Merge branch 'chunwei/framework-support-cl' into 'incubate/lite'

framework support cl

See merge request inference/paddlelite!48

paddle/fluid/framework/ir/graph_viz_pass.cc

@@ -18,6 +18,7 @@ limitations under the License. */
 #include <unordered_set>
 #include "paddle/fluid/framework/op_proto_maker.h"
 #include "paddle/fluid/inference/analysis/dot.h"
+#include "paddle/fluid/lite/utils/string.h"
 #include "paddle/fluid/string/printf.h"
 
 namespace paddle {
@@ -84,7 +85,8 @@ void GraphVizPass::ApplyImpl(ir::Graph* graph) const {
   auto marked_nodes = ConsumeMarkedNodes(graph);
   // Create nodes
   for (const Node* n : graph->Nodes()) {
-    std::string node_id = FormatName(n) + "(" + std::to_string(n->id()) + ")";
+    std::string node_id =
+        lite::string_format("%s(%d)", FormatName(n).c_str(), n->id());
     if (n->IsOp()) {
       decltype(op_attrs) attr =
           marked_nodes.count(n) ? marked_op_attrs : op_attrs;

paddle/fluid/inference/analysis/dot.h

@@ -58,9 +58,11 @@ class Dot {
     std::vector<Attr> attrs;
 
     Node(const std::string& name, const std::vector<Attr>& attrs)
-        : name(name),
-          attrs(attrs),
-          id_("node_" + std::to_string(dot_node_counter++)) {}
+        : name(name), attrs(attrs) {
+      std::stringstream ss;
+      ss << "node_" << dot_node_counter++;
+      id_ = ss.str();
+    }
 
     std::string id() const { return id_; }
 

paddle/fluid/lite/CMakeLists.txt

@@ -37,7 +37,7 @@ endfunction()
 function (lite_deps TARGET)
   set(options "")
   set(oneValueArgs "")
-  set(multiValueArgs DEPS X86_DEPS CUDA_DEPS ARM_DEPS PROFILE_DEPS LIGHT_DEPS HVY_DEPS ARGS)
+  set(multiValueArgs DEPS X86_DEPS CUDA_DEPS ARM_DEPS PROFILE_DEPS LIGHT_DEPS HVY_DEPS CL_DEPS ARGS)
   cmake_parse_arguments(lite_deps "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
 
   set(deps ${lite_deps_DEPS})
@@ -78,6 +78,12 @@ function (lite_deps TARGET)
     endforeach(var)
   endif()
 
+  if (LITE_WITH_OPENCL)
+    foreach(var ${lite_deps_CL_DEPS})
+      set(deps ${deps} ${var})
+    endforeach(var)
+  endif()
+
   set(${TARGET} ${deps} PARENT_SCOPE)
 
 endfunction()

paddle/fluid/lite/api/CMakeLists.txt

@@ -53,6 +53,7 @@ if(LITE_WITH_LIGHT_WEIGHT_FRAMEWORK AND WITH_TESTING)
 
     lite_cc_test(test_mobilenetv1_lite SRCS mobilenetv1_test.cc
        DEPS ${lite_model_test_DEPS}
+       CL_DEPS ${opencl_kernels}
        ARGS --model_dir=${LITE_MODEL_DIR}/mobilenet_v1 SERIAL)
     add_dependencies(test_mobilenetv1_lite extern_lite_download_mobilenet_v1_tar_gz)
 

paddle/fluid/lite/api/cxx_api.cc

@@ -23,10 +23,7 @@ namespace paddle {
 namespace lite {
 
 void Predictor::SaveModel(const std::string &dir) {
-#ifndef LITE_WITH_ARM
   MkDirRecur(dir);
-#else
-#endif
   program_->PersistModel(dir, program_desc_);
   LOG(INFO) << "Save model to " << dir;
 }

paddle/fluid/lite/api/mobilenetv1_test.cc

@@ -25,16 +25,13 @@
 namespace paddle {
 namespace lite {
 
-#ifdef LITE_WITH_ARM
-TEST(MobileNetV1, test) {
+void TestModel(const std::vector<Place>& valid_places,
+               const Place& preferred_place) {
   DeviceInfo::Init();
   DeviceInfo::Global().SetRunMode(LITE_POWER_HIGH, FLAGS_threads);
   lite::Predictor predictor;
-  std::vector<Place> valid_places({Place{TARGET(kHost), PRECISION(kFloat)},
-                                   Place{TARGET(kARM), PRECISION(kFloat)}});
 
-  predictor.Build(FLAGS_model_dir, Place{TARGET(kARM), PRECISION(kFloat)},
-                  valid_places);
+  predictor.Build(FLAGS_model_dir, preferred_place, valid_places);
 
   auto* input_tensor = predictor.GetInput(0);
   input_tensor->Resize(DDim(std::vector<DDim::value_type>({1, 3, 224, 224})));
@@ -70,7 +67,26 @@ TEST(MobileNetV1, test) {
   ASSERT_EQ(out->dims()[0], 1);
   ASSERT_EQ(out->dims()[1], 1000);
 }
-#endif
+
+TEST(MobileNetV1, test_arm) {
+  std::vector<Place> valid_places({
+      Place{TARGET(kHost), PRECISION(kFloat)},
+      Place{TARGET(kARM), PRECISION(kFloat)},
+      // Place{TARGET(kOpenCL), PRECISION(kFloat)},
+  });
+
+  TestModel(valid_places, Place({TARGET(kARM), PRECISION(kFloat)}));
+}
+
+TEST(MobileNetV1, test_opencl) {
+  std::vector<Place> valid_places({
+      Place{TARGET(kHost), PRECISION(kFloat)},
+      Place{TARGET(kARM), PRECISION(kFloat)},
+      Place{TARGET(kOpenCL), PRECISION(kFloat)},
+  });
+
+  TestModel(valid_places, Place({TARGET(kOpenCL), PRECISION(kFloat)}));
+}
 
 }  // namespace lite
 }  // namespace paddle

paddle/fluid/lite/core/context.cc

@@ -14,6 +14,10 @@
 
 #include "paddle/fluid/lite/core/context.h"
 
+#ifdef LITE_WITH_OPENCL
+DEFINE_string(cl_path, "/data/local/tmp/opencl", "The OpenCL kernels path.");
+#endif
+
 namespace paddle {
 namespace lite {}  // namespace lite
 }  // namespace paddle

paddle/fluid/lite/core/context.h

@@ -23,6 +23,11 @@
 #include "paddle/fluid/framework/operator.h"
 #include "paddle/fluid/platform/device_context.h"
 #endif
+#ifdef LITE_WITH_OPENCL
+#include "paddle/fluid/lite/opencl/cl_context.h"
+#include "paddle/fluid/lite/opencl/cl_engine.h"
+#include "paddle/fluid/lite/opencl/cl_helper.h"
+#endif
 #include <map>
 #include <memory>
 #include <set>
@@ -34,6 +39,10 @@
 #include "paddle/fluid/lite/core/target_wrapper.h"
 #include "paddle/fluid/lite/utils/all.h"
 
+#ifdef LITE_WITH_OPENCL
+DECLARE_string(cl_path);
+#endif
+
 namespace paddle {
 namespace lite {
 
@@ -44,6 +53,7 @@ using HostContext = Context<TargetType::kHost>;
 using X86Context = Context<TargetType::kX86>;
 using CUDAContext = Context<TargetType::kCUDA>;
 using ARMContext = Context<TargetType::kARM>;
+using OpenClContext = Context<TargetType::kOpenCL>;
 
 template <>
 class Context<TargetType::kHost> {
@@ -51,7 +61,7 @@ class Context<TargetType::kHost> {
   // NOTE: InitOnce should only be used by ContextScheduler
   void InitOnce() {}
 
-  void CopyShared(const HostContext* ctx) {}
+  void CopySharedTo(const HostContext* ctx) {}
 
   std::string name() const { return "HostContext"; }
 };
@@ -69,7 +79,7 @@ class Context<TargetType::kARM> {
   // NOTE: InitOnce should only be used by ContextScheduler
   void InitOnce() { DeviceInfo::Init(); }
 
-  void CopyShared(const ARMContext* ctx) {}
+  void CopySharedTo(const ARMContext* ctx) {}
 
   void SetRunMode(PowerMode mode, int threads) {
     return DeviceInfo::Global().SetRunMode(mode, threads);
@@ -109,7 +119,7 @@ class Context<TargetType::kCUDA> {
     cublas_fp32_ = std::make_shared<lite::cuda::Blas<float>>();
   }
 
-  void CopyShared(const CUDAContext* ctx) {
+  void CopySharedTo(const CUDAContext* ctx) {
     CHECK(ctx);
     CHECK(cublas_fp32_) << "cublas_fp32 should be set first";
     ctx->cublas_fp32_ = cublas_fp32_;
@@ -175,7 +185,7 @@ class Context<TargetType::kX86> {
   // NOTE: InitOnce should only be used by ContextScheduler
   void InitOnce() {}
 
-  void CopyShared(const X86Context* ctx) {}
+  void CopySharedTo(const X86Context* ctx) {}
 
   const device_ctx_t* x86_device_context() { return x86_device_context_.get(); }
   void SetX86DeviceContext(std::unique_ptr<device_ctx_t>&& ctx) {
@@ -202,6 +212,40 @@ class Context<TargetType::kX86> {
 };
 #endif
 
+#ifdef LITE_WITH_OPENCL
+template <>
+class Context<TargetType::kOpenCL> {
+  mutable std::shared_ptr<CLContext> cl_context_;
+  mutable std::shared_ptr<CLHelper> cl_helper_;
+
+ public:
+  CLContext* cl_context() { return cl_context_.get(); }
+  CLHelper* cl_helper() { return cl_helper_.get(); }
+
+  void InitOnce() {
+    // Init cl engine.
+    CHECK(CLEngine::Global()->IsInitSuccess()) << "OpenCL engine init failed";
+    CLEngine::Global()->set_cl_path(FLAGS_cl_path);
+
+    cl_context_ = std::make_shared<CLContext>();
+    cl_helper_ = std::make_shared<CLHelper>();
+    cl_helper_->set_context(cl_context_.get());
+
+    PrepareKernels();
+  }
+
+  void CopySharedTo(const OpenClContext* ctx) {
+    ctx->cl_context_ = cl_context_;
+  }
+
+ private:
+  void PrepareKernels() {
+    cl_helper_->AddKernel("elementwise_add", "elementwise_add_kernel.cl");
+    cl_helper_->AddKernel("pool_max", "pool_kernel.cl");
+  }
+};
+#endif
+
 // Context for running a kernel.
 // Holds the necessary resource and information.
 class KernelContext {
@@ -230,26 +274,32 @@ class ContextScheduler {
     std::unique_ptr<KernelContext> ctx(new KernelContext);
     switch (target) {
       case TARGET(kHost):
-        kernel_contexts_[TargetType::kHost].As<HostContext>().CopyShared(
+        kernel_contexts_[TargetType::kHost].As<HostContext>().CopySharedTo(
             &ctx->As<HostContext>());
         break;
 #ifdef LITE_WITH_X86
       case TARGET(kX86):
-        kernel_contexts_[TargetType::kX86].As<X86Context>().CopyShared(
+        kernel_contexts_[TargetType::kX86].As<X86Context>().CopySharedTo(
             &ctx->As<X86Context>());
         break;
 #endif
 #ifdef LITE_WITH_CUDA
       case TARGET(kCUDA):
-        kernel_contexts_[TargetType::kCUDA].As<CUDAContext>().CopyShared(
+        kernel_contexts_[TargetType::kCUDA].As<CUDAContext>().CopySharedTo(
             &ctx->As<CUDAContext>());
         break;
 #endif
 #ifdef LITE_WITH_ARM
       case TARGET(kARM):
-        kernel_contexts_[TargetType::kARM].As<ARMContext>().CopyShared(
+        kernel_contexts_[TargetType::kARM].As<ARMContext>().CopySharedTo(
             &ctx->As<ARMContext>());
         break;
+#endif
+#ifdef LITE_WITH_OPENCL
+      case TARGET(kOpenCL):
+        kernel_contexts_[TargetType::kOpenCL].As<OpenClContext>().CopySharedTo(
+            &ctx->As<OpenClContext>());
+        break;
 #endif
       default:
         LOG(FATAL) << "unsupported target " << TargetToStr(target);
@@ -273,6 +323,9 @@ class ContextScheduler {
 #endif
 #ifdef LITE_WITH_ARM
     InitContext<TargetType::kARM, ARMContext>();
+#endif
+#ifdef LITE_WITH_OPENCL
+    InitContext<TargetType::kOpenCL, OpenClContext>();
 #endif
   }
 

paddle/fluid/lite/core/mir/fusion/quant_dequant_op_fuser.cc

@@ -15,6 +15,7 @@
 #include "paddle/fluid/lite/core/mir/fusion/quant_dequant_op_fuser.h"
 #include <memory>
 #include <vector>
+#include "paddle/fluid/lite/utils/string.h"
 
 namespace paddle {
 namespace lite {
@@ -57,24 +58,24 @@ void QuantDequantOpFuser::BuildPattern() {
                            ->AsIntermediate();
   std::vector<PMNode*> nodes;
   for (int i = 0; i < times_; i++) {
-    nodes.push_back(VarNode("quantized_op_weight" + std::to_string(i))
+    nodes.push_back(VarNode(string_format("quantized_op_weight%d", i))
                         ->assert_is_op_input(op_type_, weight_name)
                         ->AsInput());
 
-    nodes.push_back(OpNode("quantized_op" + std::to_string(i), op_type_)
+    nodes.push_back(OpNode(string_format("quantized_op%d", i), op_type_)
                         ->assert_is_op(op_type_)
                         ->AsIntermediate());
 
-    nodes.push_back(VarNode("quantized_op_out" + std::to_string(i))
+    nodes.push_back(VarNode(string_format("quantized_op_out%d", i))
                         ->assert_is_op_output(op_type_)
                         ->assert_is_op_input("fake_dequantize_max_abs", "X")
                         ->AsIntermediate());
 
     nodes.push_back(
-        OpNode("dequant_op" + std::to_string(i), "fake_dequantize_max_abs")
+        OpNode(string_format("dequant_op%d", i), "fake_dequantize_max_abs")
             ->assert_is_op("fake_dequantize_max_abs")
             ->AsIntermediate());
-    nodes.push_back(VarNode("dequant_op_out" + std::to_string(i))
+    nodes.push_back(VarNode(string_format("dequant_op_out%d", i))
                         ->assert_is_op_output("fake_dequantize_max_abs", "Out")
                         ->AsOutput());
   }
@@ -108,11 +109,11 @@ void QuantDequantOpFuser::InsertNewNode(SSAGraph* graph,
 
   std::vector<Node*> nodes;
   for (int i = 0; i < times_; i++) {
-    nodes.push_back(matched.at("quantized_op_weight" + std::to_string(i)));
-    nodes.push_back(matched.at("quantized_op" + std::to_string(i)));
-    nodes.push_back(matched.at("quantized_op_out" + std::to_string(i)));
-    nodes.push_back(matched.at("dequant_op" + std::to_string(i)));
-    nodes.push_back(matched.at("dequant_op_out" + std::to_string(i)));
+    nodes.push_back(matched.at(string_format("quantized_op_weight%d", i)));
+    nodes.push_back(matched.at(string_format("quantized_op%d", i)));
+    nodes.push_back(matched.at(string_format("quantized_op_out%d", i)));
+    nodes.push_back(matched.at(string_format("dequant_op%d", i)));
+    nodes.push_back(matched.at(string_format("dequant_op_out%d", i)));
   }
   int bit_length = quant_op->stmt()->op_info()->GetAttr<int>("bit_length");
   auto* scope = quant_op->stmt()->op()->scope();

paddle/fluid/lite/core/mir/graph_visualize_pass.cc

@@ -17,6 +17,7 @@
 #include <set>
 #include <string>
 #include "paddle/fluid/lite/core/mir/pass_registry.h"
+#include "paddle/fluid/lite/utils/string.h"
 
 namespace paddle {
 namespace lite {
@@ -39,7 +40,7 @@ std::string Visualize(mir::SSAGraph* graph) {
     if (node.IsArg()) {
       key = node.AsArg().name;
     } else {
-      key = node.AsStmt().op_type() + std::to_string(id++);
+      key = string_format("%s%d", node.AsStmt().op_type().c_str(), id++);
     }
 
     if (node.IsStmt()) {

paddle/fluid/lite/core/mir/pattern_matcher.cc

@@ -325,7 +325,7 @@ std::string PMPattern::DotString() const {
   // Create Nodes
   std::unordered_map<PMNode *, std::string> node2dot;
   for (const auto &node : nodes()) {
-    std::string node_id = "Node" + std::to_string(id++);
+    std::string node_id = string_format("Node%d", id++);
     dot.AddNode(node_id, {}, node->name());
     node2dot[node.get()] = node_id;
   }

paddle/fluid/lite/core/mir/pattern_matcher.h

@@ -30,6 +30,7 @@
 #include "paddle/fluid/lite/core/mir/node.h"
 #include "paddle/fluid/lite/core/mir/ssa_graph.h"
 #include "paddle/fluid/lite/model_parser/pb/op_desc.h"
+#include "paddle/fluid/lite/utils/string.h"
 
 namespace paddle {
 namespace lite {
@@ -228,7 +229,7 @@ class PMPattern {
   FRIEND_TEST(PMPattern, NewNode);
 #endif
 
-  static std::string NewID() { return "pmnode-" + std::to_string(id_++); }
+  static std::string NewID() { return string_format("pmnode-%d", id_++); }
 
   std::vector<std::unique_ptr<PMNode>> nodes_;
   std::vector<edge_t> edges_;

paddle/fluid/lite/core/mir/type_target_transform_pass.cc

@@ -20,6 +20,7 @@
 #include <vector>
 #include "paddle/fluid/lite/core/mir/graph_visualize_pass.h"
 #include "paddle/fluid/lite/core/mir/pass_registry.h"
+#include "paddle/fluid/lite/utils/string.h"
 
 namespace paddle {
 namespace lite {
@@ -80,7 +81,7 @@ void TypeTargetTransformPass::AddIoCopyInst(
   CHECK(in->IsArg());
   auto node_id = [&] { return graph->nodes().size(); };
   auto io_copy_output_name =
-      in->AsArg().name + "/trans/" + std::to_string(node_id());
+      string_format("%s/trans/%d", in->AsArg().name.c_str(), node_id());
   auto* io_copy_output_arg = graph->NewArgumentNode(io_copy_output_name);
   auto* io_copy_inst = graph->NewInstructNode();
 

paddle/fluid/lite/core/op_lite.cc

@@ -30,6 +30,8 @@ std::vector<std::unique_ptr<KernelBase>> OpLite::CreateKernels(
   auto pick_kernel = [&](const Place &place) {
     auto ks = KernelRegistry::Global().Create(op_type_, place.target,
                                               place.precision, place.layout);
+    VLOG(5) << "pick kernel for " << op_info()->Type() << " " << place
+            << " get " << ks.size() << " kernels";
     for (auto &&it : ks) {
       AttachKernel(it.get());
       kernels.emplace_back(std::move(it));

paddle/fluid/lite/core/op_registry.cc

@@ -62,6 +62,9 @@ std::list<std::unique_ptr<KernelBase>> KernelRegistry::Create(
     case TARGET(kARM): {
       CREATE_KERNEL(kARM);
     } break;
+    case TARGET(kOpenCL): {
+      CREATE_KERNEL(kOpenCL);
+    } break;
     default:
       CHECK(false) << "not supported kernel target " << TargetToStr(target);
   }
@@ -99,6 +102,10 @@ KernelRegistry::KernelRegistry()
   INIT_FOR(kARM, kInt8, kNCHW);
   INIT_FOR(kARM, kAny, kNCHW);
   INIT_FOR(kARM, kAny, kAny);
+
+  INIT_FOR(kOpenCL, kFloat, kNCHW);
+  INIT_FOR(kOpenCL, kAny, kNCHW);
+  INIT_FOR(kOpenCL, kAny, kAny);
 #undef INIT_FOR
 }
 

paddle/fluid/lite/core/op_registry.h

@@ -82,6 +82,10 @@ class KernelRegistry final {
               KernelRegistryForTarget<TARGET(kARM), PRECISION(kFloat),
                                       DATALAYOUT(kNCHW)> *,  //
               KernelRegistryForTarget<TARGET(kARM), PRECISION(kInt8),
+                                      DATALAYOUT(kNCHW)> *,  //
+              KernelRegistryForTarget<TARGET(kOpenCL), PRECISION(kFloat),
+                                      DATALAYOUT(kNCHW)> *,  //
+              KernelRegistryForTarget<TARGET(kOpenCL), PRECISION(kInt8),
                                       DATALAYOUT(kNCHW)> *  //
               >;
 

paddle/fluid/lite/core/optimizer.h

@@ -52,11 +52,20 @@ class Optimizer {
       RunPasses(std::vector<std::string>{{
           "lite_quant_dequant_fuse_pass",  //
           "lite_conv_bn_fuse_pass",        //
+// This pass is disabled to force some opencl kernels selected for final
+// running, otherwise, they will be fused to ARM fusion kernels, and the OpenCL
+// devices will be discarded.
+// TODO(Superjomn) Refine the fusion related design to select fusion kernels for
+// devices automatically.
+#ifndef LITE_WITH_OPENCL
           "lite_conv_elementwise_add_activation_fuse_pass",  //
+#endif
           "lite_fc_fuse_pass",              //
           "identity_scale_eliminate_pass",  //
 #ifdef LITE_WITH_LIGHT_WEIGHT_FRAMEWORK
+#ifndef LITE_WITH_OPENCL
           "lite_elementwise_add_activation_fuse_pass",  //
+#endif
 #endif
           "static_kernel_pick_pass",        //
           "variable_place_inference_pass",  //

paddle/fluid/lite/core/program.h

@@ -140,7 +140,8 @@ class RuntimeProgram {
 
   void Run() {
     for (auto& inst : instructions_) {
-      VLOG(3) << ">> Running kernel: " << inst.op()->op_info()->Repr();
+      VLOG(4) << ">> Running kernel: " << inst.op()->op_info()->Repr()
+              << " on Target " << TargetToStr(inst.kernel()->target());
       inst.Run();
     }
   }

paddle/fluid/lite/core/target_wrapper.h

@@ -31,6 +31,7 @@ enum class TargetType : int {
   kX86,
   kCUDA,
   kARM,
+  kOpenCL,
   kAny,  // any target
   NUM,   // number of fields.
 };
@@ -69,8 +70,8 @@ static size_t PrecisionTypeLength(PrecisionType type) {
 #define DATALAYOUT(item__) paddle::lite::DataLayoutType::item__
 
 static const std::string& TargetToStr(TargetType target) {
-  static const std::string target2string[] = {"unk",  "host", "x86",
-                                              "cuda", "arm",  "any"};
+  static const std::string target2string[] = {"unk", "host",   "x86", "cuda",
+                                              "arm", "opencl", "any"};
   auto x = static_cast<int>(target);
   CHECK_LT(x, static_cast<int>(TARGET(NUM)));
   return target2string[x];
@@ -92,8 +93,8 @@ static const std::string& DataLayoutToStr(DataLayoutType layout) {
 }
 
 static const std::string& TargetRepr(TargetType target) {
-  static const std::string target2string[] = {"kUnk", "kHost", "kX86", "kCUDA",
-                                              "kAny"};
+  static const std::string target2string[] = {
+      "kUnk", "kHost", "kX86", "kCUDA", "kARM", "kOpenCL", "kAny"};
   auto x = static_cast<int>(target);
   CHECK_LT(x, static_cast<int>(TARGET(NUM)));
   return target2string[x];

paddle/fluid/lite/kernels/CMakeLists.txt

@@ -4,5 +4,5 @@ add_subdirectory(host)
 add_subdirectory(arm)
 add_subdirectory(cuda)
 add_subdirectory(x86)
- 
+add_subdirectory(opencl)
  

paddle/fluid/lite/kernels/opencl/CMakeLists.txt

+if (NOT LITE_WITH_OPENCL)
+    return ()
+endif()
+
+set(cl_kernel_deps op_params_lite cl_caller cl_engine cl_context cl_wrapper)
+
+cc_library(elementwise_add_opencl SRCS elementwise_add_compute.cc DEPS ${cl_kernel_deps})
+
+lite_cc_test(test_elementwise_add_opencl SRCS elementwise_add_compute_test.cc DEPS elementwise_add_opencl
+  op_registry_lite program_lite
+  context_lite
+  )
+
+set(opencl_kernels
+  elementwise_add_opencl
+  CACHE INTERNAL "")

paddle/fluid/lite/kernels/opencl/elementwise_add_compute.cc

+// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "paddle/fluid/lite/core/kernel.h"
+#include "paddle/fluid/lite/core/op_registry.h"
+#include "paddle/fluid/lite/operators/op_params.h"
+// NOTE ugly here, hide these.
+#include "paddle/fluid/lite/opencl/cl_caller.h"
+
+namespace paddle {
+namespace lite {
+namespace kernels {
+namespace opencl {
+
+class ElementwiseAddCompute
+    : public KernelLite<TARGET(kOpenCL), PRECISION(kFloat), DATALAYOUT(kNCHW)> {
+ public:
+  using param_t = operators::ElementwiseParam;
+
+  void Run() override {
+    auto& param = *param_.get_mutable<param_t>();
+    auto& context = ctx_->As<OpenClContext>();
+    CHECK(context.cl_context());
+
+    elementwise_add(
+        context.cl_context(), static_cast<const float*>(param.X->raw_data()),
+        param.X->dims(), static_cast<const float*>(param.Y->raw_data()),
+        param.Y->dims(), param.Out->mutable_data<float>(), param.Out->dims());
+  }
+};
+
+}  // namespace opencl
+}  // namespace kernels
+}  // namespace lite
+}  // namespace paddle
+
+REGISTER_LITE_KERNEL(elementwise_add, kOpenCL, kFloat, kNCHW,
+                     paddle::lite::kernels::opencl::ElementwiseAddCompute, def)
+    .BindInput("X", {LiteType::GetTensorTy(TARGET(kHost))})
+    .BindInput("Y", {LiteType::GetTensorTy(TARGET(kHost))})
+    .BindOutput("Out", {LiteType::GetTensorTy(TARGET(kHost))})
+    .Finalize();

paddle/fluid/lite/kernels/opencl/elementwise_add_compute_test.cc

+// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <gtest/gtest.h>
+#include "paddle/fluid/lite/core/op_registry.h"
+
+namespace paddle {
+namespace lite {
+
+TEST(elementwise_add, init) {
+  LOG(INFO) << "to get kernel ...";
+  auto kernels = KernelRegistry::Global().Create(
+      "elementwise_add", TARGET(kOpenCL), PRECISION(kFloat), DATALAYOUT(kNCHW));
+  ASSERT_FALSE(kernels.empty());
+
+  auto kernel = std::move(kernels.front());
+
+  LOG(INFO) << "get kernel";
+
+  lite::Tensor X, Y, Out;
+  operators::ElementwiseParam param;
+  param.X = &X;
+  param.Y = &Y;
+  param.Out = &Out;
+
+  std::unique_ptr<KernelContext> context(new KernelContext);
+  context->As<OpenClContext>().InitOnce();
+
+  kernel->SetParam(param);
+  kernel->SetContext(std::move(context));
+
+  X.Resize({1, 10});
+  Y.Resize({1, 10});
+  Out.Resize({1, 10});
+
+  auto* x_data = X.mutable_data<float>();
+  auto* y_data = Y.mutable_data<float>();
+  auto* out_data = Out.mutable_data<float>();
+
+  for (int i = 0; i < 10; i++) {
+    x_data[i] = 1.1 * i;
+    y_data[i] = 2.3 * i;
+  }
+
+  kernel->Launch();
+
+  for (int i = 0; i < 10; i++) {
+    EXPECT_NEAR(out_data[i], 3.4 * i, 1e-1);
+  }
+}
+
+}  // namespace lite
+}  // namespace paddle
+
+USE_LITE_KERNEL(elementwise_add, kOpenCL, kFloat, kNCHW, def);

paddle/fluid/lite/kernels/use_kernels.h

@@ -61,3 +61,6 @@ USE_LITE_KERNEL(mul, kCUDA, kFloat, kNCHW, def);
 USE_LITE_KERNEL(io_copy, kCUDA, kAny, kAny, host_to_device);
 USE_LITE_KERNEL(io_copy, kCUDA, kAny, kAny, device_to_host);
 #endif
+#ifdef LITE_WITH_OPENCL
+USE_LITE_KERNEL(elementwise_add, kOpenCL, kFloat, kNCHW, def);
+#endif

paddle/fluid/lite/opencl/CMakeLists.txt

@@ -2,18 +2,16 @@ if (NOT LITE_WITH_OPENCL)
     return()
 endif()
 
-if (WITH_LITE AND LITE_WITH_LIGHT_WEIGHT_FRAMEWORK)
-    cc_library(cl_wrapper SRCS cl_wrapper.cc)
-    cc_library(cl_tool SRCS cl_tool.cc)
-    target_compile_options(cl_tool BEFORE PUBLIC -Wno-ignored-qualifiers)
-    cc_library(cl_half SRCS cl_half.cc)
-    target_compile_options(cl_half BEFORE PUBLIC -fno-strict-aliasing)
-    cc_library(cl_engine SRCS cl_engine.cc DEPS cl_tool)
-    cc_library(cl_context SRCS cl_context.cc DEPS cl_engine)
-    cc_library(cl_helper SRCS cl_helper.cc DEPS cl_context)
-    cc_library(cl_image_converter SRCS cl_image_converter.cc DEPS cl_half lite_tensor)
-    cc_library(cl_image SRCS cl_image.cc DEPS cl_half lite_tensor cl_image_converter cl_engine)
-    cc_library(cl_caller SRCS cl_caller.cc  DEPS cl_helper cl_image)
-    lite_cc_test(test_cl_runtime SRCS cl_test.cc DEPS cl_helper cl_image cl_caller cl_wrapper)
-    add_dependencies(cl_tool opencl_clhpp)
-endif()
+cc_library(cl_wrapper SRCS cl_wrapper.cc)
+cc_library(cl_tool SRCS cl_tool.cc)
+target_compile_options(cl_tool BEFORE PUBLIC -Wno-ignored-qualifiers)
+cc_library(cl_half SRCS cl_half.cc)
+target_compile_options(cl_half BEFORE PUBLIC -fno-strict-aliasing)
+cc_library(cl_engine SRCS cl_engine.cc DEPS cl_tool)
+cc_library(cl_context SRCS cl_context.cc DEPS cl_engine)
+cc_library(cl_helper SRCS cl_helper.cc DEPS cl_context)
+cc_library(cl_image_converter SRCS cl_image_converter.cc DEPS cl_half lite_tensor)
+cc_library(cl_image SRCS cl_image.cc DEPS cl_half lite_tensor cl_image_converter cl_engine)
+cc_library(cl_caller SRCS cl_caller.cc  DEPS cl_helper cl_image)
+lite_cc_test(test_cl_runtime SRCS cl_test.cc DEPS cl_helper cl_image cl_caller cl_wrapper)
+add_dependencies(cl_tool opencl_clhpp)

paddle/fluid/lite/opencl/cl_caller.cc

@@ -49,12 +49,12 @@ bool InitOpenCLEngine(std::string cl_path) {
   return engine->IsInitSuccess();
 }
 
-void elementwise_add(CLContext* context, float* in, const DDim& in_dim,
-                     float* bias, const DDim& bias_dim, float* out,
+void elementwise_add(CLContext* context, const float* in, const DDim& in_dim,
+                     const float* bias, const DDim& bias_dim, float* out,
                      const DDim& out_dim) {
   CLHelper helper(context);
   helper.AddKernel("elementwise_add", "elementwise_add_kernel.cl");
-  auto kernel = helper.KernelAt(0);
+  auto kernel = helper.GetKernel(0);
   CLImage in_image;
   in_image.set_tensor_data(in, in_dim);
   in_image.InitNormalCLImage(helper.OpenCLContext());

paddle/fluid/lite/opencl/cl_caller.h

@@ -22,8 +22,13 @@ namespace paddle {
 namespace lite {
 
 bool InitOpenCLEngine(std::string cl_path);
-void elementwise_add(CLContext* context, float* in, const DDim& in_dim,
-                     float* bias, const DDim& bias_dim, float* out,
+
+/// An elementwise_add method to embed OpenCL logic inside, it is used as a
+/// black box so that the framework can remain simple.
+/// NOTE Currently, these methods are quite expensive, we will optimize them
+/// latter.
+void elementwise_add(CLContext* context, const float* in, const DDim& in_dim,
+                     const float* bias, const DDim& bias_dim, float* out,
                      const DDim& out_dim);
 
 }  // namespace lite

paddle/fluid/lite/opencl/cl_helper.cc

@@ -29,17 +29,18 @@ void CLHelper::AddKernel(const std::string &kernel_name,
   CHECK(context_ != nullptr) << "Please use set_context first!";
   VLOG(3) << " --- begin to add kernel ---";
   auto kernel = context_->GetKernel(kernel_name, file_name, options);
-  kernels.emplace_back(std::move(kernel));
+  kernels_.emplace_back(std::move(kernel));
+  kernel_offset_[kernel_name] = kernels_.size() - 1;
   VLOG(3) << " --- end to add kernel --- ";
 }
 
-cl::Kernel &CLHelper::KernelAt(const int index) {
-  VLOG(3) << " --- kernel count: " << kernels.size() << " --- ";
-  CHECK(static_cast<size_t>(index) < kernels.size())
+cl::Kernel &CLHelper::GetKernel(const int index) {
+  VLOG(3) << " --- kernel count: " << kernels_.size() << " --- ";
+  CHECK(static_cast<size_t>(index) < kernels_.size())
       << "The index must be less than the size of kernels.";
-  CHECK(kernels[index] != nullptr)
+  CHECK(kernels_[index] != nullptr)
       << "The target kernel pointer cannot be null.";
-  return *(kernels[index]);
+  return *(kernels_[index]);
 }
 
 cl::CommandQueue &CLHelper::OpenCLCommandQueue() {

paddle/fluid/lite/opencl/cl_helper.h

@@ -14,6 +14,7 @@ limitations under the License. */
 
 #pragma once
 
+#include <map>
 #include <memory>
 #include <string>
 #include <vector>
@@ -35,7 +36,12 @@ class CLHelper {
   void AddKernel(const std::string &kernel_name, const std::string &file_name,
                  const std::string &options = "");
 
-  cl::Kernel &KernelAt(const int index);
+  cl::Kernel &GetKernel(const int index);
+  cl::Kernel &GetKernel(const std::string &name) {
+    auto it = kernel_offset_.find(name);
+    CHECK(it != kernel_offset_.end());
+    return GetKernel(it->second);
+  }
 
   cl::CommandQueue &OpenCLCommandQueue();
 
@@ -45,7 +51,8 @@ class CLHelper {
 
  private:
   CLContext *context_{nullptr};
-  std::vector<std::unique_ptr<cl::Kernel>> kernels;
+  std::map<std::string, int> kernel_offset_;
+  std::vector<std::unique_ptr<cl::Kernel>> kernels_;
 };
 
 }  // namespace lite

paddle/fluid/lite/opencl/cl_image.cc

@@ -53,7 +53,7 @@ std::ostream& operator<<(std::ostream& os, const CLImage& cl_image) {
   return os;
 }
 
-void CLImage::set_tensor_data(float* tensor_data, const DDim& dim) {
+void CLImage::set_tensor_data(const float* tensor_data, const DDim& dim) {
 #ifdef LITE_WITH_LIGHT_WEIGHT_FRAMEWORK
   auto numel = dim.product();
 #else

paddle/fluid/lite/opencl/cl_image.h

@@ -33,7 +33,7 @@ class CLImage {
   /*
    * Will not hold input tensor data, memcpy in this method.
    * */
-  void set_tensor_data(float* tensor_data, const DDim& dim);
+  void set_tensor_data(const float* tensor_data, const DDim& dim);
 
   bool IsInit() { return initialized_; }
   /*

paddle/fluid/lite/opencl/cl_test.cc

@@ -65,7 +65,7 @@ TEST(cl_test, kernel_test) {
   helper->AddKernel("elementwise_add", "elementwise_add_kernel.cl");
   helper->AddKernel("pool_max", "pool_kernel.cl");
   helper->AddKernel("elementwise_add", "elementwise_add_kernel.cl");
-  auto kernel = helper->KernelAt(2);
+  auto kernel = helper->GetKernel(2);
 
   std::unique_ptr<float[]> in_data(new float[1024 * 512]);
   for (int i = 0; i < 1024 * 512; i++) {