gpu codex warning fix

mindspore/ccsrc/kernel/gpu/cuda_impl/layer_norm_impl.cu

@@ -96,7 +96,7 @@ inline __device__ void BlockReduce(const int& col_dim, T* mean, T* var, T* num,
   __syncthreads();
 
   if (threadIdx.x == 0) {
-    mean_addr[blockIdx.x] = share_mem[0];  // todo: blockDim.x < row
+    mean_addr[blockIdx.x] = share_mem[0];
     share_mem[1] /= col_dim;
     var_addr[blockIdx.x] = share_mem[1];
   }

mindspore/ccsrc/kernel/gpu/data/dataset_iterator_kernel.cc

@@ -96,7 +96,8 @@ bool DatasetIteratorKernel::Launch(const std::vector<AddressPtr> &, const std::v
   }
 
   for (size_t i = 0; i < output_size_list_.size(); i++) {
-    CHECK_CUDA_RET_WITH_EXCEPT(cudaMemcpyAsync(outputs[i]->addr, addr, output_size_list_[i], cudaMemcpyDeviceToDevice,
+    void *output_addr = GetDeviceAddress<void>(outputs, i);
+    CHECK_CUDA_RET_WITH_EXCEPT(cudaMemcpyAsync(output_addr, addr, output_size_list_[i], cudaMemcpyDeviceToDevice,
                                                reinterpret_cast<cudaStream_t>(stream)),
                                "Cuda Memcpy Failed");
     addr = reinterpret_cast<unsigned char *>(addr) + output_size_list_[i];

mindspore/ccsrc/kernel/gpu/math/broadcast_gpu_kernel.h

@@ -68,14 +68,14 @@ class BroadcastOpGpuKernel : public GpuKernel {
       output_shape_[i] = shape3[i];
       output_num_ *= shape3[i];
     }
-    int offset = shape3.size() - shape1.size();
+    int lhs_offset = shape3.size() - shape1.size();
     for (size_t j = 0; j < shape1.size(); j++) {
-      lhs_shape_[j + offset] = shape1[j];
+      lhs_shape_[j + lhs_offset] = shape1[j];
       input1_num_ *= shape1[j];
     }
-    offset = shape3.size() - shape2.size();
+    int rhs_offset = shape3.size() - shape2.size();
     for (size_t k = 0; k < shape2.size(); k++) {
-      rhs_shape_[k + offset] = shape2[k];
+      rhs_shape_[k + rhs_offset] = shape2[k];
       input2_num_ *= shape2[k];
     }
 

mindspore/ccsrc/kernel/gpu/math/broadcast_grad_gpu_kernel.h

@@ -74,14 +74,14 @@ class BroadcastOpGradGpuKernel : public GpuKernel {
       dy_shape_[i] = shape3[i];
       output_num_ *= shape3[i];
     }
-    int offset = shape3.size() - shape1.size();
+    int x1_offset = shape3.size() - shape1.size();
     for (size_t i = 0; i < shape1.size(); i++) {
-      x1_shape_[i + offset] = shape1[i];
+      x1_shape_[i + x1_offset] = shape1[i];
       input1_num_ *= shape1[i];
     }
-    offset = shape3.size() - shape2.size();
+    int x2_offset = shape3.size() - shape2.size();
     for (size_t i = 0; i < shape2.size(); i++) {
-      x2_shape_[i + offset] = shape2[i];
+      x2_shape_[i + x2_offset] = shape2[i];
       input2_num_ *= shape2[i];
     }
 

mindspore/ccsrc/kernel/gpu/nn/dropout_gpu_kernel.cc

@@ -68,14 +68,12 @@ void DropoutGpuFwdKernel::DestroyResource() noexcept {}
 
 void DropoutGpuFwdKernel::InitSizeLists() {
   size_t input_size = num_count_ * sizeof(float);
-  size_t workspace_size = 0;
   input_size_list_.push_back(input_size);
   output_size_list_.push_back(input_size);  // output size: the same with input size
   output_size_list_.push_back(input_size);  // mask size: the same with input size
-  workspace_size_list_.push_back(workspace_size);
 }
 
-bool DropoutGpuFwdKernel::Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+bool DropoutGpuFwdKernel::Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &,
                                  const std::vector<AddressPtr> &outputs, void *stream_ptr) {
   if (is_null_input_) {
     return true;

mindspore/ccsrc/kernel/gpu/nn/dropout_grad_kernel.cc

@@ -66,15 +66,13 @@ void DropoutGradGpuFwdKernel::InitSizeLists() {
   size_t dy_size = num_count_ * sizeof(float);
   size_t mask_size = dy_size;
   size_t dx_size = dy_size;
-  size_t workspace_size = 0;
 
   input_size_list_.push_back(dy_size);
   input_size_list_.push_back(mask_size);
   output_size_list_.push_back(dx_size);
-  workspace_size_list_.push_back(workspace_size);
 }
 
-bool DropoutGradGpuFwdKernel::Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &workspace,
+bool DropoutGradGpuFwdKernel::Launch(const std::vector<AddressPtr> &inputs, const std::vector<AddressPtr> &,
                                      const std::vector<AddressPtr> &outputs, void *stream_ptr) {
   if (is_null_input_) {
     return true;