| /external/tensorflow/tensorflow/lite/delegates/gpu/gl/kernels/ |
| D | conv.cc | 234 auto dst_depth = DivideRoundUp(ctx.output_shapes[0][3], 4); in GenerateCode() local
237 if (dst_depth >= 2) { in GenerateCode()
240 if (dst_depth >= 4) { in GenerateCode()
243 if (dst_depth >= 8) { in GenerateCode()
246 if (dst_depth >= 32) { in GenerateCode()
249 if (dst_depth >= 64) { in GenerateCode()
253 if (dst_depth >= 2) { in GenerateCode()
256 if (dst_depth >= 4) { in GenerateCode()
259 if (dst_depth >= 8) { in GenerateCode()
262 if (dst_depth >= 32) { in GenerateCode()
[all …]
|
| D | fully_connected.cc | 43 const int dst_depth = DivideRoundUp(attr.weights.shape.o, 4); in GenerateCode() local
53 {"dst_depth", dst_depth}, in GenerateCode()
110 /*workload=*/uint3(dst_depth, kWorkgroupHintY, 1), in GenerateCode()
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/common/tasks/ |
| D | conv_buffer_1x1.cc | 83 int dst_depth) { in GetBestParams() argument
103 int task_size = shape.w * shape.b * shape.h * dst_depth; in GetBestParams()
117 if (definition.precision == CalculationsPrecision::F32 && dst_depth < 32) { in GetBestParams()
123 if (dst_depth >= 32) { in GetBestParams()
135 int src_depth, int dst_depth) { in GetBestParams() argument
348 const int dst_depth = DivideRoundUp(attr.weights.shape.o, 4); in CreateConvBuffer1x1() local
353 GetBestParams(gpu_info, definition, *shape, src_depth, dst_depth); in CreateConvBuffer1x1()
355 conv_params = GetBestParams(gpu_info, definition, src_depth, dst_depth); in CreateConvBuffer1x1()
366 const int dst_depth = DivideRoundUp(attr.weights.shape.o, 4); in CreateConvBuffer1x1() local
371 GetBestParams(gpu_info, definition, *shape, src_depth, dst_depth); in CreateConvBuffer1x1()
[all …]
|
| D | conv_powervr.cc | 1060 int dst_depth, bool x_kernel_is_1, bool y_kernel_is_1, in GuessBestParams() argument
1085 if (dst_depth % 4 == 0 || dst_depth >= 8) { in GuessBestParams()
1087 } else if (dst_depth % 2 == 0 || dst_depth >= 4) { in GuessBestParams()
1090 conv_params.block_size.w = dst_depth; in GuessBestParams()
1093 int task_size = dst_shape->w * dst_shape->b * dst_shape->h * dst_depth; in GuessBestParams()
1134 if (dst_depth % 8 == 0 || dst_depth >= 32) { in GuessBestParams()
1136 } else if (dst_depth % 4 == 0 || dst_depth >= 8) { in GuessBestParams()
1138 } else if (dst_depth % 2 == 0 || dst_depth >= 4) { in GuessBestParams()
1141 conv_params.block_size.w = dst_depth; in GuessBestParams()
1181 if (dst_depth % 8 == 0 || dst_depth >= 32) { in GuessBestParams()
[all …]
|
| D | conv_constants.h | 36 const int dst_depth = DivideRoundUp(weights.shape.o, 4); in RearrangeWeightsForConvConstants() local
45 for (int d = 0; d < dst_depth; ++d) { in RearrangeWeightsForConvConstants()
73 const int dst_depth = DivideRoundUp(weights.shape.o, 4); in RearrangeWeightsForConvConstantsDot() local
82 for (int d = 0; d < dst_depth; ++d) { in RearrangeWeightsForConvConstantsDot()
112 const int dst_depth = DivideRoundUp(weights.shape.o, 4); in UploadWeightsForConvConstants() local
119 : weights.shape.i * dst_depth * 4; in UploadWeightsForConvConstants()
|
| D | add.cc | 29 int dst_depth = DivideRoundUp(dst_channels, 4); in CreateAdd() local
32 add.linkable_ = dst_depth == src0_depth; in CreateAdd()
33 if (src0_depth < dst_depth) { in CreateAdd()
|
| D | fully_connected.h | 96 const int dst_depth = DivideRoundUp(dst_channels, 4); in RearrangeFCWeightsToOIO4I4() local
99 for (int d = 0; d < dst_depth; ++d) { in RearrangeFCWeightsToOIO4I4()
151 const int dst_depth = DivideRoundUp(weights.shape.o, 4); in UploadWeights() local
153 const int elements_count = src_depth * dst_depth * 4; in UploadWeights()
178 desc.size = int2(src_depth * 4, dst_depth); in UploadWeights()
|
| D | convolution_transposed.h | 119 const int dst_depth = in UploadWeights() local
124 int texture_width = dst_depth; in UploadWeights()
143 const int dst_depth = in UploadWeights() local
151 kernel_x * kernel_y * kernel_z * src_depth * dst_depth * 4; in UploadWeights()
186 int texture_width = dst_depth; in UploadWeights()
|
| D | convolution_transposed_3x3_thin.cc | 42 const OperationDef& op_def, int src_depth, int dst_depth) { in GenerateConvolutionTransposedCode() argument
99 for (int d = 0; d < dst_depth; ++d) { in GenerateConvolutionTransposedCode()
147 for (int d = 0; d < dst_depth; ++d) { in GenerateConvolutionTransposedCode()
169 for (int d = 0; d < dst_depth; ++d) { in GenerateConvolutionTransposedCode()
|
| D | conv_constants.cc | 234 const int dst_depth = DivideRoundUp(dst_channels, 4); in IsDotConvBetter() local
235 return dst_channels * src_depth < src_channels * dst_depth; in IsDotConvBetter()
253 const int dst_depth = DivideRoundUp(w_shape.o, 4); in IsConvConstantsSupported() local
255 use_dot_conv ? w_shape.o * src_depth * 4 : w_shape.i * dst_depth * 4; in IsConvConstantsSupported()
|
| D | convolution_transposed.cc | 59 const int dst_depth = DivideRoundUp(attr.weights.shape.o, 4); in ConvolutionTransposed() local
60 if (dst_depth == 1 || dst_depth == 3) { in ConvolutionTransposed()
105 const int dst_depth = DivideRoundUp(attr.weights.shape.o, 4); in ConvolutionTransposed() local
106 if (dst_depth == 1 || dst_depth == 3) { in ConvolutionTransposed()
|
| D | convolution_transposed_3x3_thin.h | 77 int src_depth, int dst_depth);
|
| D | depthwise_conv.h | 40 const int dst_depth = DivideRoundUp(dst_channels, 4); in RearrangeWeightsForDWConv2D() local
45 for (int d = 0; d < dst_depth; ++d) { in RearrangeWeightsForDWConv2D()
|
| D | conv_buffer_1x1.h | 127 const int dst_depth = DivideRoundUp(weights.shape.o, 4); in UploadWeights() local
133 const int dst_depth_aligned = AlignByN(dst_depth, conv_params_.block_size.z); in UploadWeights()
|
| D | conv_powervr.h | 189 int dst_depth, bool x_kernel_is_1,
277 const int dst_depth = in UploadWeights() local
282 int texture_width = dst_depth; in UploadWeights()
|
| D | conv_weights_converter_test_util.cc | 63 const int dst_depth = AlignByN(DivideRoundUp(weights.shape.o, 4), in ConvolutionWeightsConverterTest() local
68 int texture_width = dst_depth; in ConvolutionWeightsConverterTest()
|
| D | conv_metal.cc | 44 const int dst_depth = DivideRoundUp(dst_channels, 4); in GetNumOutputSlices() local
45 if (dst_depth % 4 == 0 || dst_depth >= 16) { in GetNumOutputSlices()
47 } else if (dst_depth % 2 == 0 || dst_depth >= 4) { in GetNumOutputSlices()
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/common/tasks/special/ |
| D | fc_fc_add.h | 72 const int dst_depth = DivideRoundUp(dst_channels, 4); in RearrangeFCWeightsToOIO4I4() local
75 for (int d = 0; d < dst_depth; ++d) { in RearrangeFCWeightsToOIO4I4()
128 const int dst_depth = DivideRoundUp(weights.shape.o, 4); in UploadWeights() local
130 const int elements_count = src_depth * dst_depth * 4; in UploadWeights()
158 desc.size = int2(src_depth * 4, dst_depth); in UploadWeights()
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/cl/ |
| D | util.h | 41 const int dst_depth = dst.size(); in CopyLinearFLT4() local
42 for (int d = 0; d < dst_depth; ++d) { in CopyLinearFLT4()
|
| /external/deqp/external/openglcts/modules/gl/ |
| D | gl4cStencilTexturingTests.cpp | 1163 GLfloat dst_depth = 0.0f; in verifyTexture() local
1167 memcpy(&dst_depth, dst_pixel_data, 4); in verifyTexture()
1172 if (de::abs(src_depth_f - dst_depth) > 0.0001f) in verifyTexture()
1175 << "], got: " << dst_depth << " expected: " << src_depth_f in verifyTexture()
1183 GLfloat dst_depth = 0.0f; in verifyTexture() local
1186 memcpy(&dst_depth, dst_pixel_data, 4); in verifyTexture()
1189 if (de::abs(src_depth - dst_depth) > 0.0001f) in verifyTexture()
1192 << "], got: " << dst_depth << " expected: " << src_depth in verifyTexture()
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/common/selectors/ |
| D | operation_selector.cc | 50 const int dst_depth = DivideRoundUp(attr.weights.shape.o, 4); in IsRecommendedForWinograd4x4To6x6() local
74 src_depth >= min_depth && dst_depth >= min_depth; in IsRecommendedForWinograd4x4To6x6()
353 const int dst_depth = AlignByN(DivideRoundUp(weights_shape.o, 4), in GPUOperationFromNode() local
364 int texture_width = dst_depth; in GPUOperationFromNode()
|
| /external/tensorflow/tensorflow/lite/delegates/gpu/common/ |
| D | convert.cc | 135 const int dst_depth = DivideRoundUp(shape.o, 4); in ConvertToPHWO4I4() local
139 for (int f = 0; f < dst_depth; ++f) { in ConvertToPHWO4I4()
|
| /external/mesa3d/src/intel/compiler/ |
| D | brw_fs_visitor.cpp | 442 const fs_reg dst_depth = fetch_payload_reg(bld, payload.dest_depth_reg); in emit_single_fb_write() local
456 color0, color1, src0_alpha, src_depth, dst_depth, src_stencil, in emit_single_fb_write()
|
| D | brw_fs.cpp | 4369 const fs_reg &dst_depth = inst->src[FB_WRITE_LOGICAL_SRC_DST_DEPTH]; in lower_fb_write_logical_send() local
4534 if (dst_depth.file != BAD_FILE) { in lower_fb_write_logical_send()
4535 sources[length] = dst_depth; in lower_fb_write_logical_send()
|
| /external/mesa3d/src/mesa/state_tracker/ |
| D | st_cb_texture.c | 2326 unsigned dst_depth = 1; in fallback_copy_texsubimage() local
2357 dst_width, dst_height, dst_depth, in fallback_copy_texsubimage()
|