fix:delete useless code

.gitignore

@@ -104,10 +104,3 @@ metal/paddle-mobile-demo/paddle-mobile-demo/Resources
 metal/paddle-mobile-demo/paddle-mobile-demo/Resources/images
 metal/paddle-mobile-demo/paddle-mobile-demo/Resources/models
 metal/MobileNetDemo/MobileNetDemo/Resources
-
-# generated files
-lite/api/paddle_use_kernels.h
-lite/api/paddle_use_ops.h
-lite/backends/arm/math/dotprod/gemm_sdot.h
-lite/tools/cmake_tools/ast.pyc
-

CMakeLists.txt

@@ -22,8 +22,6 @@ if (WITH_PADDLE_MOBILE)
     return()
 endif(WITH_PADDLE_MOBILE)
 
-# set(CMAKE_BUILD_TYPE DEBUG)
-
 set(PADDLE_SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR})
 set(PADDLE_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR})
 set(CMAKE_CXX_STANDARD 11)

fpga.sh

-./lite/tools/build.sh \
-  --arm_os=armlinux \
-  --arm_abi=armv8 \
-  --arm_lang=gcc \
-  test

lite/backends/fpga/KD/llapi/filter.cpp

@@ -221,7 +221,6 @@ int8_t* format_filter(float* data_in,
       align_to_x(num_per_div_before_alignment, filter_num_alignment);
   int div_num =
       (num + num_per_div_before_alignment - 1) / num_per_div_before_alignment;
-  // int num_after_alignment = num_per_div_after_alignment * div_num;
   int residual = num % num_per_div_before_alignment;
   int num_after_alignment = num_per_div_after_alignment *
                                 ((residual == 0) ? div_num : (div_num - 1)) +

lite/backends/fpga/KD/llapi/zynqmp_api.cpp

@@ -62,7 +62,6 @@ void reset_device() {
 // memory management;
 void *fpga_malloc(size_t size) {
 #ifdef ENABLE_DEBUG
-  std::cout << "fpga_malloc:" << size << std::endl;
 #endif
 #ifdef PADDLE_OS_LINUX
   void *ptr = reinterpret_cast<void *>(

lite/backends/fpga/KD/pes/conv_process.hpp

@@ -364,7 +364,6 @@ inline void split_filter_num(const ConvParam& c_param) {
     args.image.height = input->shape().height();
     args.image.pad_width = param.paddings[1];
     args.image.pad_height = param.paddings[0];
-
     args.dilation = param.dilations[0];
 
     args.output.address = out_address;
@@ -420,7 +419,6 @@ inline void split_channel(const ConvParam& c_param) {
     }
     scale.flush();
     bias.flush();
-    // Shape sb_shape(N, {2 * channel});
     format_scale_bias(&scale,
                       &bias,
                       &conv_param->filter,
@@ -448,7 +446,6 @@ inline void split_channel(const ConvParam& c_param) {
     args.image.height = conv_param->input.shape().height();
     args.image.pad_width = param.paddings[1];
     args.image.pad_height = param.paddings[0];
-
     args.dilation = param.dilations[0];
     args.output.address = conv_param->output.mutableData<void>();
     args.output.scale_address = conv_param->output.scale();
@@ -479,7 +476,6 @@ inline bool compute_conv(const ConvParam& c_conv_params) {
   }
   size_t size = params.size();
   if (ret == 0 && size > 1) {
-    // Tensor* output = conv_params.output;
     Tensor& img = params[0]->output;
     for (int i = 0; i < 1; i++) {
       for (int i = 0; i < img.shape().numel(); i++) {

lite/backends/fpga/KD/pes/depthwise_conv_pe.hpp

@@ -62,7 +62,6 @@ class DepthwiseConvPE : public PE {
       float16* scale_data = param_.scale()->data<float16>();
       float16* filter_data = param.quantizedFilter()->mutableData<float16>(
           FP16, param.filter->shape());
-
       memcpy(filter_data,
              scale_data,
              param.filter->shape().numel() * sizeof(float16));

lite/backends/fpga/KD/pes/gru_pe.hpp

@@ -119,7 +119,6 @@ class GRUPE : public PE {
         prev_hidden_.copyFrom(value.pre_output);
       }
       mul_pe_.dispatch();
-      // reset_hidden_.saveToFile("reset_hidden_.txt");
       update_gate_data += stride_update;
       reset_gate_data += stride_update;
 
@@ -170,7 +169,6 @@ class GRUPE : public PE {
   zynqmp::Tensor bias_;
   zynqmp::Tensor weight_;
   zynqmp::Tensor state_weight_;
-
   zynqmp::Tensor update_gate_;
   zynqmp::Tensor reset_gate_;
   zynqmp::Tensor cell_state_;

lite/backends/fpga/KD/tensor.hpp

@@ -348,19 +348,9 @@ class Tensor {
     if (placeHolder_ == nullptr) {
       return;
     }
-    std::cout << scale()[0] << " , " << scale()[1] << std::endl;
   }
 
-  void printScale(std::string type) {
-    std::cout << type << " : "
-              << std::to_string(shape_->num()) + "_" +
-                     std::to_string(shape_->channel()) + "_" +
-                     std::to_string(shape_->height()) + "_" +
-                     std::to_string(shape_->width())
-              << std::endl;
-    std::cout << type << " \n";
-    printScale();
-  }
+  void printScale(std::string type) { printScale(); }
 
   std::string dimsFileName() {
     return std::to_string(shape_->num()) + "_" +
@@ -388,7 +378,6 @@ class Tensor {
     static int counter = 0;
     std::string npath = std::to_string(counter) + "_" + path;
     counter++;
-    std::cout << "======== saving file:" << npath << " ============\n";
     save_file_with_name(npath);
   }
 

lite/backends/fpga/lite_tensor.h

@@ -165,9 +165,6 @@ class TensorLite {
 
   TargetType target() const { return target_; }
 
-  // template <typename T>
-  // TensorLite Slice(int64_t begin, int64_t end) const;
-
   zynqmp::Tensor *ZynqTensor() const { return zynq_tensor_; }
 
   friend std::ostream &operator<<(std::ostream &os, const TensorLite &tensor) {
@@ -257,7 +254,6 @@ TensorLite TensorLite::Slice(int64_t begin, int64_t end) const {
     int64_t base = numel() / dims_[0];
 
     TensorLite dst;
-
     dst.target_ = target_;
     auto dst_dims = dims_;
     dst_dims[0] = end - begin;

lite/core/program.cc

@@ -146,7 +146,7 @@ void RuntimeProgram::Run() {
 #ifdef LITE_WITH_PROFILE
 #ifdef LITE_WITH_PRECISION_PROFILE
 #ifndef LITE_WITH_FPGA
-// LITE_PRECISION_PROFILE(inst)
+    LITE_PRECISION_PROFILE(inst)
 #endif
 #endif  // LITE_WITH_PRECISION_PROFILE
 #endif  // LITE_WITH_PROFILE

lite/kernels/arm/lookup_table_compute.cc

@@ -28,7 +28,6 @@ namespace arm {
 
 void LookupTableCompute::Run() {
   auto& param = this->Param<param_t>();
-  auto& ctx = this->ctx_->template As<ARMContext>();
   // inputs
   auto w = param.W;
   auto ids = param.Ids;
@@ -76,3 +75,13 @@ REGISTER_LITE_KERNEL(lookup_table,
     .BindInput("Ids", {LiteType::GetTensorTy(TARGET(kARM))})
     .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kARM))})
     .Finalize();
+REGISTER_LITE_KERNEL(lookup_table_v2,
+                     kARM,
+                     kFloat,
+                     kNCHW,
+                     paddle::lite::kernels::arm::LookupTableCompute,
+                     def)
+    .BindInput("W", {LiteType::GetTensorTy(TARGET(kARM))})
+    .BindInput("Ids", {LiteType::GetTensorTy(TARGET(kARM))})
+    .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kARM))})
+    .Finalize();

lite/kernels/fpga/io_copy_compute.cc

@@ -191,8 +191,6 @@ class IoCopyFpgaToHostCHWCompute
     param.y->ZynqTensor()->flush();
     auto out_lod = param.y->mutable_lod();
     *out_lod = param.x->lod();
-    // param.x->ZynqTensor()->saveToFile("io_x", true);
-    // param.y->ZynqTensor()->saveToFile("io_y", true);
   }
   std::string doc() const override { return "Copy IO from FPGA to HOST"; }
 };

lite/kernels/fpga/prior_box_compute.cc

@@ -78,7 +78,6 @@ void PriorBoxCompute::PrepareForRun() {
 
   param.boxes->mutable_data<float>();
   param.variances->mutable_data<float>();
-
   zynqmp::PriorBoxParam& priobox_param = pe_.param();
   priobox_param.input = param.input->ZynqTensor();
   priobox_param.image = param.image->ZynqTensor();

lite/kernels/host/reshape_compute.cc

@@ -63,26 +63,6 @@ REGISTER_LITE_KERNEL(reshape,
                                        DATALAYOUT(kAny))})
     .Finalize();
 
-// REGISTER_LITE_KERNEL(reshape,
-//                      kFPGA,
-//                      kFP16,
-//                      kNHWC,
-//                      paddle::lite::kernels::host::ReshapeCompute,
-//                      def)
-//     .BindInput("X",
-//                {LiteType::GetTensorTy(
-//                    TARGET(kFPGA), PRECISION(kFP16), DATALAYOUT(kNHWC))})
-//     .BindInput("ShapeTensor",
-//                {LiteType::GetTensorTy(
-//                    TARGET(kHost), PRECISION(kAny), DATALAYOUT(kAny))})
-//     .BindInput("Shape",
-//                {LiteType::GetTensorTy(
-//                    TARGET(kHost), PRECISION(kAny), DATALAYOUT(kAny))})
-//     .BindOutput("Out",
-//                 {LiteType::GetTensorTy(
-//                     TARGET(kHost), PRECISION(kFloat), DATALAYOUT(kNCHW))})
-//     .Finalize();
-
 REGISTER_LITE_KERNEL(reshape2,
                      kHost,
                      kAny,

lite/tools/build.sh

@@ -13,7 +13,7 @@ readonly NUM_PROC=${LITE_BUILD_THREADS:-4}
 
 
 # global variables
-BUILD_EXTRA=ON
+BUILD_EXTRA=OFF
 BUILD_JAVA=ON
 BUILD_PYTHON=OFF
 BUILD_DIR=$(pwd)