Fix potential mkldnn concat/pool/conv kernel issues (#18393)

1. some key generation method is not aligned with PR#17965
2. enlarge ptr lifetime to avoid memory release if SetBlob fails
   otherwise it will get core dump.

test=develop

paddle/fluid/operators/mkldnn/concat_mkldnn_op.cc

@@ -81,6 +81,13 @@ std::string CreateKey(const paddle::framework::ExecutionContext& ctx,
   platform::MKLDNNHandler::AppendKey(&key, std::to_string(dt));
   platform::MKLDNNHandler::AppendKey(&key,
                                      std::to_string(multi_input[0]->format()));
+  if (platform::get_cur_thread_id() != -1) {
+    auto tid = std::this_thread::get_id();
+    std::stringstream ss;
+    ss << tid;
+    platform::MKLDNNHandler::AppendKey(&key, "-t:");
+    platform::MKLDNNHandler::AppendKey(&key, ss.str());
+  }
   return key;
 }
 

paddle/fluid/operators/mkldnn/conv_mkldnn_op.cc

@@ -220,7 +220,7 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
     auto weights_memory_p = handler.AcquireWeightsMemoryFromPrimitive(
         user_weights_memory_p, pipeline, is_test);
 
-    std::shared_ptr<mkldnn::memory> dst_memory_p;
+    std::shared_ptr<mkldnn::memory> dst_memory_p, user_residual_memory_p;
 
     if (fuse_residual_conn) {
       auto residual_param = ctx.Input<Tensor>("ResidualData");
@@ -243,7 +243,7 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
 
         auto user_residual_md = platform::MKLDNNMemDesc(
             residual_data_tz, residual_data_type, residual_param->format());
-        auto user_residual_memory_p = handler.AcquireResidualDataMemory(
+        user_residual_memory_p = handler.AcquireResidualDataMemory(
             user_residual_md, to_void_cast<T>(residual_param_data));
 
         dst_memory_p = handler.AcquireDstMemoryFromResidualDataMemory(
@@ -263,14 +263,15 @@ class ConvMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
 
     // create convolution op primitive
     std::shared_ptr<mkldnn::convolution_forward> conv_p;
+    std::shared_ptr<mkldnn::memory> user_bias_memory_p, bias_memory_p;
     if (bias) {
       const T* bias_data = bias->data<T>();
       auto user_bias_md = platform::MKLDNNMemDesc(
           {bias_tz}, platform::MKLDNNGetDataType<T>(), memory::format::x);
-      auto user_bias_memory_p =
+      user_bias_memory_p =
           handler.AcquireBiasMemory(user_bias_md, to_void_cast<T>(bias_data));
 
-      auto bias_memory_p =
+      bias_memory_p =
           handler.AcquireBiasMemoryFromPrimitive(user_bias_memory_p, pipeline);
       conv_p = handler.AcquireConvolution(src_memory_p, weights_memory_p,
                                           bias_memory_p, dst_memory_p);

paddle/fluid/operators/mkldnn/pool_mkldnn_op.cc

@@ -48,6 +48,13 @@ std::string CreateKey(const paddle::framework::ExecutionContext& ctx,
   platform::MKLDNNHandler::AppendKey(&key, std::to_string(dt));
   platform::MKLDNNHandler::AppendKey(&key, std::to_string(fmt));
   platform::MKLDNNHandler::AppendKey(&key, suffix);
+  if (platform::get_cur_thread_id() != -1) {
+    auto tid = std::this_thread::get_id();
+    std::stringstream ss;
+    ss << tid;
+    platform::MKLDNNHandler::AppendKey(&key, "-t:");
+    platform::MKLDNNHandler::AppendKey(&key, ss.str());
+  }
   return key;
 }
 
@@ -128,6 +135,10 @@ class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
     const std::string key_pool_workspace_memory =
         key + "@pool_workspace_memory";
 
+    std::shared_ptr<mkldnn::memory> src_memory, dst_memory;
+    std::shared_ptr<mkldnn::pooling_forward::primitive_desc> pool_pd;
+    std::shared_ptr<mkldnn::memory> pool_src_memory_p, pool_dst_memory_p;
+
     auto pool_p =
         std::static_pointer_cast<pooling_forward>(dev_ctx.GetBlob(key_pool_p));
     if (pool_p == nullptr) {
@@ -158,9 +169,9 @@ class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
       // save pool_pd into global device context to be referred in backward path
       if (!is_test) dev_ctx.SetBlob(key_pool_pd, pool_pd);
 
-      auto src_memory = std::make_shared<memory>(pool_pd->src_primitive_desc(),
+      src_memory = std::make_shared<memory>(pool_pd->src_primitive_desc(),
                                             to_void_cast<T>(input_data));
-      auto dst_memory =
+      dst_memory =
           std::make_shared<memory>(pool_pd->dst_primitive_desc(), output_data);
 
       dev_ctx.SetBlob(key_pool_src_mem_p, src_memory);
@@ -186,11 +197,11 @@ class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
           (memory::format)dst_memory->get_primitive_desc().desc().data.format;
     } else {
       // Primitives already exist
-      auto pool_src_memory_p =
+      pool_src_memory_p =
           std::static_pointer_cast<memory>(dev_ctx.GetBlob(key_pool_src_mem_p));
       PADDLE_ENFORCE(pool_src_memory_p != nullptr,
                      "Fail to find pooling src mem_p in device context");
-      auto pool_dst_memory_p =
+      pool_dst_memory_p =
           std::static_pointer_cast<memory>(dev_ctx.GetBlob(key_pool_dst_mem_p));
       PADDLE_ENFORCE(pool_dst_memory_p != nullptr,
                      "Fail to find pooling dst mem_p in device context");

paddle/fluid/platform/mkldnn_reuse.h

@@ -38,6 +38,9 @@ class MKLDNNHandler {
     std::stringstream ss;
     ss << tid;
     key_ = key_common_ + "-t:" + ss.str();
+    if (platform::get_cur_thread_id() == -1) {
+      key_ = key_common_;
+    }
   }
 
   std::shared_ptr<mkldnn::memory> AcquireSrcMemory(