diff --git a/lite/api/CMakeLists.txt b/lite/api/CMakeLists.txt
index 84f8a09860edea1bd0f29a0e7a726b1011a02ffd..874b5bb34dc12c1d979bc2e662f506c22350a776 100644
--- a/lite/api/CMakeLists.txt
+++ b/lite/api/CMakeLists.txt
@@ -84,6 +84,7 @@ if (NOT LITE_ON_TINY_PUBLISH)
                         SRCS cxx_api.cc
                         DEPS ${cxx_api_deps} ${ops} ${host_kernels} program
                         X86_DEPS ${x86_kernels}
+                        CUDA_DEPS ${cuda_kernels}
                         ARM_DEPS ${arm_kernels}
                         CV_DEPS paddle_cv_arm
                         NPU_DEPS ${npu_kernels}
@@ -120,6 +121,7 @@ if(WITH_TESTING)
        DEPS cxx_api mir_passes lite_api_test_helper
        ${ops} ${host_kernels}
        X86_DEPS ${x86_kernels}
+       CUDA_DEPS ${cuda_kernels}
        ARM_DEPS ${arm_kernels}
        NPU_DEPS ${npu_kernels}
        XPU_DEPS ${xpu_kernels}
@@ -305,7 +307,6 @@ if(NOT IOS)
         FPGA_DEPS ${fpga_kernels}
         X86_DEPS ${x86_kernels}
         CUDA_DEPS ${cuda_kernels})
-
     lite_cc_binary(benchmark_bin SRCS benchmark.cc DEPS paddle_api_full paddle_api_light gflags utils
         ${ops} ${host_kernels}
         ARM_DEPS ${arm_kernels}
diff --git a/lite/core/arena/CMakeLists.txt b/lite/core/arena/CMakeLists.txt
index 6c0c917a3e6b18f926a5fa768131e36296301432..1c85353d5386fea1ae7f4a0f1869a95f8a2478af 100644
--- a/lite/core/arena/CMakeLists.txt
+++ b/lite/core/arena/CMakeLists.txt
@@ -6,5 +6,5 @@ endif()
 lite_cc_library(arena_framework SRCS framework.cc DEPS program gtest)
 
 if((NOT LITE_WITH_OPENCL) AND (LITE_WITH_X86 OR LITE_WITH_ARM))
-  lite_cc_test(test_arena_framework SRCS framework_test.cc DEPS arena_framework ${npu_kernels} ${xpu_kernels} ${x86_kernels} ${fpga_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+  lite_cc_test(test_arena_framework SRCS framework_test.cc DEPS arena_framework ${npu_kernels} ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${fpga_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
 endif()
diff --git a/lite/tests/kernels/CMakeLists.txt b/lite/tests/kernels/CMakeLists.txt
index 69d22dc8258324c9bb23c7ca694d4efd578b0cff..7eda7acd69b6bfdc6be92857f23adb35c71f0487 100644
--- a/lite/tests/kernels/CMakeLists.txt
+++ b/lite/tests/kernels/CMakeLists.txt
@@ -1,69 +1,69 @@
 if((NOT LITE_WITH_OPENCL AND NOT LITE_WITH_FPGA) AND (LITE_WITH_X86 OR LITE_WITH_ARM))
-    lite_cc_test(test_kernel_scale_compute SRCS scale_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_power_compute SRCS power_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_shuffle_channel_compute SRCS shuffle_channel_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_yolo_box_compute SRCS yolo_box_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_fc_compute SRCS fc_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_elementwise_compute SRCS elementwise_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_lrn_compute SRCS lrn_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_decode_bboxes_compute SRCS decode_bboxes_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_box_coder_compute SRCS box_coder_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_activation_compute SRCS activation_compute_test.cc DEPS arena_framework ${npu_kernels} ${xpu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_argmax_compute SRCS argmax_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_axpy_compute SRCS axpy_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_conv2d_transpose_compute SRCS conv2d_transpose_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_norm_compute SRCS norm_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_cast_compute SRCS cast_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_instance_norm_compute SRCS instance_norm_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_grid_sampler_compute SRCS grid_sampler_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_sequence_softmax_compute SRCS sequence_softmax_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_im2sequence_compute SRCS im2sequence_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_compare_compute SRCS compare_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_logical_xor_compute SRCS logical_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_topk_compute SRCS topk_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_increment_compute SRCS increment_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_write_to_array_compute SRCS write_to_array_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_read_from_array_compute SRCS read_from_array_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_concat_compute SRCS concat_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_transpose_compute SRCS transpose_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_reshape_compute SRCS reshape_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_layer_norm_compute SRCS layer_norm_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_dropout_compute SRCS dropout_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_softmax_compute SRCS softmax_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_mul_compute SRCS mul_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_batch_norm_compute SRCS batch_norm_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_scale_compute SRCS scale_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_power_compute SRCS power_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_shuffle_channel_compute SRCS shuffle_channel_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_yolo_box_compute SRCS yolo_box_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_fc_compute SRCS fc_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_elementwise_compute SRCS elementwise_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_lrn_compute SRCS lrn_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_decode_bboxes_compute SRCS decode_bboxes_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_box_coder_compute SRCS box_coder_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_activation_compute SRCS activation_compute_test.cc DEPS arena_framework ${npu_kernels} ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_argmax_compute SRCS argmax_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_axpy_compute SRCS axpy_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_conv2d_transpose_compute SRCS conv2d_transpose_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_norm_compute SRCS norm_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_cast_compute SRCS cast_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_instance_norm_compute SRCS instance_norm_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_grid_sampler_compute SRCS grid_sampler_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_sequence_softmax_compute SRCS sequence_softmax_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_im2sequence_compute SRCS im2sequence_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_compare_compute SRCS compare_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_logical_xor_compute SRCS logical_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_topk_compute SRCS topk_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_increment_compute SRCS increment_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_write_to_array_compute SRCS write_to_array_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_read_from_array_compute SRCS read_from_array_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_concat_compute SRCS concat_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_transpose_compute SRCS transpose_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_reshape_compute SRCS reshape_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_layer_norm_compute SRCS layer_norm_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_dropout_compute SRCS dropout_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_softmax_compute SRCS softmax_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_mul_compute SRCS mul_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_batch_norm_compute SRCS batch_norm_compute_test.cc DEPS arena_framework ${xpu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
 
 if(LITE_BUILD_EXTRA)
-    lite_cc_test(test_gru_unit SRCS gru_unit_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_sequence_pool_compute SRCS sequence_pool_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_reduce_max_compute SRCS reduce_max_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_unsqueeze_compute SRCS unsqueeze_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_assign_compute SRCS assign_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_assign_value_compute SRCS assign_value_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_box_clip_compute SRCS box_clip_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_reduce_mean_compute SRCS reduce_mean_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_reduce_prod_compute SRCS reduce_prod_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_stack_compute SRCS stack_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_range_compute SRCS range_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_affine_channel_compute SRCS affine_channel_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_anchor_generator_compute SRCS anchor_generator_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_generate_proposals_compute SRCS generate_proposals_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    #lite_cc_test(test_kernel_roi_align_compute SRCS roi_align_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_search_aligned_mat_mul_compute SRCS search_aligned_mat_mul_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_search_seq_fc_compute SRCS search_seq_fc_compute_test.cc DEPS arena_framework ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_lookup_table_compute SRCS lookup_table_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_gather_compute SRCS gather_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_gru_unit SRCS gru_unit_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_sequence_pool_compute SRCS sequence_pool_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_reduce_max_compute SRCS reduce_max_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_unsqueeze_compute SRCS unsqueeze_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_assign_compute SRCS assign_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_assign_value_compute SRCS assign_value_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_box_clip_compute SRCS box_clip_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_reduce_mean_compute SRCS reduce_mean_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_reduce_prod_compute SRCS reduce_prod_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_stack_compute SRCS stack_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_range_compute SRCS range_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_affine_channel_compute SRCS affine_channel_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_anchor_generator_compute SRCS anchor_generator_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_generate_proposals_compute SRCS generate_proposals_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    #lite_cc_test(test_kernel_roi_align_compute SRCS roi_align_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_search_aligned_mat_mul_compute SRCS search_aligned_mat_mul_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_search_seq_fc_compute SRCS search_seq_fc_compute_test.cc DEPS arena_framework ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_lookup_table_compute SRCS lookup_table_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_gather_compute SRCS gather_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
 endif()
-    lite_cc_test(test_kernel_pad2d_compute SRCS  pad2d_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_prior_box_compute SRCS  prior_box_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_negative_compute SRCS negative_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_bilinear_interp_compute SRCS bilinear_interp_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_nearest_interp_compute SRCS nearest_interp_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_shape_compute SRCS shape_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_crop_compute SRCS crop_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_sequence_expand_compute SRCS sequence_expand_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_squeeze_compute SRCS squeeze_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_slice_compute SRCS slice_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_expand_compute SRCS expand_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
-    lite_cc_test(test_kernel_matmul_compute SRCS matmul_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_pad2d_compute SRCS  pad2d_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_prior_box_compute SRCS  prior_box_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_negative_compute SRCS negative_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_bilinear_interp_compute SRCS bilinear_interp_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_nearest_interp_compute SRCS nearest_interp_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_shape_compute SRCS shape_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_crop_compute SRCS crop_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_sequence_expand_compute SRCS sequence_expand_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_squeeze_compute SRCS squeeze_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_slice_compute SRCS slice_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_expand_compute SRCS expand_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
+    lite_cc_test(test_kernel_matmul_compute SRCS matmul_compute_test.cc DEPS arena_framework ${xpu_kernels} ${npu_kernels} ${x86_kernels} ${cuda_kernels} ${arm_kernels} ${lite_ops} ${host_kernels})
 endif()