| /third_party/mindspore/mindspore/lite/test/ut/src/runtime/kernel/arm/fp32/ |
| D | reduce_fp32_tests.cc | 48 int thread_num_ = 1; member in mindspore::TestReduceFp32
93 ctx_->thread_num_ = thread_num_; in Prepare()
117 thread_num_ = 2; in TEST_F()
144 thread_num_ = 2; in TEST_F()
172 thread_num_ = 1; in TEST_F()
198 thread_num_ = 2; in TEST_F()
225 thread_num_ = 2; in TEST_F()
254 thread_num_ = 2; in TEST_F()
282 thread_num_ = 2; in TEST_F()
307 thread_num_ = 2; in TEST_F()
[all …]
|
| D | crop_fp32_test.cc | 39 crop_param.op_parameter_.thread_num_ = 1; in TEST_F()
62 crop_param.op_parameter_.thread_num_ = 1; in TEST_F()
82 crop_param.op_parameter_.thread_num_ = 1; in TEST_F()
102 crop_param.op_parameter_.thread_num_ = 2; in TEST_F()
187 crop_param.op_parameter_.thread_num_ = 2; in TEST_F()
213 crop_param.op_parameter_.thread_num_ = 2; in TEST_F()
237 crop_param.op_parameter_.thread_num_ = 2; in TEST_F()
270 ctx->thread_num_ = 2; in TEST_F()
|
| /third_party/mindspore/mindspore/lite/test/ut/src/runtime/kernel/arm/int8/ |
| D | reduce_int8_tests.cc | 43 int thread_num_ = 1; member in mindspore::TestReduceInt8
83 ctx_.thread_num_ = thread_num_; in Prepare()
106 thread_num_ = 2; in TEST_F()
129 thread_num_ = 2; in TEST_F()
153 thread_num_ = 2; in TEST_F()
176 thread_num_ = 2; in TEST_F()
199 thread_num_ = 2; in TEST_F()
221 thread_num_ = 2; in TEST_F()
244 thread_num_ = 2; in TEST_F()
266 thread_num_ = 2; in TEST_F()
[all …]
|
| D | scale_int8.cc | 37 int thread_num_ = 1; member in mindspore::TestScaleInt8
96 ctx_.thread_num_ = thread_num_; in Prepare()
119 thread_num_ = 2; in TEST_F()
145 thread_num_ = 2; in TEST_F()
171 thread_num_ = 2; in TEST_F()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/base/ |
| D | reshape_base.cc | 58 …thread_num_ = MSMIN(static_cast<size_t>(op_parameter_->thread_num_), UP_DIV(size, kMinCostPerThrea… in Run()
59 if (thread_num_ < 1) { in Run()
60 thread_num_ = 1; in Run()
62 auto block_size = UP_DIV(size, thread_num_); in Run()
63 thread_num_ = UP_DIV(size, block_size); in Run()
70 if (task_id == (thread_num_ - 1)) { in Run()
77 if (thread_num_ == 1) { in Run()
81 return lite::ParallelLaunch(this->ms_context_, Copy, nullptr, thread_num_); in Run()
|
| D | quant_dtype_cast.h | 28 : InnerKernel(parameter, inputs, outputs, ctx), thread_num_(ctx->thread_num_) {} in QuantDTypeCastCPUKernel()
37 int thread_num_;
|
| D | detection_post_process_base.h | 32 : InnerKernel(parameter, inputs, outputs, ctx), thread_num_(ctx->thread_num_) { in DetectionPostProcessBaseCPUKernel()
42 int thread_num_ = 1; variable
|
| /third_party/mindspore/mindspore/lite/micro/coder/opcoders/nnacl/int8/ |
| D | conv2d_3x3_int8_coder.cc | 90 tile_buffer_size_ = thread_num_ * TILE_NUM * 16 * ic8 * C8NUM * sizeof(int16_t); in InitTmpBuffer()
94 block_unit_buffer_size_ = thread_num_ * 4 * 4 * C8NUM * sizeof(int16_t); in InitTmpBuffer()
99 tmp_dst_buffer_size_ = thread_num_ * TILE_NUM * 16 * oc4 * C4NUM * sizeof(int32_t); in InitTmpBuffer()
116 conv_param_->thread_num_ = thread_num_; in Prepare()
118 conv_param_->op_parameter_.thread_num_ = thread_num_; in Prepare()
140 if (thread_num_ > 1) { in DoCode()
177 if (thread_num_ > 1) { in DoCode()
|
| D | deconvolution_int8_coder.cc | 37 conv_param_->thread_num_ = thread_num_; in Prepare()
38 conv_param_->op_parameter_.thread_num_ = thread_num_; in Prepare()
39 thread_count_ = thread_num_; in Prepare()
59 thread_count_ = MSMIN(conv_param_->op_parameter_.thread_num_, oc4); in InitParam()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp16/ |
| D | pad_fp16.cc | 32 PadFp16(input_, output_, in_, out_, pad_param_->paddings_, task_id, op_parameter_->thread_num_); in RunImpl()
49 …16(input_data, output_data, in_, out_, pad_param_->paddings_, task_id, op_parameter_->thread_num_); in RunMirrorPadImpl()
52 for (size_t i = task_id; i < mirror_pad_block_.size(); i += op_parameter_->thread_num_) { in RunMirrorPadImpl()
74 MS_CHECK_FALSE(op_parameter_->thread_num_ == 0, RET_ERROR); in RunMirrorPadImpl()
75 int unit = UP_DIV(out_tensors_.at(0)->ElementsNum(), op_parameter_->thread_num_); in RunMirrorPadImpl()
116 ret = ParallelLaunch(this->ms_context_, PadImpl, this, op_parameter_->thread_num_); in Run()
128 ret = ParallelLaunch(this->ms_context_, MirrorPadImpl, this, op_parameter_->thread_num_); in Run()
|
| D | deconvolution_winograd_fp16.cc | 128 deconv_param_->thread_num_ = MSMAX(1, op_parameter_->thread_num_); in InitParameter()
129 deconv_param_->thread_num_ = MSMIN(deconv_param_->thread_num_, deconv_param_->in_tile_count_); in InitParameter()
131 thread_num_hw_ = MSMIN(op_parameter_->thread_num_, deconv_param_->output_plane_); in InitParameter()
134 …int size = deconv_param_->thread_num_ * DECONV_WINOGRAD_DEFAULT_UNIT * DECONV_WINOGRAD_DEFAULT_UNI… in InitParameter()
144 size = deconv_param_->thread_num_ * deconv_param_->out_tile_w_ * deconv_param_->out_tile_h_ * in InitParameter()
162 malloc(deconv_param_->thread_num_ * size * sizeof(float16_t)); in InitParameter()
167 malloc(deconv_param_->thread_num_ * size * sizeof(float16_t)); in InitParameter()
173 …unit.winograd_.b_buffer_ = malloc(deconv_param_->thread_num_ * unit.winograd_.kh_ * unit.winograd_… in InitParameter()
178 … unit.tmp_buffer_ = malloc(deconv_param_->thread_num_ * unit.winograd_.kh_ * unit.winograd_.kw_ * in InitParameter()
184 …unit.tmp_buffer_ = malloc(deconv_param_->thread_num_ * deconv_param_->oc_div4_ * unit.w_size_ * un… in InitParameter()
[all …]
|
| D | addn_fp16.cc | 45 int count_per_thread = UP_DIV(elements_num_, op_parameter_->thread_num_); in AddNParallelRun()
64 if (static_cast<int>(elements_num_) < op_parameter_->thread_num_) { in Run()
95 auto ret = ParallelLaunch(this->ms_context_, AddNLaunch, this, op_parameter_->thread_num_); in Run()
104 ret = ParallelLaunch(this->ms_context_, AddNLaunch, this, op_parameter_->thread_num_); in Run()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp32/ |
| D | glu_fp32.cc | 92 usable_thread_num_ = MSMIN(op_parameter_->thread_num_, num_unit_); in ReSize()
121 MS_CHECK_TRUE_RET(op_parameter_->thread_num_ != 0, RET_ERROR); in Sigmoid()
122 int stride = UP_DIV(length, op_parameter_->thread_num_); in Sigmoid()
136 MS_CHECK_TRUE_RET(op_parameter_->thread_num_ != 0, RET_ERROR); in Mul()
137 int stride = UP_DIV(length, op_parameter_->thread_num_); in Mul()
176 ret = ParallelLaunch(this->ms_context_, SigmoidRun, this, op_parameter_->thread_num_); in Run()
183 ret = ParallelLaunch(this->ms_context_, MulRun, this, op_parameter_->thread_num_); in Run()
|
| D | l2_norm_fp32.cc | 93 int unit = UP_DIV(l2_norm_param_->data_num_, op_parameter_->thread_num_); in CalcSquareSum()
104 int unit = UP_DIV(l2_norm_param_->data_num_, op_parameter_->thread_num_); in DivSqrtSum()
121 int unit = UP_DIV(outer_size, op_parameter_->thread_num_); in CalcL2NormTrailingAxis()
170 ret = ParallelLaunch(this->ms_context_, SquareSumRun, this, op_parameter_->thread_num_); in Run()
176 for (int i = 0; i < op_parameter_->thread_num_; ++i) { in Run()
180 ret = ParallelLaunch(this->ms_context_, L2NormRun, this, op_parameter_->thread_num_); in Run()
186 … ret = ParallelLaunch(this->ms_context_, L2NormTrailingAxisRun, this, op_parameter_->thread_num_); in Run()
|
| D | deconvolution_winograd_fp32.cc | 115 deconv_param_->thread_num_ = MSMAX(1, op_parameter_->thread_num_); in InitParameter()
116 deconv_param_->thread_num_ = MSMIN(deconv_param_->thread_num_, deconv_param_->in_tile_count_); in InitParameter()
118 thread_num_hw_ = MSMIN(op_parameter_->thread_num_, deconv_param_->output_plane_); in InitParameter()
121 …int size = deconv_param_->thread_num_ * DECONV_WINOGRAD_DEFAULT_UNIT * DECONV_WINOGRAD_DEFAULT_UNI… in InitParameter()
145 malloc(deconv_param_->thread_num_ * size * sizeof(float)); in InitParameter()
151 malloc(deconv_param_->thread_num_ * size * sizeof(float)); in InitParameter()
158 …unit.winograd_.b_buffer_ = malloc(deconv_param_->thread_num_ * unit.winograd_.kh_ * unit.winograd_… in InitParameter()
164 … unit.tmp_buffer_ = malloc(deconv_param_->thread_num_ * unit.winograd_.kh_ * unit.winograd_.kw_ * in InitParameter()
171 …unit.tmp_buffer_ = malloc(deconv_param_->thread_num_ * deconv_param_->oc_div4_ * unit.w_size_ * un… in InitParameter()
391 … = task_id; tile_index < deconv_param_->in_tile_count_; tile_index += deconv_param_->thread_num_) { in DoDeconv()
[all …]
|
| D | convolution_depthwise_3x3_fp32.cc | 55 conv_param_->thread_num_ = MSMIN(thread_count_, conv_param_->output_h_); in ReSize()
63 if (conv_param_->thread_num_ == 0) { in DoExecute()
67 int step_oh = UP_DIV(conv_param_->output_h_, conv_param_->thread_num_); in DoExecute()
88 int buffer_size = units * c4 * C12NUM * conv_param_->thread_num_; in Run()
106 auto ret = ParallelLaunch(this->ms_context_, ConvDw3x3Run, this, conv_param_->thread_num_); in Run()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp16_grad/ |
| D | bn_fp16_grad.cc | 89 int thread_num = thread_num_; in DoExecute()
166 thread_num_ = ms_context_->thread_num_; in Run()
167 if (thread_num_ == 1) { in Run()
168 int error_code = ParallelLaunch(this->ms_context_, BNGradFp16Run, this, thread_num_); in Run()
174 const std::vector<int> threads = {thread_num_, 1, thread_num_}; in Run()
|
| D | pooling_fp16_grad.cc | 81 MS_CHECK_TRUE_RET(thread_num_ > 0, RET_ERROR); in DoExecute()
82 int stride = UP_DIV(pool_param->output_batch_, thread_num_); in DoExecute()
114 thread_num_ = ms_context_->thread_num_; in Run()
115 int error_code = ParallelLaunch(this->ms_context_, PoolingFp16GradImpl, this, thread_num_); in Run()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp32_grad/ |
| D | bn_grad.cc | 89 int thread_num = thread_num_; in Execute()
179 thread_num_ = op_parameter_->thread_num_; in Run()
181 if (thread_num_ == 1) { in Run()
182 error_code = ParallelLaunch(this->ms_context_, BNGradRun, this, thread_num_); in Run()
188 const std::vector<int> threads = {thread_num_, 1, thread_num_}; in Run()
|
| D | pooling_grad.cc | 82 MS_CHECK_TRUE_RET(thread_num_ > 0, RET_ERROR); in Execute()
83 int stride = UP_DIV(pool_param->output_batch_, thread_num_); in Execute()
115 thread_num_ = op_parameter_->thread_num_; in Run()
116 int error_code = ParallelLaunch(this->ms_context_, PoolingGradImpl, this, thread_num_); in Run()
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp16/ |
| D | conv_fp16.c | 31 NNACL_CHECK_ZERO_RETURN(conv_param->thread_num_); in ConvFp16()
34 int block_per_thread = UP_DIV(UP_DIV(output_hw, tile_n), conv_param->thread_num_); in ConvFp16()
73 NNACL_CHECK_ZERO_RETURN(conv_param->op_parameter_.thread_num_); in ConvOutNc8hw8Fp16()
77 int block_per_thread = UP_DIV(input_block, conv_param->thread_num_); in ConvOutNc8hw8Fp16()
121 NNACL_CHECK_ZERO_RETURN(param->op_parameter_.thread_num_); in Conv1x1OutNc8hw8MultiThreadByInputFp16()
125 int block_per_thread = UP_DIV(input_block, param->op_parameter_.thread_num_); in Conv1x1OutNc8hw8MultiThreadByInputFp16()
163 NNACL_CHECK_ZERO_RETURN(param->op_parameter_.thread_num_); in Conv1x1OutNc8hw8MultiThreadByWeightFp16()
167 int block_per_thread = UP_DIV(weight_block, param->op_parameter_.thread_num_); in Conv1x1OutNc8hw8MultiThreadByWeightFp16()
197 NNACL_CHECK_ZERO_RETURN(conv_param->thread_num_); in ConvWinogardFp16()
202 int per_thread_num = UP_DIV(output_count, conv_param->thread_num_); in ConvWinogardFp16()
[all …]
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/string/ |
| D | lsh_projection.cc | 67 …aram_->thread_stride_ = op_parameter_->thread_num_ > 1 ? UP_DIV(param_->hash_shape_[0], op_paramet… in Run()
73 ret = ParallelLaunch(this->ms_context_, LshProjectionRun, this, op_parameter_->thread_num_); in Run()
83 … static_cast<char **>(ms_context_->allocator->Malloc(op_parameter_->thread_num_ * sizeof(char *))); in MallocKeys()
88 for (int i = 0; i < op_parameter_->thread_num_; i++) { in MallocKeys()
101 for (int i = 0; i < op_parameter_->thread_num_; i++) { in FreeKeys()
|
| /third_party/mindspore/mindspore/lite/micro/example/mnist_stm32f746/mnist_stm32f746/operator_library/kernels/nnacl/fp32/ |
| D | exp_fp32.c | 24 for (size_t i = task_id; i < parameter->element_num_; i += parameter->thread_num_) { in Exp()
28 for (size_t i = task_id; i < parameter->element_num_; i += parameter->thread_num_) { in Exp()
33 for (size_t i = task_id; i < parameter->element_num_; i += parameter->thread_num_) { in Exp()
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/nnacl/fp32/ |
| D | batchnorm_fp32.c | 24 if (param->op_parameter_.thread_num_ == 0) { in BatchNormFp32()
27 int units_per_thread = UP_DIV(param->unit_, param->op_parameter_.thread_num_); in BatchNormFp32()
44 if (param->op_parameter_.thread_num_ == 0) { in FusedBatchNormFp32()
47 int units_per_thread = UP_DIV(param->unit_, param->op_parameter_.thread_num_); in FusedBatchNormFp32()
|
| /third_party/mindspore/mindspore/ccsrc/backend/kernel_compiler/cpu/ |
| D | sparse_optimizer_cpu_kernel.h | 86 size_t thread_num_{0};
173 if (param.thread_num_ < 1) { in SplitAndCalculateSegmentBucketSize()
176 size_t thread_indices_size = input_grad->indices_size_ / param.thread_num_; in SplitAndCalculateSegmentBucketSize()
177 size_t left_indices_size = input_grad->indices_size_ % param.thread_num_; in SplitAndCalculateSegmentBucketSize()
179 tasks.reserve(param.thread_num_); in SplitAndCalculateSegmentBucketSize()
180 segments.reserve(param.thread_num_); in SplitAndCalculateSegmentBucketSize()
183 for (size_t i = 0; i < param.thread_num_; ++i) { in SplitAndCalculateSegmentBucketSize()
184 …oid)segment_bucket_sizes.emplace_back(std::make_shared<std::vector<size_t>>(param.thread_num_, 0)); in SplitAndCalculateSegmentBucketSize()
210 if (param.thread_num_ == 0) { in CopySegmentIndicesToBucket()
213 std::vector<size_t> bucket_data_num(param.thread_num_, 0); in CopySegmentIndicesToBucket()
[all …]
|