first commit (#38143)

paddle/fluid/operators/fused/attn_bias_add.cu.h

@@ -34,7 +34,7 @@ namespace cub = hipcub;
 #include "paddle/fluid/operators/elementwise/elementwise_op_broadcast.cu.h"
 #include "paddle/fluid/operators/kernel_primitives/kernel_primitives.h"
 #include "paddle/fluid/operators/reduce_ops/reduce_op.cu.h"
-#include "paddle/fluid/platform/fast_divmod.h"
+#include "paddle/phi/kernels/funcs/fast_divmod.h"
 
 namespace paddle {
 namespace operators {

paddle/fluid/operators/optimizers/lars_momentum_op.cu

@@ -15,7 +15,7 @@ limitations under the License. */
 #include "paddle/fluid/operators/optimizers/lars_momentum_op.h"
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/operators/amp/fp16_type_traits.h"
-#include "paddle/fluid/platform/fast_divmod.h"
+#include "paddle/phi/kernels/funcs/aligned_vector.h"
 #include "paddle/phi/kernels/funcs/math_cuda_utils.h"
 
 #if CUDA_VERSION >= 11000

paddle/fluid/platform/fast_divmod.h → paddle/phi/kernels/funcs/fast_divmod.h

@@ -23,8 +23,8 @@ limitations under the License. */
 #define __forceinline__ __inline__
 #endif
 
-namespace paddle {
-namespace platform {
+namespace phi {
+namespace funcs {
 
 struct FastDivMod {
   // 1st value represents the result of input number divides by recorded divisor
@@ -64,5 +64,5 @@ struct FastDivMod {
   uint32_t multiplier;
 };
 
-}  // namespace platform
-}  // namespace paddle
+}  // namespace funcs
+}  // namespace phi