou

2018-11-21 14:17:42 +0000

[diff] [blame]

74

const int pad_left_x = *((__global int *)offset(&pad, 0, 0));

75

const int pad_right_x = *((__global int *)offset(&pad, 1, 0));

76

const int pad_left_y = *((__global int *)offset(&pad, 0, 1));

77

const int pad_right_y = *((__global int *)offset(&pad, 1, 1));

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

78

79

int block_x = *((__global int *)vector_offset(&block, 0));

80

int block_y = *((__global int *)vector_offset(&block, 1));

81

82

const int out_x = get_global_id(0);

83

const int out_y = get_global_id(1);

84

const int z = get_global_id(2);

85

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

86

const int pos_x = out_x * block_x + ((batch_id / BATCH_IN) % block_x);

87

const int pos_y = out_y * block_y + ((batch_id / BATCH_IN) / block_x);

88

89

if(((pos_y >= pad_left_y) && (pos_y < pad_left_y + HEIGHT_IN) && (pos_x >= pad_left_x) && (pos_x < pad_left_x + WIDTH_IN)))

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

90

{

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

91

const int w = batch_id % BATCH_IN;

92

const int in_x = pos_x - pad_left_x;

93

const int in_y = pos_y - pad_left_y;

94

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

95

*((__global DATA_TYPE *)out.ptr) = *((__global DATA_TYPE *)tensor4D_offset(&in, in_x, in_y, z, w));

96

}

97

}

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

98

/** Calculate the space to batch conversion. (NHWC)

99

*

100

* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float

101

* @note The block shape tensor rank must be passed at compile time using -DBLOCK_SHAPE_DIM. e.g. -DBLOCK_SHAPE_DIM=2

102

*

103

* @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32

104

* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)

105

* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)

106

* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)

107

* @param[in] input_step_y input_stride_y * number of elements along Y processed per workitem(in bytes)

108

* @param[in] input_stride_z Stride of the source tensor in Z dimension (in bytes)

109

* @param[in] input_step_z input_stride_z * number of elements along Z processed per workitem(in bytes)

110

* @param[in] input_offset_first_element_in_bytes The offset of the first element in the first source image

111

* @param[in] paddings_ptr Pointer to the second source image. Supported data types: S32

112

* @param[in] paddings_stride_x Stride of the paddinds tensor in X dimension (in bytes)

113

* @param[in] paddings_step_x paddings_stride_x * number of elements along X processed per workitem(in bytes)

114

* @param[in] paddings_stride_y Stride of the paddinds tensor in Y dimension (in bytes)

115

* @param[in] paddings_step_y paddings_stride_y * number of elements along Y processed per workitem(in bytes)

116

* @param[in] paddingse_offset_first_element_in_bytes The offset of the first element in the second source image

117

* @param[in] block_shape_ptr Pointer to the block shape tensor. Supported data types: S32

118

* @param[in] block_shape_stride_x Stride of the block shape tensor in X dimension (in bytes)

119

* @param[in] block_shape_step_x block_shape_stride_x * number of elements along X processed per workitem(in bytes)

120

* @param[in] block_shape_stride_y Stride of the block shape tensor in Y dimension (in bytes)

121

* @param[in] block_shape_step_y block_shape_stride_y * number of elements along Y processed per workitem(in bytes)

122

* @param[in] block_shape_offset_first_element_in_bytes The offset of the first element in the block shapetensor

123

* @param[in] batch_id The output tensor batch id

124

* @param[out] output_ptr Pointer to the destination tensor. Supported data types: same as @p input_ptr

125

* @param[in] output_stride_x Stride of the destination tensor in X dimension (in bytes)

126

* @param[in] output_step_x output_stride_x * number of elements along X processed per workitem(in bytes)

127

* @param[in] output_stride_y Stride of the destination tensor in Y dimension (in bytes)

128

* @param[in] output_step_y output_stride_y * number of elements along Y processed per workitem(in bytes)

129

* @param[in] output_stride_z Stride of the destination tensor in Z dimension (in bytes)

130

* @param[in] output_step_z output_stride_z * number of elements along Z processed per workitem(in bytes)

131

* @param[in] output_offset_first_element_in_bytes The offset of the first element in the destination image

132

*/

133

__kernel void space_to_batch_nhwc(

134

TENSOR4D_DECLARATION(input),

135

IMAGE_DECLARATION(paddings),

136

VECTOR_DECLARATION(block_shape),

137

const int batch_id,

138

TENSOR3D_DECLARATION(output))

139

{

140

Tensor4D in = CONVERT_TO_TENSOR4D_STRUCT_NO_STEP(input, 0);

141

Image pad = CONVERT_TO_IMAGE_STRUCT_NO_STEP(paddings);

142

Vector block = CONVERT_TO_VECTOR_STRUCT_NO_STEP(block_shape);

143

Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(output);

144

Michalis Spyrou

2018-11-21 14:17:42 +0000

[diff] [blame]

145

const int pad_left_x = *((__global int *)offset(&pad, 0, 0));

146

const int pad_right_x = *((__global int *)offset(&pad, 1, 0));

147

const int pad_left_y = *((__global int *)offset(&pad, 0, 1));

148

const int pad_right_y = *((__global int *)offset(&pad, 1, 1));

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

149

150

int block_x = *((__global int *)vector_offset(&block, 0));

151

int block_y = *((__global int *)vector_offset(&block, 1));

152

153

const int out_x = get_global_id(1);

154

const int out_y = get_global_id(2);

155

const int z = get_global_id(0);

156

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

157

const int pos_x = out_x * block_x + ((batch_id / BATCH_IN) % block_x);

158

const int pos_y = out_y * block_y + ((batch_id / BATCH_IN) / block_x);

159

160

if(((pos_y >= pad_left_y) && (pos_y < pad_left_y + HEIGHT_IN) && (pos_x >= pad_left_x) && (pos_x < pad_left_x + WIDTH_IN)))

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

161

{

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

162

const int w = batch_id % BATCH_IN;

163

const int in_x = pos_x - pad_left_x;

164

const int in_y = pos_y - pad_left_y;

165

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

166

*((__global DATA_TYPE *)out.ptr) = *((__global DATA_TYPE *)tensor4D_offset(&in, z, in_x, in_y, w));

167

}

168

}

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

169

#endif // defined(BATCH_SIZE) && defined(DATA_TYPE) && defined(WIDTH_IN) && defined(HEIGHT_IN)

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

170

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

171

#if defined(BATCH_SIZE) && defined(DATA_TYPE) && defined(BLOCK_SHAPE_X) && defined(BLOCK_SHAPE_Y) && defined(PAD_LEFT_X) && defined(PAD_RIGHT_X) && defined(PAD_LEFT_Y) && defined(PAD_RIGHT_Y) && defined(WIDTH_IN) && defined(HEIGHT_IN)

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

172

/** Calculate the space to batch conversion.

173

*

174

* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float

175

* @note The input tensor batch size must be passed at compile time using -DBATCH_SIZE. e.g. -DBATCH_SIZE=2

176

* @note The block shape x must be passed at compile time using -DBLOCK_SHAPE_X. e.g. -DBLOCK_SHAPE_X=2

177

* @note The block shape y must be passed at compile time using -DBLOCK_SHAPE_Y. e.g. -DBLOCK_SHAPE_Y=2

178

* @note The starting pad value of x must be passed at compile time using -DPAD_LEFT_X. e.g. -DPAD_LEFT_X=2

179

* @note The ending pad value of x must be passed at compile time using -DPAD_RIGHT_X. e.g. -DPAD_RIGHT_X=2

180

* @note The starting pad value of y must be passed at compile time using -DPAD_LEFT_Y. e.g. -DPAD_LEFT_Y=2

181

* @note The ending pad value of y must be passed at compile time using -DPAD_RIGHT_Y. e.g. -DPAD_RIGHT_X=2

182

*

183

* @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32

184

* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)

185

* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)

186

* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)

187

* @param[in] input_step_y input_stride_y * number of elements along Y processed per workitem(in bytes)

188

* @param[in] input_stride_z Stride of the source tensor in Z dimension (in bytes)

189

* @param[in] input_step_z input_stride_z * number of elements along Z processed per workitem(in bytes)

190

* @param[in] input_offset_first_element_in_bytes The offset of the first element in the first source image

191

* @param[in] batch_id The output tensor batch id

192

* @param[out] output_ptr Pointer to the destination tensor. Supported data types: same as @p input_ptr

193

* @param[in] output_stride_x Stride of the destination tensor in X dimension (in bytes)

194

* @param[in] output_step_x output_stride_x * number of elements along X processed per workitem(in bytes)

195

* @param[in] output_stride_y Stride of the destination tensor in Y dimension (in bytes)

196

* @param[in] output_step_y output_stride_y * number of elements along Y processed per workitem(in bytes)

197

* @param[in] output_stride_z Stride of the source tensor in Z dimension (in bytes)

198

* @param[in] output_step_z output_stride_z * number of elements along Z processed per workitem(in bytes)

199

* @param[in] output_offset_first_element_in_bytes The offset of the first element in the destination image

200

*/

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

201

__kernel void space_to_batch_static_nchw(

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

202

TENSOR4D_DECLARATION(input),

203

const int batch_id,

204

TENSOR3D_DECLARATION(output))

205

{

206

Tensor4D in = CONVERT_TO_TENSOR4D_STRUCT_NO_STEP(input, 0);

207

Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(output);

208

209

int block_x = BLOCK_SHAPE_X;

Michalis Spyrou

2018-11-21 14:17:42 +0000

[diff] [blame]

210

int block_y = BLOCK_SHAPE_Y;

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

211

212

const int out_x = get_global_id(0);

213

const int out_y = get_global_id(1);

214

const int z = get_global_id(2);

215

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

216

const int pos_x = out_x * block_x + ((batch_id / BATCH_IN) % block_x);

217

const int pos_y = out_y * block_y + ((batch_id / BATCH_IN) / block_x);

218

219

if(pos_y >= PAD_LEFT_Y && pos_y < PAD_LEFT_Y + HEIGHT_IN && pos_x >= PAD_LEFT_X && pos_x < PAD_LEFT_X + WIDTH_IN)

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

220

{

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

221

const int w = batch_id % BATCH_IN;

222

const int in_x = pos_x - PAD_LEFT_X;

223

const int in_y = pos_y - PAD_LEFT_Y;

224

Michalis Spyrou

2018-08-21 18:03:58 +0100

[diff] [blame]

225

*((__global DATA_TYPE *)out.ptr) = *((__global DATA_TYPE *)tensor4D_offset(&in, in_x, in_y, z, w));

226

}

227

}

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

228

/** Calculate the space to batch conversion. (NHWC)

229

*

230

* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float

231

* @note The input tensor batch size must be passed at compile time using -DBATCH_SIZE. e.g. -DBATCH_SIZE=2

232

* @note The block shape x must be passed at compile time using -DBLOCK_SHAPE_X. e.g. -DBLOCK_SHAPE_X=2

233

* @note The block shape y must be passed at compile time using -DBLOCK_SHAPE_Y. e.g. -DBLOCK_SHAPE_Y=2

234

* @note The starting pad value of x must be passed at compile time using -DPAD_LEFT_X. e.g. -DPAD_LEFT_X=2

235

* @note The ending pad value of x must be passed at compile time using -DPAD_RIGHT_X. e.g. -DPAD_RIGHT_X=2

236

* @note The starting pad value of y must be passed at compile time using -DPAD_LEFT_Y. e.g. -DPAD_LEFT_Y=2

237

* @note The ending pad value of y must be passed at compile time using -DPAD_RIGHT_Y. e.g. -DPAD_RIGHT_X=2

238

*

239

* @param[in] input_ptr Pointer to the source tensor. Supported data types: U8/S8/QASYMM8/U16/S16/F16/U32/S32/F32

240

* @param[in] input_stride_x Stride of the source tensor in X dimension (in bytes)

241

* @param[in] input_step_x input_stride_x * number of elements along X processed per workitem(in bytes)

242

* @param[in] input_stride_y Stride of the source image in Y dimension (in bytes)

243

* @param[in] input_step_y input_stride_y * number of elements along Y processed per workitem(in bytes)

244

* @param[in] input_stride_z Stride of the source tensor in Z dimension (in bytes)

245

* @param[in] input_step_z input_stride_z * number of elements along Z processed per workitem(in bytes)

246

* @param[in] input_offset_first_element_in_bytes The offset of the first element in the first source image

247

* @param[in] batch_id The output tensor batch id

248

* @param[out] output_ptr Pointer to the destination tensor. Supported data types: same as @p input_ptr

249

* @param[in] output_stride_x Stride of the destination tensor in X dimension (in bytes)

250

* @param[in] output_step_x output_stride_x * number of elements along X processed per workitem(in bytes)

251

* @param[in] output_stride_y Stride of the destination tensor in Y dimension (in bytes)

252

* @param[in] output_step_y output_stride_y * number of elements along Y processed per workitem(in bytes)

253

* @param[in] output_stride_z Stride of the source tensor in Z dimension (in bytes)

254

* @param[in] output_step_z output_stride_z * number of elements along Z processed per workitem(in bytes)

255

* @param[in] output_offset_first_element_in_bytes The offset of the first element in the destination image

256

*/

257

__kernel void space_to_batch_static_nhwc(

258

TENSOR4D_DECLARATION(input),

259

const int batch_id,

260

TENSOR3D_DECLARATION(output))

261

{

262

Tensor4D in = CONVERT_TO_TENSOR4D_STRUCT_NO_STEP(input, 0);

263

Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(output);

264

265

int block_x = BLOCK_SHAPE_X;

Michalis Spyrou

2018-11-21 14:17:42 +0000

[diff] [blame]

266

int block_y = BLOCK_SHAPE_Y;

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

267

268

const int out_x = get_global_id(1);

269

const int out_y = get_global_id(2);

270

const int z = get_global_id(0);

271

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

272

const int pos_x = out_x * block_x + ((batch_id / BATCH_IN) % block_x);

273

const int pos_y = out_y * block_y + ((batch_id / BATCH_IN) / block_x);

274

275

if(pos_y >= PAD_LEFT_Y && pos_y < PAD_LEFT_Y + HEIGHT_IN && pos_x >= PAD_LEFT_X && pos_x < PAD_LEFT_X + WIDTH_IN)

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

276

{

Isabella Gottardi

2018-12-14 11:40:40 +0000

[diff] [blame]

277

const int w = batch_id % BATCH_IN;

278

const int in_x = pos_x - PAD_LEFT_X;

279

const int in_y = pos_y - PAD_LEFT_Y;

280

Michalis Spyrou

2018-09-18 13:09:30 +0100

[diff] [blame]

281

*((__global DATA_TYPE *)out.ptr) = *((__global DATA_TYPE *)tensor4D_offset(&in, z, in_x, in_y, w));

282

}

283

}

Isabella Gottardi