| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp16/ |
| D | convolution_depthwise_slidewindow_fp16.cc | 38 …packed_input_ = reinterpret_cast<float16_t *>(ms_context_->allocator->Malloc(pack_input_size * siz… in InitPackedInputOutput()
39 if (packed_input_ == nullptr) { in InitPackedInputOutput()
49 ms_context_->allocator->Free(packed_input_); in InitPackedInputOutput()
50 packed_input_ = nullptr; in InitPackedInputOutput()
133 ConvDwC8Fp16(packed_output_, packed_input_, reinterpret_cast<float16_t *>(packed_weight_), in DoExecute()
167 PackNHWCToNHWC8Fp16(input_ptr, packed_input_, conv_param_->input_batch_, in Run()
170 packed_input_ = input_ptr; in Run()
192 ms_context_->allocator->Free(packed_input_); in FreePackedInputOutput()
194 packed_input_ = nullptr; in FreePackedInputOutput()
|
| D | convolution_fp16.h | 45 if (packed_input_ != nullptr) { in FreeTmpBuffer()
46 ctx_->allocator->Free(packed_input_); in FreeTmpBuffer()
47 packed_input_ = nullptr; in FreeTmpBuffer()
54 float16_t *packed_input_ = nullptr; variable
|
| D | deconvolution_depthwise_fp16.cc | 53 …packed_input_ = reinterpret_cast<float16_t *>(ms_context_->allocator->Malloc(pack_input_size * siz… in InitPackedInputOutput()
54 if (packed_input_ == nullptr) { in InitPackedInputOutput()
152 DeconvDwC8Fp16(packed_output_, packed_input_, reinterpret_cast<float16_t *>(packed_weight_), in DoExecute()
192 PackNHWCToNHWC8Fp16(input_ptr, packed_input_, conv_param_->input_batch_, in Run()
195 packed_input_ = input_ptr; in Run()
218 ms_context_->allocator->Free(packed_input_); in FreePackedInputOutput()
220 packed_input_ = nullptr; in FreePackedInputOutput()
|
| D | convolution_fp16.cc | 81 …packed_input_ = reinterpret_cast<float16_t *>(ctx_->allocator->Malloc(unit_size * sizeof(float16_t… in InitTmpBuffer()
82 if (packed_input_ == nullptr) { in InitTmpBuffer()
154 ConvOutNc8hw8Fp16(input_ptr, packed_input_, reinterpret_cast<float16_t *>(packed_weight_), in RunImpl()
157 ConvFp16(input_ptr, packed_input_, reinterpret_cast<float16_t *>(packed_weight_), in RunImpl()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/fp32/ |
| D | convolution_fp32.h | 44 if (packed_input_ != nullptr) { in FreeTmpBuffer()
45 ctx_->allocator->Free(packed_input_); in FreeTmpBuffer()
46 packed_input_ = nullptr; in FreeTmpBuffer()
61 float *packed_input_ = nullptr; variable
|
| D | convolution_depthwise_slidewindow_x86_fp32.cc | 40 …packed_input_ = reinterpret_cast<float *>(ms_context_->allocator->Malloc(pack_input_size * sizeof(… in InitPackedInputOutput()
41 if (packed_input_ == nullptr) { in InitPackedInputOutput()
96 DepthwiseSWAvxFp32(packed_output_, packed_input_, reinterpret_cast<float *>(packed_weight_), in Execute()
128 PackNHWCToNHWCXFp32(input_ptr, packed_input_, conv_param_->input_batch_, in Run()
131 packed_input_ = input_ptr; in Run()
158 ms_context_->allocator->Free(packed_input_); in FreePackedInputOutput()
159 packed_input_ = nullptr; in FreePackedInputOutput()
|
| D | convolution_depthwise_slidewindow_fp32.cc | 37 …packed_input_ = reinterpret_cast<float *>(ms_context_->allocator->Malloc(pack_input_size * sizeof(… in InitPackedInputOutput()
38 if (packed_input_ == nullptr) { in InitPackedInputOutput()
95 ConvDwSWFp32(packed_output_, packed_input_, reinterpret_cast<float *>(packed_weight_), in DoExecute()
127 PackNHWCToNHWC4Fp32(input_ptr, packed_input_, conv_param_->input_batch_, in Run()
130 packed_input_ = input_ptr; in Run()
156 ms_context_->allocator->Free(packed_input_); in FreePackedInputOutput()
158 packed_input_ = nullptr; in FreePackedInputOutput()
|
| D | convolution_depthwise_indirect_fp32.cc | 123 …packed_input_ = reinterpret_cast<float *>(ms_context_->allocator->Malloc(pack_input_size * sizeof(… in MallocPackedInput()
124 if (packed_input_ == nullptr) { in MallocPackedInput()
147 PackNHWCToNHWC8Fp32(input_ptr, packed_input_, conv_param_->input_batch_, in Run()
150 PackNHWCToNHWC4Fp32(input_ptr, packed_input_, conv_param_->input_batch_, in Run()
154 packed_input_ = input_ptr; in Run()
163 ConvDwInitIndirection(indirect_buffer_, packed_input_, zero_ptr_, conv_param_, step_h, step_w); in Run()
171 ms_context_->allocator->Free(packed_input_); in Run()
|
| D | deconvolution_depthwise_fp32.cc | 52 …packed_input_ = reinterpret_cast<float *>(ms_context_->allocator->Malloc(pack_input_size * sizeof(… in InitPackedInputOutput()
53 if (packed_input_ == nullptr) { in InitPackedInputOutput()
119 DeconvDwSWFp32(packed_output_, packed_input_, reinterpret_cast<float *>(packed_weight_), in DoExecute()
165 PackNHWCToNHWC4Fp32(input_addr, packed_input_, conv_param_->input_batch_, in Run()
168 packed_input_ = input_addr; in Run()
221 ms_context_->allocator->Free(packed_input_); in FreePackedInputOutput()
223 packed_input_ = nullptr; in FreePackedInputOutput()
|
| D | convolution_fp32.cc | 49 packed_input_ = reinterpret_cast<float *>(ctx_->allocator->Malloc(unit_size * sizeof(float))); in InitTmpBuffer()
50 if (packed_input_ == nullptr) { in InitTmpBuffer()
115 ConvFp32(ori_input_data, packed_input_, reinterpret_cast<float *>(packed_weight_), in RunImpl()
119 ConvFp32OutNC4HW4(ori_input_data, packed_input_, reinterpret_cast<float *>(packed_weight_), in RunImpl()
|
| /third_party/mindspore/mindspore/lite/src/runtime/kernel/arm/int8/ |
| D | convolution_int8.h | 58 if (packed_input_ != nullptr) { in FreeTmpBuffer()
59 ctx_->allocator->Free(packed_input_); in FreeTmpBuffer()
60 packed_input_ = nullptr; in FreeTmpBuffer()
77 int8_t *packed_input_ = nullptr; variable
|
| D | deconvolution_depthwise_int8.cc | 101 …packed_input_ = reinterpret_cast<int16_t *>(ms_context_->allocator->Malloc(pack_input_size * sizeo… in InitBuffer()
102 if (packed_input_ == nullptr) { in InitBuffer()
170 …DeconvDwInt8(packed_output_, buffer, packed_input_, packed_weight_, reinterpret_cast<int32_t *>(bi… in DoExecute()
193 ms_context_->allocator->Free(packed_input_); in Run()
194 packed_input_ = nullptr; in Run()
214 PackDepthwiseInt8Input(input_addr, packed_input_, conv_param_); in Run()
232 ms_context_->allocator->Free(packed_input_); in Run()
233 packed_input_ = nullptr; in Run()
|
| D | convolution_1x1_int8.cc | 119 packed_input_ = reinterpret_cast<int8_t *>(ctx_->allocator->Malloc(size * sizeof(int8_t))); in InitRunBuf()
120 if (packed_input_ == nullptr) { in InitRunBuf()
128 if (packed_input_ != nullptr) { in FreeRunBuf()
129 ctx_->allocator->Free(packed_input_); in FreeRunBuf()
130 packed_input_ = nullptr; in FreeRunBuf()
431 …int8_t *hw_packed_in = packed_input_ + task_id * thread_stride_hw_ * C4NUM * matmul_param_->deep_1… in RunArmHw()
458 …int8_t *hw_packed_in = packed_input_ + task_id * thread_stride_hw_ * C4NUM * matmul_param_->deep_4… in RunArm64OptHw()
491 …Conv1x1Int8Opt(packed_input_, packed_weight_ + cur_stride * matmul_param_->deep_4_, output_ptr_ + … in RunArm64OptOc()
520 …Conv1x1Int8(packed_input_, packed_weight_ + cur_stride * matmul_param_->deep_16_, output_ptr_ + cu… in RunArmOc()
535 …int8_t *hw_packed_in = packed_input_ + task_id * thread_stride_hw_ * C4NUM * matmul_param_->deep_4… in OcOptPre()
[all …]
|
| D | convolution_depthwise_slidewindow_int8.cc | 78 packed_input_ = in InitPackedInputOutput()
80 if (packed_input_ == nullptr) { in InitPackedInputOutput()
293 …ConvDwInt8SW(packed_output_, packed_input_, packed_weight_, reinterpret_cast<int32_t *>(bias_data_… in DoExecute()
321 PackNHWCToNHWC8Int8(input_addr, packed_input_, conv_param_->output_batch_, in Run()
324 packed_input_ = input_addr; in Run()
348 ms_context_->allocator->Free(packed_input_); in FreePackedInputOutput()
350 packed_input_ = nullptr; in FreePackedInputOutput()
|
| D | convolution_int8.cc | 160 …packed_input_ = reinterpret_cast<int8_t *>(ctx_->allocator->Malloc(tmp_size * thread_count_ * tile… in InitTmpBuffer()
161 if (packed_input_ == nullptr) { in InitTmpBuffer()
208 …ConvInt8(ori_input_data, packed_input_, matmul_packed_input_, packed_weight_, reinterpret_cast<int… in RunImpl()
|
| D | deconvolution_depthwise_int8.h | 45 int16_t *packed_input_ = nullptr; variable
|
| /third_party/mindspore/mindspore/lite/micro/coder/opcoders/nnacl/fp32/ |
| D | deconv2d_fp32_coder.cc | 48 …packed_input_ = reinterpret_cast<float *>(allocator_->Malloc(kNumberTypeFloat32, pack_input_size_,… in InitRunBuf()
49 MS_CHECK_PTR(packed_input_); in InitRunBuf()
166 code.CodeFunction("memset", packed_input_, "0", pack_input_size_); in DoCode()
180 …code.CodeFunction("RowMajor2Col4Major", input_ptr_, packed_input_, matmul_param_.row_, matmul_para… in DoCode()
182 …code.CodeFunction("RowMajor2Col12Major", input_ptr_, packed_input_, matmul_param_.row_, matmul_par… in DoCode()
184 …code.CodeBaseStruct("DeConvFp32Args", kRunArgs, packed_input_, packed_weight_, packed_bias_, packe… in DoCode()
|
| D | convolution_fp32_coder.cc | 35 …packed_input_ = reinterpret_cast<float *>(allocator_->Malloc(kNumberTypeFloat32, packed_input_size… in InitTmpBuffer()
36 MS_CHECK_PTR(packed_input_); in InitTmpBuffer()
163 code.CodeFunction("memset", packed_input_, "0", packed_input_size_); in DoCode()
166 …code.CodeFunction("ConvFp32", input_tensor_, packed_input_, packed_weight_, bias_data_, col_major_… in DoCode()
|
| /third_party/mindspore/mindspore/lite/micro/coder/wrapper/int8/ |
| D | conv1x1_run_int8_wrapper.c | 42 …int8_t *hw_packed_in = args->packed_input_ + task_id * args->thread_stride_hw_ * C4NUM * args->mat… in OcOptPre()
73 …Conv1x1Int8Opt(args->packed_input_, args->packed_weight_ + cur_stride * args->matmul_param_->deep_… in RunArm64OptOc()
104 …Conv1x1Int8(args->packed_input_, args->packed_weight_ + cur_stride * args->matmul_param_->deep_16_, in RunArmOc()
121 …int8_t *hw_packed_in = args->packed_input_ + task_id * args->thread_stride_hw_ * C4NUM * args->mat… in RunArm64OptHw()
149 args->packed_input_ + task_id * args->thread_stride_hw_ * C4NUM * args->matmul_param_->deep_16_; in RunArmHw()
|
| D | convolution_int8_wrapper.h | 27 int8_t *packed_input_; member
|
| D | convolution_int8_wrapper.c | 21 …ConvInt8(args->input_data_, args->packed_input_, args->matmul_input_, args->packed_weight_, args->… in ConvolutionInt8Run()
|
| /third_party/mindspore/mindspore/lite/micro/coder/opcoders/nnacl/int8/ |
| D | conv2d_int8_coder.cc | 51 …packed_input_ = static_cast<int8_t *>(allocator_->Malloc(kNumberTypeInt8, packed_input_size_, kWor… in InitTmpBuffer()
52 MS_CHECK_PTR(packed_input_); in InitTmpBuffer()
224 code.CodeFunction("memset", packed_input_, 0, packed_input_size_); in DoCode()
230 …code.CodeBaseStruct("ConvolutionInt8Args", kRunArgs, input_tensor_, packed_input_, matmul_packed_i… in DoCode()
234 …code.CodeBaseStruct("ConvolutionInt8Args", kRunArgs, input_tensor_, packed_input_, matmul_packed_i… in DoCode()
|
| D | conv2d_int8_coder.h | 42 packed_input_ = nullptr; in ~Conv2DINT8Coder()
67 int8_t *packed_input_{nullptr};
|
| D | conv2d_1x1_int8_coder.cc | 76 … multiplier_, packed_weight_, bias_data_, packed_input_, nullptr, nullptr, 0, 0, 0, 0, in DoCode()
266 …packed_input_ = reinterpret_cast<int8_t *>(allocator_->Malloc(kNumberTypeInt8, size * sizeof(int8_… in InitRunBuf()
267 MS_CHECK_PTR(packed_input_); in InitRunBuf()
|
| /third_party/mindspore/mindspore/lite/micro/coder/wrapper/fp32/ |
| D | deconvolution_fp32_wrapper.h | 25 const float *packed_input_; member
|