use funcs/cub.h [fluid_ops] (#76674)

Author: co63oc

paddle/phi/kernels/fusion/gpu/mmha_util.cu.h

@@ -59,17 +59,14 @@
 #include <hip/hip_bf16.h>
 #include <hip/hip_fp16.h>
 #include <hip/hip_runtime.h>
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
 #define __nv_bfloat16 __hip_bfloat16
 #define __nv_bfloat162 __hip_bfloat162
 #else
 #include <cuda_fp16.h>
 #include <float.h>
-#include <cub/cub.cuh>
 #endif
-
 #include "paddle/phi/common/datatype_traits.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 
 #ifdef PADDLE_WITH_HIP

paddle/phi/kernels/gpu/affine_channel_grad_kernel.cu

@@ -12,18 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
+#include "paddle/phi/kernels/affine_channel_grad_kernel.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_primitives.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/affine_channel_grad_kernel.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 
 namespace phi {
 

paddle/phi/kernels/gpu/affine_channel_kernel.cu

@@ -12,18 +12,11 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
+#include "paddle/phi/kernels/affine_channel_kernel.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_primitives.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/affine_channel_kernel.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 
 namespace phi {
 

paddle/phi/kernels/gpu/arg_min_max_kernel.cu

@@ -19,17 +19,10 @@
 
 #if defined(__NVCC__) || defined(__HIPCC__)
 
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
 #include <limits>
-
 #include "paddle/common/ddim.h"
 #include "paddle/phi/core/utils/data_type.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 namespace phi {
 

paddle/phi/kernels/gpu/argsort_grad_kernel.cu

@@ -17,17 +17,10 @@
 #include <thrust/sequence.h>
 #include <thrust/sort.h>
 
-#include "paddle/phi/kernels/argsort_kernel.h"
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
-
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/argsort_kernel.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/primitive/functor_primitives.h"
 #include "paddle/phi/kernels/transpose_kernel.h"

paddle/phi/kernels/gpu/argsort_kernel.cu

@@ -18,19 +18,12 @@
 #include <thrust/execution_policy.h>
 #include <thrust/sequence.h>
 #include <thrust/sort.h>
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
-
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_info.h"
 #include "paddle/phi/backends/gpu/gpu_launch_config.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/funcs/blas/blas.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/primitive/functor_primitives.h"
 #include "paddle/phi/kernels/transpose_kernel.h"

paddle/phi/kernels/gpu/batch_norm_kernel.cu

@@ -12,25 +12,17 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
-
+#include "paddle/phi/kernels/batch_norm_kernel.h"
 #include "glog/logging.h"
-
 #include "paddle/common/flags.h"
 #include "paddle/common/layout.h"
 #include "paddle/phi/backends/gpu/gpu_context.h"
 #include "paddle/phi/backends/gpu/gpu_dnn.h"
 #include "paddle/phi/core/enforce.h"
 #include "paddle/phi/core/kernel_registry.h"
-#include "paddle/phi/kernels/batch_norm_kernel.h"
 #include "paddle/phi/kernels/full_kernel.h"
 #include "paddle/phi/kernels/funcs/batch_norm_utils.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 #include "paddle/phi/kernels/funcs/eigen/common.h"
 #include "paddle/phi/kernels/funcs/norm_utils.cu.h"
 #include "paddle/phi/kernels/funcs/norm_utils.h"

paddle/phi/kernels/gpu/class_center_sample_kernel.cu

@@ -15,24 +15,19 @@
 #ifdef PADDLE_WITH_HIP
 #include <hiprand.h>
 #include <hiprand_kernel.h>
-
-#include <hipcub/hipcub.hpp>
 typedef hiprandState curandState;
-namespace cub = hipcub;
 #else
 #include <curand.h>
 #include <curand_kernel.h>
-
-#include <cub/cub.cuh>
 #endif
 
 #include <iterator>
 #include <random>
-
 #include "paddle/phi/common/memory_utils.h"
 #include "paddle/phi/core/enforce.h"
 #include "paddle/phi/core/tensor_utils.h"
 #include "paddle/phi/kernels/class_center_sample_kernel.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 
 #if defined(PADDLE_WITH_NCCL) || defined(PADDLE_WITH_RCCL)
 #include "paddle/common/flags.h"

paddle/phi/kernels/gpu/collect_fpn_proposals_kernel.cu

@@ -12,25 +12,18 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
-
+#include "paddle/phi/kernels/gpu/collect_fpn_proposals_kernel.h"
 #include "paddle/phi/backends/gpu/gpu_primitives.h"
 #include "paddle/phi/common/memory_utils.h"
 #include "paddle/phi/core/allocator.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/mixed_vector.h"
 #include "paddle/phi/kernels/funcs/concat_and_split_functor.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 #include "paddle/phi/kernels/funcs/detection/bbox_util.h"
 #include "paddle/phi/kernels/funcs/for_range.h"
 #include "paddle/phi/kernels/funcs/gather.cu.h"
 #include "paddle/phi/kernels/funcs/strided_memcpy.h"
-#include "paddle/phi/kernels/gpu/collect_fpn_proposals_kernel.h"
 #include "paddle/phi/kernels/impl/collect_fpn_proposals_kernel_impl.h"
 #include "paddle/utils/optional.h"
 

paddle/phi/kernels/gpu/cross_entropy_bwd_w_downcast.cu

@@ -12,23 +12,15 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
 
-#include "paddle/phi/kernels/cross_entropy_grad_kernel.h"
-
-#ifdef __NVCC__
-#include "cub/cub.cuh"
-#endif
-#ifdef __HIPCC__
-#include <hipcub/hipcub.hpp>
-namespace cub = hipcub;
-#endif
-
 #include "paddle/phi/backends/gpu/gpu_device_function.h"
 #include "paddle/phi/backends/gpu/gpu_dnn.h"
 #include "paddle/phi/common/amp_type_traits.h"
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/core/tensor_utils.h"
 #include "paddle/phi/core/visit_type.h"
+#include "paddle/phi/kernels/cross_entropy_grad_kernel.h"
 #include "paddle/phi/kernels/funcs/axis_utils.h"
+#include "paddle/phi/kernels/funcs/cub.h"
 #include "paddle/phi/kernels/funcs/for_range.h"
 #include "paddle/phi/kernels/funcs/math_function.h"
 #include "paddle/phi/kernels/funcs/softmax.h"