!4832 remove all todos

Merge pull request !4832 from wangzhe/master

mindspore/lite/src/populate_parameter.cc

@@ -263,7 +263,6 @@ OpParameter *PopulatePoolingParameter(const mindspore::lite::PrimitiveC *primiti
   pooling_param->global_ = pooling_primitive->GetGlobal();
   pooling_param->window_w_ = pooling_primitive->GetWindowW();
   pooling_param->window_h_ = pooling_primitive->GetWindowH();
-  // todo format
   auto pooling_lite_primitive = (lite::Pooling *)primitive;
   MS_ASSERT(nullptr != pooling_lite_primitive);
   pooling_param->pad_u_ = pooling_lite_primitive->PadUp();
@@ -402,7 +401,6 @@ OpParameter *PopulateConvDwParameter(const mindspore::lite::PrimitiveC *primitiv
   auto conv_primitive = dynamic_cast<const mindspore::lite::DepthwiseConv2D *>(primitive);
   conv_param->kernel_h_ = conv_primitive->GetKernelH();
   conv_param->kernel_w_ = conv_primitive->GetKernelW();
-  // todo format, group
   conv_param->stride_h_ = conv_primitive->GetStrideH();
   conv_param->stride_w_ = conv_primitive->GetStrideW();
 

mindspore/lite/src/runtime/kernel/arm/fp32/pad.cc

@@ -101,7 +101,6 @@ int PadCPUKernel::Run() {
   int output_size = output->DataSize();
 
   auto output_data = reinterpret_cast<float *>(output->Data());
-  // todo parallel memset to save time
   memset(output_data, 0, output_size * sizeof(float));
 
   int error_code = LiteBackendParallelLaunch(PadImpl, this, context_->thread_num_);

mindspore/lite/src/runtime/kernel/arm/fp32/scatter_nd.cc

@@ -87,7 +87,6 @@ int ScatterNDCPUKernel::ReSize() {
       return RET_ERROR;
     }
   }
-  // todo check indeices out of range
   // for (size_t i = 0; i < static_cast<size_t>(indice_unit_rank); i++) {}
 
   // calculate unit_size_

mindspore/lite/src/runtime/kernel/arm/nnacl/fp16/conv_fp16.c

@@ -332,7 +332,6 @@ void ConvFp16(float16_t *input_data, float16_t *packed_input, float16_t *packed_
   int out_channel = conv_param->output_channel_;
   bool relu = conv_param->is_relu_;
   bool relu6 = conv_param->is_relu6_;
-  // todo
   int thread_count = conv_param->thread_num_;
   const int tile_n = 16;
   int output_count = out_h * out_w;

mindspore/lite/src/runtime/kernel/arm/nnacl/fp32/arithmetic.c

@@ -379,7 +379,6 @@ int BroadcastSub(float *input0, float *input1, float *tile_input0, float *tile_i
   return ElementSub(tile_input0, tile_input1, output, element_size);
 }
 
-// todo c=a/b,if(b==0)
 int ElementDiv(float *input0, float *input1, float *output, int element_size) {
   for (int i = 0; i < element_size; i++) {
     if (input1[i] == 0) {

mindspore/lite/src/runtime/kernel/arm/nnacl/fp32/conv_depthwise.c

@@ -423,7 +423,6 @@ void ConvDw3x3Fp32InputTrans(const float *input_data, float *trans_input, float
   }
 }
 
-// todo yangruoqi: implement assembly
 void ConvDw3x3Fp32Winograd(float *trans_buffer, const float *weight, int out_h_block, int out_w_block) {
   const int unit = 4;
   for (int oh = 0; oh < out_h_block; oh++) {

mindspore/lite/src/runtime/kernel/arm/nnacl/int8/deconv.c

@@ -152,7 +152,7 @@ void DeConvWeightTransInt8(int8_t *src, int8_t *dst, int input_channel, int outp
       }
     }
   } else {
-    /* todo normal int8 deconv */
+    /* normal int8 deconv */
   }
   return;
 }
@@ -171,7 +171,7 @@ void DeConvPackWeightSum(int8_t *weight, int32_t *weight_sum, int32_t input_zp,
       weight_sum[c] = filter_zp * input_zp * deep16 - value * input_zp;
     }
   } else {
-    /* todo normal int8 deconv */
+    /* normal int8 deconv */
   }
   return;
 }
@@ -188,7 +188,7 @@ void DeConvPackInputSum(const int8_t *src, int32_t *dst, int32_t filter_zp, int
       dst[r] = tmp_value * filter_zp;
     }
   } else {
-    /* todo normal int8 deconv */
+    /* normal int8 deconv */
   }
   return;
 }
@@ -199,7 +199,7 @@ int DeConvInt8(const int8_t *input, const int8_t *weight, int32_t *output, int32
   if (matmul_func != NULL) {
     matmul_func(input, weight, output, act_row, act_col, act_deep, input_sum, weight_sum);
   } else {
-    /* todo normal int8 deconv */
+    /* normal int8 deconv */
   }
   return NNACL_OK;
 }
@@ -210,7 +210,7 @@ int DeConvPostInt8(const int32_t *src, const int32_t *bias, int32_t *tmp, int8_t
   if (support_optimize) {
     error_code = DeConvPostInt8C4(src, bias, tmp, out, output_channel, conv_param);
   } else {
-    /* todo normal int8 deconv post */
+    /* normal int8 deconv post */
   }
   return error_code;
 }

mindspore/lite/src/runtime/kernel/arm/nnacl/int8/reduce_int8.c

@@ -377,7 +377,7 @@ int ReduceProdInt8(const int outer_size, const int inner_size, const int axis_si
       if (isAddOverflow(prod, quant->in_zp_)) {
         return NNACL_ERRCODE_ADD_OVERFLOW;
       }
-      *inner_dst = prod + quant->in_zp_;  // todo overflow
+      *inner_dst = prod + quant->in_zp_;
     }
   }
   return NNACL_OK;

mindspore/lite/src/runtime/opencl/opencl_executor.cc

@@ -130,7 +130,6 @@ int OpenCLExecutor::TransformTensorLayoutToBuffer(tensor::Tensor *tensor, schema
     tensor->SetFormat(dst_format);
     return RET_OK;
   } else if (dst_format == schema::Format_NHWC) {
-    // TODO(wandongdong): add support !!
     return RET_OK;
   } else {
     MS_LOG(ERROR) << "Unsupported layout transform: " << schema::EnumNameFormat(tensor->GetFormat()) << " to "
@@ -200,7 +199,6 @@ int OpenCLExecutor::TransformTensorLayoutUint8(tensor::Tensor *tensor, schema::F
   MS_ASSERT(nullptr != tensor);
   MS_ASSERT(4 == tensor->shape().size());
   //  auto src_format = tensor->GetFormat();
-  // todo
   MS_LOG(ERROR) << "Unsupported layout transform: " << schema::EnumNameFormat(tensor->GetFormat()) << " to "
                 << schema::EnumNameFormat(dst_format) << " in uint8";
   return RET_ERROR;

mindspore/lite/test/ut/src/runtime/kernel/arm/common/pack_tests.cc

@@ -138,7 +138,6 @@ TEST_F(TestPack, PackWeightFp32) {
 
 #ifdef ENABLE_FP16
 TEST_F(TestPack, PackInputFp16) {
-  // todo
   size_t input_size;
   std::string input_path = "./test_data/conv/convfp32_input_1_28_28_3.bin";
   auto input_data = reinterpret_cast<float *>(mindspore::lite::ReadFile(input_path.c_str(), &input_size));

mindspore/lite/test/ut/src/runtime/kernel/arm/fp16/convolution_fp16_tests.cc

@@ -328,7 +328,6 @@ TEST_F(TestConvolutionFp16, ConvTest2) {
 TEST_F(TestConvolutionFp16, Conv3x3Test1) {
   auto conv_param = new ConvParameter();
   InitConvParamGroup1Fp16(conv_param);
-  // todo
   int thread_count = 1;
   int tile_num = 16;
   int output_batch = conv_param->output_batch_;
@@ -474,7 +473,6 @@ TEST_F(TestConvolutionFp16, Conv3x3Test1) {
 TEST_F(TestConvolutionFp16, Conv3x3Test2) {
   auto conv_param = new ConvParameter();
   InitConvParamGroup2Fp16(conv_param);
-  // todo
   int thread_count = 1;
   int tile_num = 16;
   int output_batch = conv_param->output_batch_;

mindspore/lite/test/ut/src/runtime/kernel/arm/int8/resize_bilinear_int8_tests.cc

@@ -90,7 +90,7 @@ TEST_F(TestResizeBilinearInt8, Bilinear0) {
   int8_t expect[16] = {4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 6, 6, 5, 5, 6, 6};
 
   Prepare(in_shape, out_shape, input_data, output_data, quant_in, quant_out, align_corners, thread_num);
-  kernel_->Init();  // todo delete
+  kernel_->Init();
   kernel_->Run();
 
   CompareOutputInt8(output_data, expect, 16, err_percent_);

mindspore/lite/test/ut/src/runtime/kernel/arm/int8/resize_nearest_neighbor_int8_tests.cc

@@ -92,7 +92,7 @@ TEST_F(TestResizeNearestNeighborInt8, NearestNeighbor0) {
   err_percent_ = 0.25f;
 
   Prepare(in_shape, out_shape, input_data, output_data, quant_in, quant_out, false, thread_num);
-  kernel_->Init();  // todo delete
+  kernel_->Init();
   kernel_->Run();
 
   CompareOutputInt8(output_data, expect, 16, err_percent_);

mindspore/lite/tools/converter/legacy_optimizer/fusion/batchnorm_fold_fusion_pass.cc

@@ -348,7 +348,7 @@ STATUS BatchNormFoldFusionPass::GenNewBiasTensor() {  // bias has no quant
     MS_LOG(ERROR) << "new BiasTensor failed";
     return RET_ERROR;
   }
-  newBiasTensor->dataType = 0;  // todo is float
+  newBiasTensor->dataType = 0;
   newBiasTensor->format = Format_NUM_OF_FORMAT;
   newBiasTensor->refCount = schema::NodeType_ValueNode;
   newBiasTensor->dims = biasShape;

mindspore/lite/tools/converter/legacy_optimizer/graph/weight_format_hardcode_pass.cc

@@ -29,7 +29,7 @@ void WeightFormatHardCodePass::SetFmkType(converter::FmkType fmkType) { this->fm
 // pre set tensor format
 // non quant, filterFormat:
 //           conv     deconv     depth     dedepth
-// caffe   K(C/g)HW  C(K/g)HW      /         /     // todo with deconvOp
+// caffe   K(C/g)HW  C(K/g)HW      /         /
 // tf        HWCK     HWKC       HWCK      HWKC
 // onnx    K(C/g)HW  C(K/g)HW      /         /
 

mindspore/lite/tools/converter/parser/caffe/caffe_model_parser.cc

@@ -174,7 +174,6 @@ STATUS CaffeModelParser::ParseLayer(const caffe::NetParameter &proto, const caff
         break;
       }
     }
-    // todo y00520784 : layer.input_param().shape(0)
     if (layer.type() == "Input") {
       std::unique_ptr<schema::TensorT> msTensor = std::make_unique<schema::TensorT>();
       for (int j = 0; j < layer.input_param().shape(0).dim_size(); j++) {

mindspore/lite/tools/converter/parser/caffe/caffe_scale_parser.cc

@@ -43,7 +43,6 @@ STATUS CaffeScaleParser::Parse(const caffe::LayerParameter &proto, const caffe::
   attr->axis = axis;
 
   // parse scale
-  // todo expect only weight as scale not bias
   if (weight.blobs().size() == 1) {
     auto scale = ConvertWeight(weight.blobs(0));
     if (scale == nullptr) {

mindspore/lite/tools/converter/parser/onnx/onnx_arithmetic_operation_parser.cc

@@ -66,8 +66,7 @@ STATUS OnnxDivParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::Node
 STATUS OnnxPowParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::NodeProto &onnx_node, schema::CNodeT *op) {
   MS_LOG(DEBUG) << "onnx PowParser";
   if (op != nullptr) {
-    // TODO(wangzhe) attr power need populate
-    std::unique_ptr<schema::PowerT> attr = std::make_unique<schema::PowerT>();
+    std::unique_ptr<schema::PowerT> attr(new schema::PowerT());
     op->primitive = std::make_unique<schema::PrimitiveT>();
     op->primitive->value.type = schema::PrimitiveType_Power;
     op->primitive->value.value = attr.release();

mindspore/lite/tools/converter/parser/onnx/onnx_conv_parser.cc

@@ -65,7 +65,6 @@ STATUS OnnxConvParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::Nod
         MS_LOG(ERROR) << "dilations size " << onnx_node_attr.ints().size() << " is not 2";
         return RET_ERROR;
       }
-      // TODO(wangzhe) verify the change
       attr->dilateH = static_cast<int32_t>(onnx_node_attr.ints(0));
       attr->dilateW = static_cast<int32_t>(onnx_node_attr.ints(1));
     } else if (onnx_node_attr.name() == "kernels") {
@@ -80,7 +79,6 @@ STATUS OnnxConvParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::Nod
         MS_LOG(ERROR) << "kernel_shape size " << onnx_node_attr.ints().size() << " is not 2";
         return RET_ERROR;
       }
-      // TODO(wangzhe) verify the change
       attr->kernelH = static_cast<int32_t>(onnx_node_attr.ints(0));
       attr->kernelW = static_cast<int32_t>(onnx_node_attr.ints(1));
     } else if (onnx_node_attr.name() == "auto_pad") {
@@ -99,7 +97,6 @@ STATUS OnnxConvParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::Nod
         MS_LOG(ERROR) << "strides size " << onnx_node_attr.ints().size() << " is not 2";
         return RET_ERROR;
       }
-      // TODO(wangzhe) verify the change
       attr->strideH = static_cast<int32_t>(onnx_node_attr.ints(0));
       attr->strideW = static_cast<int32_t>(onnx_node_attr.ints(1));
     } else if (onnx_node_attr.name() == "order") {
@@ -143,7 +140,6 @@ STATUS OnnxConvParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::Nod
       dims.insert(dims.begin(), iter->ints().begin(), iter->ints().end());
     }
     attr->channelOut = dims[0];
-    // TODO(wangzhe) verify this code
     attr->channelIn = dims[3] * attr->group;
   }
   attr->format = schema::Format_NCHW;

mindspore/lite/tools/converter/parser/onnx/onnx_model_parser.cc

@@ -241,7 +241,6 @@ STATUS OnnxModelParser::ParseOnnxGivenFillNode(const onnx::NodeProto &onnx_node,
       std::for_each(shape.begin(), shape.end(), [&data_count](int dim) { data_count *= dim; });
       size_t data_size = 0;
       if (onnx_node.op_type() == "Int8GivenIntTensorFill") {
-        // todo  how to read onnx-ori-dataType
         tensor->dataType = kNumberTypeInt32;
         data_size = data_count * sizeof(int32_t) / sizeof(uint8_t);
         tensor->data.resize(data_size);
@@ -252,9 +251,7 @@ STATUS OnnxModelParser::ParseOnnxGivenFillNode(const onnx::NodeProto &onnx_node,
           castedTensorData[i] = int32_t(iter->ints().data()[i]);
         }
       } else if (onnx_node.op_type() == "Int8GivenTensorFill") {
-        // todo  how to read onnx-ori-dataType
         tensor->dataType = kNumberTypeUInt8;
-        // todo: add * sizof(string)
         data_size = data_count;
         tensor->data.resize(data_size);
         MS_LOG(DEBUG) << "tensor data size " << data_size << ", s: " << sizeof(iter->s().data());

mindspore/lite/tools/converter/parser/onnx/onnx_relu_parser.cc

@@ -65,7 +65,7 @@ STATUS OnnxPReluParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::No
   if (slope_size == 1) {
     attr->slope.push_back(*slope_raw_data);
     attr->channelShared = true;
-  } else {  // TODO(wangzhe) we don't check input tensor's channel size, this may cause problem
+  } else {
     attr->slope.resize(slope_size);
     attr->channelShared = false;
     if (memcpy_s(attr->slope.data(), slope_size * sizeof(float), slope_raw_data, slope_size * sizeof(float)) != 0) {

mindspore/lite/tools/converter/parser/onnx/onnx_reshape_parser.cc

@@ -26,7 +26,6 @@ STATUS OnnxReshapeParser::Parse(const onnx::GraphProto &onnx_graph, const onnx::
   std::unique_ptr<schema::ReshapeT> attr = std::make_unique<schema::ReshapeT>();
   attr->format = schema::Format_NCHW;
   std::vector<onnx::TensorProto> params;
-  // TODO(wangzhe) shape may also come from other op, there need refactor to introduce tensor_cache
   for (int i = 0; i < onnx_node.input_size(); ++i) {
     const auto &input_name = onnx_node.input(i);
     for (const auto &it : onnx_graph.initializer()) {

mindspore/lite/tools/converter/parser/tflite/schema.fbs

@@ -247,7 +247,7 @@ enum BuiltinOperator : byte {
   SPACE_TO_DEPTH = 26,
   SVDF = 27,
   TANH = 28,
-  // TODO(aselle): Consider rename to CONCATENATE_EMBEDDINGS
+  // Consider rename to CONCATENATE_EMBEDDINGS
   CONCAT_EMBEDDINGS = 29,
   SKIP_GRAM = 30,
   CALL = 31,

mindspore/lite/tools/converter/quantizer/calc_quant_param.cc

@@ -168,7 +168,6 @@ int LinearCalcer::Calc(MetaGraphT *graph, const CNodeT &node) {
       if (outQuantParam->inited) {
         continue;
       }
-      // todo copy quant params
       outTensor->quantParams.front() = std::move(outQuantParam);
     }
   }

mindspore/lite/tools/converter/quantizer/post_training_quantizer.cc

@@ -802,7 +802,6 @@ STATUS PostTrainingQuantizer::CheckTensorVec(const std::string &nodeName,
  **/
 STATUS PostTrainingQuantizer::DoInference() {
   for (size_t i = 0; i < calibrator_->GetBatchNum(); i++) {
-    // TODO(x) when model has inputs count > 1
     // get input tensor
     vector<mindspore::tensor::MSTensor *> inputs = session_->GetInputs();
     if (inputs.size() > 1) {
@@ -854,7 +853,6 @@ STATUS PostTrainingQuantizer::DoInference() {
 
 STATUS PostTrainingQuantizer::CollectDataFrequency() {
   for (size_t i = 0; i < calibrator_->GetBatchNum(); i++) {
-    // TODO(x) when model has inputs count > 1
     // get input tensor
     vector<mindspore::tensor::MSTensor *> inputs = session_->GetInputs();
     if (inputs.size() > 1) {

mindspore/lite/tools/optimizer/common/node_pass_extends.cc

@@ -33,11 +33,11 @@ bool NodePass::Run(const FuncGraphPtr &func_graph) {
   manager->AddFuncGraph(func_graph);
 
   std::unordered_set<AnfNodePtr> seen_node;
-  std::deque<AnfNodePtr> todo{func_graph->output()};
+  std::deque<AnfNodePtr> to_process{func_graph->output()};
   bool changes = false;
-  while (!todo.empty()) {
-    AnfNodePtr node = todo.front();
-    todo.pop_front();
+  while (!to_process.empty()) {
+    AnfNodePtr node = to_process.front();
+    to_process.pop_front();
     if (seen_node.count(node) > 0 || !manager->all_nodes().contains(node)) {
       continue;
     }
@@ -53,15 +53,15 @@ bool NodePass::Run(const FuncGraphPtr &func_graph) {
     if (new_node && IsValueNode<FuncGraph>(new_node)) {
       auto const_func_graph = GetValueNode<FuncGraphPtr>(new_node);
       MS_EXCEPTION_IF_NULL(const_func_graph);
-      todo.push_back(const_func_graph->output());
+      to_process.push_back(const_func_graph->output());
     } else if (new_node && new_node->isa<CNode>()) {
       if (IsGraphKernel(new_node)) {
-        todo.push_back(new_node);
+        to_process.push_back(new_node);
       }
       auto cnode = new_node->cast<CNodePtr>();
       MS_EXCEPTION_IF_NULL(cnode);
       auto inputs = cnode->inputs();
-      (void)todo.insert(todo.end(), inputs.begin(), inputs.end());
+      (void)to_process.insert(to_process.end(), inputs.begin(), inputs.end());
     }
     changes = changes || change;
   }