remove duplicate cmake targets of sequence-pool (#2467)

lite/kernels/cuda/CMakeLists.txt

@@ -9,7 +9,7 @@ add_kernel(io_copy_compute_cuda CUDA basic SRCS io_copy_compute.cc DEPS ${lite_k
 add_kernel(leaky_relu_compute_cuda CUDA basic SRCS leaky_relu_compute.cu DEPS ${lite_kernel_deps})
 add_kernel(relu_compute_cuda CUDA basic SRCS relu_compute.cu DEPS ${lite_kernel_deps})
 add_kernel(yolo_box_compute_cuda CUDA basic SRCS yolo_box_compute.cu DEPS ${lite_kernel_deps})
-#add_kernel(sequence_pool_compute_cuda CUDA extra SRCS sequence_pool_compute.cu DEPS ${lite_kernel_deps})
+add_kernel(sequence_pool_compute_cuda CUDA extra SRCS sequence_pool_compute.cu DEPS ${lite_kernel_deps})
 add_kernel(transpose_compute_cuda CUDA basic SRCS transpose_compute.cu DEPS ${lite_kernel_deps} ${math_cuda} cuda_transpose)
 add_kernel(nearest_interp_compute_cuda CUDA basic SRCS nearest_interp_compute.cu DEPS ${lite_kernel_deps})
 add_kernel(conv2d_cuda CUDA basic SRCS conv_compute.cc DEPS ${lite_kernel_deps} ${math_cuda})
@@ -44,7 +44,6 @@ nv_test(yolo_box_compute_cuda_test SRCS yolo_box_compute_test.cc DEPS yolo_box_c
 nv_test(transpose_compute_cuda_test SRCS transpose_compute_test.cc DEPS transpose_compute_cuda)
 nv_test(concat_compute_cuda_test SRCS concat_compute_test.cc DEPS concat_compute_cuda)
 nv_test(elementwise_add_compute_cuda_test SRCS elementwise_add_compute_test.cc DEPS elementwise_add_compute_cuda)
-#nv_test(sequence_pool_compute_cuda_test SRCS sequence_pool_compute_test.cc DEPS sequence_pool_compute_cuda)
 nv_test(softmax_compute_cuda_test SRCS softmax_compute_test.cc DEPS softmax_compute_cuda)
 #nv_test(layout_cuda_test SRCS layout_compute_test.cc DEPS layout_compute_cuda)
 nv_test(mul_compute_cuda_test SRCS mul_compute_test.cc DEPS mul_compute_cuda)
@@ -60,7 +59,7 @@ if(LITE_BUILD_EXTRA)
     nv_test(search_seq_depadding_compute_cuda_test SRCS search_seq_depadding_compute_test.cc DEPS search_seq_depadding_compute_cuda)
     nv_test(match_matrix_tensor_compute_cuda_test SRCS match_matrix_tensor_compute_test.cc DEPS match_matrix_tensor_compute_cuda)
     nv_test(search_grnn_compute_cuda_test SRCS search_grnn_compute_test.cc DEPS search_grnn_compute_cuda)
-   #nv_test(sequence_pool_compute_cuda_test SRCS sequence_pool_compute_test.cc DEPS sequence_pool_compute_cuda sequence_pooling)
+    nv_test(sequence_pool_compute_cuda_test SRCS sequence_pool_compute_test.cc DEPS sequence_pool_compute_cuda)
     nv_test(lookup_table_compute_cuda_test SRCS lookup_table_compute_test.cc DEPS lookup_table_compute_cuda)
     nv_test(search_aligned_mat_mul_compute_cuda_test SRCS search_aligned_mat_mul_compute_test.cc DEPS search_aligned_mat_mul_compute_cuda)
     nv_test(search_seq_fc_compute_cuda_test SRCS search_seq_fc_compute_test.cc DEPS search_seq_fc_compute_cuda)

lite/kernels/cuda/sequence_pool_compute.cu

@@ -13,6 +13,7 @@
 // limitations under the License.
 
 #include <vector>
+#include "lite/backends/cuda/cuda_utils.h"
 #include "lite/core/op_registry.h"
 #include "lite/core/target_wrapper.h"
 #include "lite/kernels/cuda/sequence_pool_compute.h"
@@ -22,6 +23,10 @@ namespace lite {
 namespace kernels {
 namespace cuda {
 
+#define CUDA_KERNEL_LOOP(i, n)                                 \
+  for (int i = blockIdx.x * blockDim.x + threadIdx.x; i < (n); \
+       i += blockDim.x * gridDim.x)
+
 template <typename Dtype>
 __global__ void seq_pool_average_kernel(Dtype* dst,
                                         const Dtype* src_in,