prefetch support gpu

test=develop

paddle/fluid/operators/distributed/parameter_prefetch.cc

@@ -59,14 +59,13 @@ static std::vector<int64_t> ToAbsoluteSection(
 }
 
 static std::vector<std::vector<int64_t>> SplitIds(
-    const std::string& id_name, const std::vector<int>& height_section,
-    framework::Scope* scope) {
-  auto& id_tensor = scope->FindVar(id_name)->Get<framework::LoDTensor>();
-  auto* id_data = id_tensor.data<int64_t>();
+    const std::vector<int64_t>& ids_vector,
+    const std::vector<int>& height_section, framework::Scope* scope) {
   std::set<int64_t> all_ids;
-  for (size_t i = 0; i < id_tensor.numel(); ++i) {
-    all_ids.insert(id_data[i]);
+  for (auto id : ids_vector) {
+    all_ids.insert(id);
   }
+
   auto abs_sections = ToAbsoluteSection(height_section);
   std::vector<std::vector<int64_t>> splited_ids;
   splited_ids.resize(height_section.size() + 1);
@@ -78,7 +77,7 @@ static std::vector<std::vector<int64_t>> SplitIds(
 }
 
 static void SplitIdsIntoMultipleVarsBySection(
-    const std::string& id_name, const std::vector<std::string>& in_var_names,
+    const std::vector<std::string>& in_var_names,
     const std::vector<int>& height_section,
     const std::vector<std::vector<int64_t>>& splited_ids,
     framework::Scope* scope) {
@@ -99,8 +98,8 @@ static void SplitIdsIntoMultipleVarsBySection(
 }
 
 static void MergeMultipleVarsIntoOneBySection(
-    const std::string& id_name, const std::string& out_name,
-    const std::vector<std::string>& out_var_names,
+    const std::string& id_name, const std::vector<int64_t>& ids_vector,
+    const std::string& out_name, const std::vector<std::string>& out_var_names,
     const std::vector<int>& height_section,
     const std::vector<std::vector<int64_t>>& splited_ids,
     const framework::ExecutionContext& context, framework::Scope* scope) {
@@ -109,16 +108,20 @@ static void MergeMultipleVarsIntoOneBySection(
   auto cpu_place = platform::CPUPlace();
 
   auto abs_sections = ToAbsoluteSection(height_section);
-  auto& id_tensor = scope->FindVar(id_name)->Get<framework::LoDTensor>();
-  auto* id_data = id_tensor.data<int64_t>();
   std::unordered_map<int64_t, std::vector<size_t>> id_to_offset;
-  for (size_t i = 0; i < id_tensor.numel(); ++i) {
-    id_to_offset[id_data[i]].push_back(i);
+  for (size_t i = 0; i < ids_vector.size(); ++i) {
+    id_to_offset[ids_vector[i]].push_back(i);
   }
 
+  auto& id_tensor = scope->FindVar(id_name)->Get<framework::LoDTensor>();
   auto* out_tensor =
       scope->FindVar(out_name)->GetMutable<framework::LoDTensor>();
-  auto* out_tensor_data = out_tensor->mutable_data<float>(context.GetPlace());
+  auto* out_tensor_data = out_tensor->mutable_data<float>(id_tensor.place());
+
+  bool is_on_cpu_place = true;
+  if (!platform::is_cpu_place(id_tensor.place())) {
+    is_on_cpu_place = false;
+  }
 
   for (size_t section_idx = 0; section_idx < out_var_names.size();
        ++section_idx) {
@@ -140,9 +143,20 @@ static void MergeMultipleVarsIntoOneBySection(
         auto& offsets = id_to_offset[origin_id];
         for (auto& offset : offsets) {
           // should support GPU tensor
+          if (is_on_cpu_place) {
             memory::Copy(cpu_place, out_tensor_data + offset * row_numel,
                          cpu_place, out_var_data + i * row_numel,
                          sizeof(float) * row_numel);
+          } else {
+#ifndef PADDLE_WITH_CUDA
+            PADDLE_THROW("paddle is not compiled with CUDA!");
+#else
+            memory::Copy(boost::get<platform::CUDAPlace>(id_tensor.place()),
+                         out_tensor_data + offset * row_numel, cpu_place,
+                         out_var_data + i * row_numel,
+                         sizeof(float) * row_numel);
+#endif
+          }
         }
       }
     } else {
@@ -159,7 +173,7 @@ void prefetch(const std::string& id_name, const std::string& out_name,
   auto& local_scope = context.scope().NewScope();
 
   platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
-  auto& ctx = *pool.Get(context.GetPlace());
+  auto& cpu_ctx = *pool.Get(platform::CPUPlace());
 
   distributed::RPCClient* rpc_client =
       distributed::RPCClient::GetInstance<RPCCLIENT_T>(
@@ -172,9 +186,34 @@ void prefetch(const std::string& id_name, const std::string& out_name,
     out_var_names.push_back(out_name + "@" + epmap[i]);
   }
 
-  auto splited_ids = SplitIds(id_name, height_sections, &local_scope);
-  SplitIdsIntoMultipleVarsBySection(id_name, in_var_names, height_sections,
-                                    splited_ids, &local_scope);
+  auto& id_tensor = local_scope.FindVar(id_name)->Get<framework::LoDTensor>();
+  std::vector<int64_t> ids_vector;
+  if (platform::is_cpu_place(id_tensor.place())) {
+    auto* id_data = id_tensor.data<int64_t>();
+    for (size_t i = 0; i < id_tensor.numel(); ++i) {
+      ids_vector.push_back(id_data[i]);
+    }
+  } else {
+#ifndef PADDLE_WITH_CUDA
+    PADDLE_THROW("paddle is not compiled with CUDA!");
+#else
+    auto cpu_place = platform::CPUPlace();
+    framework::Tensor cpu_tensor;
+    auto* cpu_tensor_data =
+        cpu_tensor.mutable_data<int64_t>(id_tensor.dims(), cpu_place);
+    memory::Copy(cpu_place, cpu_tensor_data,
+                 boost::get<platform::CUDAPlace>(id_tensor.place()),
+                 id_tensor.data<int64_t>(),
+                 sizeof(int64_t) * id_tensor.numel());
+    for (size_t i = 0; i < id_tensor.numel(); ++i) {
+      ids_vector.push_back(cpu_tensor_data[i]);
+    }
+#endif
+  }
+
+  auto splited_ids = SplitIds(ids_vector, height_sections, &local_scope);
+  SplitIdsIntoMultipleVarsBySection(in_var_names, height_sections, splited_ids,
+                                    &local_scope);
 
   // create output var in local scope
   for (auto& name : out_var_names) {
@@ -187,7 +226,7 @@ void prefetch(const std::string& id_name, const std::string& out_name,
       VLOG(3) << "sending " << in_var_names[i] << " to " << epmap[i]
               << " to get " << out_var_names[i] << " back";
       rets.push_back(rpc_client->AsyncPrefetchVar(
-          epmap[i], ctx, local_scope, in_var_names[i], out_var_names[i],
+          epmap[i], cpu_ctx, local_scope, in_var_names[i], out_var_names[i],
           table_names[i]));
     } else {
       VLOG(3) << "don't send no-initialied variable: " << out_var_names[i];
@@ -198,9 +237,9 @@ void prefetch(const std::string& id_name, const std::string& out_name,
     PADDLE_ENFORCE(rets[i]->Wait(), "internal error in RPCClient");
   }
 
-  MergeMultipleVarsIntoOneBySection(id_name, out_name, out_var_names,
-                                    height_sections, splited_ids, context,
-                                    &local_scope);
+  MergeMultipleVarsIntoOneBySection(id_name, ids_vector, out_name,
+                                    out_var_names, height_sections, splited_ids,
+                                    context, &local_scope);
 
   context.scope().DeleteScope(&local_scope);
 }

python/paddle/fluid/tests/unittests/test_lookup_remote_table_op.py

@@ -184,8 +184,8 @@ class TestListenAndServOp(unittest.TestCase):
         port1 = self._get_pserver_port(p1.pid)
 
         places = [core.CPUPlace()]
-        # if core.is_compiled_with_cuda():
-        #     places.append(core.CUDAPlace(0))
+        if core.is_compiled_with_cuda():
+            places.append(core.CUDAPlace(0))
 
         for place in places:
             self._run_lookup_table_op_one_pserver(place, port0)