[PTen] Rename kernel register marco (#38861)

* rename register marco

* fix error changing

* fix format error

cmake/pten_kernel.cmake

@@ -16,12 +16,12 @@
 function(kernel_declare TARGET_LIST)
     foreach(kernel_path ${TARGET_LIST})
         file(READ ${kernel_path} kernel_impl)
-        # TODO(chenweihang): rename PT_REGISTER_CTX_KERNEL to PT_REGISTER_KERNEL
+        # TODO(chenweihang): rename PT_REGISTER_KERNEL to PT_REGISTER_KERNEL
         # NOTE(chenweihang): now we don't recommend to use digit in kernel name
-        string(REGEX MATCH "(PT_REGISTER_CTX_KERNEL|PT_REGISTER_GENERAL_KERNEL)\\([ \t\r\n]*[a-z0-9_]*," first_registry "${kernel_impl}")
+        string(REGEX MATCH "(PT_REGISTER_KERNEL|PT_REGISTER_GENERAL_KERNEL)\\([ \t\r\n]*[a-z0-9_]*," first_registry "${kernel_impl}")
         if (NOT first_registry STREQUAL "")
             # parse the first kernel name
-            string(REPLACE "PT_REGISTER_CTX_KERNEL(" "" kernel_name "${first_registry}")
+            string(REPLACE "PT_REGISTER_KERNEL(" "" kernel_name "${first_registry}")
             string(REPLACE "PT_REGISTER_GENERAL_KERNEL(" "" kernel_name "${kernel_name}")
             string(REPLACE "," "" kernel_name "${kernel_name}")
             string(REGEX REPLACE "[ \t\r\n]+" "" kernel_name "${kernel_name}")

paddle/pten/kernels/cpu/cast_kernel.cc

@@ -58,7 +58,7 @@ void CastKernel(const Context& dev_ctx,
 
 }  // namespace pten
 
-PT_REGISTER_CTX_KERNEL(cast,
+PT_REGISTER_KERNEL(cast,
                    CPU,
                    ALL_LAYOUT,
                    pten::CastKernel,

paddle/pten/kernels/cpu/complex_kernel.cc

@@ -21,7 +21,7 @@
 // See Note [ Why still include the fluid headers? ]
 #include "paddle/fluid/platform/complex.h"
 
-PT_REGISTER_CTX_KERNEL(conj,
+PT_REGISTER_KERNEL(conj,
                    CPU,
                    ALL_LAYOUT,
                    pten::ConjKernel,

paddle/pten/kernels/cpu/dot_grad_kernel.cc

@@ -20,7 +20,7 @@
 
 #include "paddle/fluid/platform/complex.h"
 
-PT_REGISTER_CTX_KERNEL(dot_grad,
+PT_REGISTER_KERNEL(dot_grad,
                    CPU,
                    ALL_LAYOUT,
                    pten::DotGradKernel,

paddle/pten/kernels/cpu/dot_kernel.cc

@@ -49,7 +49,7 @@ void DotKernel(const Context& dev_ctx,
 using complex64 = ::paddle::platform::complex<float>;
 using complex128 = ::paddle::platform::complex<double>;
 
-PT_REGISTER_CTX_KERNEL(dot,
+PT_REGISTER_KERNEL(dot,
                    CPU,
                    ALL_LAYOUT,
                    pten::DotKernel,

paddle/pten/kernels/cpu/full_kernel.cc

@@ -18,7 +18,7 @@ limitations under the License. */
 #include "paddle/pten/core/kernel_registry.h"
 #include "paddle/pten/kernels/impl/full_kernel_impl.h"
 
-PT_REGISTER_CTX_KERNEL(full,
+PT_REGISTER_KERNEL(full,
                    CPU,
                    ALL_LAYOUT,
                    pten::FullKernel,
@@ -34,7 +34,7 @@ PT_REGISTER_CTX_KERNEL(full,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(full_like,
+PT_REGISTER_KERNEL(full_like,
                    CPU,
                    ALL_LAYOUT,
                    pten::FullLikeKernel,

paddle/pten/kernels/cpu/math_kernel.cc

@@ -118,9 +118,9 @@ using complex128 = ::paddle::platform::complex<double>;
 
 // NOTE(chenweihang): using bfloat16 will cause redefine with xpu bfloat16
 // using bfloat16 = ::paddle::platform::bfloat16;
-PT_REGISTER_CTX_KERNEL(
+PT_REGISTER_KERNEL(
     mean, CPU, ALL_LAYOUT, pten::MeanKernel, float, double, bool) {}
-PT_REGISTER_CTX_KERNEL(add,
+PT_REGISTER_KERNEL(add,
                    CPU,
                    ALL_LAYOUT,
                    pten::AddKernel,
@@ -130,7 +130,7 @@ PT_REGISTER_CTX_KERNEL(add,
                    int64_t,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(subtract,
+PT_REGISTER_KERNEL(subtract,
                    CPU,
                    ALL_LAYOUT,
                    pten::SubtractKernel,
@@ -140,7 +140,7 @@ PT_REGISTER_CTX_KERNEL(subtract,
                    int64_t,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(divide,
+PT_REGISTER_KERNEL(divide,
                    CPU,
                    ALL_LAYOUT,
                    pten::DivideKernel,
@@ -150,7 +150,7 @@ PT_REGISTER_CTX_KERNEL(divide,
                    int64_t,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(multiply,
+PT_REGISTER_KERNEL(multiply,
                    CPU,
                    ALL_LAYOUT,
                    pten::MultiplyKernel,
@@ -161,7 +161,7 @@ PT_REGISTER_CTX_KERNEL(multiply,
                    bool,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(sum,
+PT_REGISTER_KERNEL(sum,
                    CPU,
                    ALL_LAYOUT,
                    pten::SumKernel,

paddle/pten/kernels/cpu/matmul_grad_kernel.cc

@@ -19,7 +19,7 @@ limitations under the License. */
 
 #include "paddle/pten/kernels/impl/matmul_grad_kernel_impl.h"
 
-PT_REGISTER_CTX_KERNEL(matmul_grad,
+PT_REGISTER_KERNEL(matmul_grad,
                    CPU,
                    ALL_LAYOUT,
                    pten::MatmulGradKernel,
@@ -28,7 +28,7 @@ PT_REGISTER_CTX_KERNEL(matmul_grad,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(matmul_double_grad,
+PT_REGISTER_KERNEL(matmul_double_grad,
                    CPU,
                    ALL_LAYOUT,
                    pten::MatmulDoubleGradKernel,
@@ -37,7 +37,7 @@ PT_REGISTER_CTX_KERNEL(matmul_double_grad,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(matmul_triple_grad,
+PT_REGISTER_KERNEL(matmul_triple_grad,
                    CPU,
                    ALL_LAYOUT,
                    pten::MatmulTripleGradKernel,

paddle/pten/kernels/cpu/matmul_kernel.cc

@@ -20,7 +20,7 @@ limitations under the License. */
 #include "paddle/fluid/platform/complex.h"
 #include "paddle/pten/kernels/impl/matmul_kernel_impl.h"
 
-PT_REGISTER_CTX_KERNEL(matmul,
+PT_REGISTER_KERNEL(matmul,
                    CPU,
                    ALL_LAYOUT,
                    pten::MatmulKernel,

paddle/pten/kernels/cpu/scale_kernel.cc

@@ -51,7 +51,7 @@ void ScaleKernel(const Context& dev_ctx,
 
 }  // namespace pten
 
-PT_REGISTER_CTX_KERNEL(scale,
+PT_REGISTER_KERNEL(scale,
                    CPU,
                    ALL_LAYOUT,
                    pten::ScaleKernel,

paddle/pten/kernels/cpu/sign_kernel.cc

@@ -21,5 +21,4 @@ limitations under the License. */
 // See Note [ Why still include the fluid headers? ]
 #include "paddle/fluid/platform/bfloat16.h"
 
-PT_REGISTER_CTX_KERNEL(sign, CPU, ALL_LAYOUT, pten::SignKernel, float, double) {
-}
+PT_REGISTER_KERNEL(sign, CPU, ALL_LAYOUT, pten::SignKernel, float, double) {}

paddle/pten/kernels/empty_kernel.cc

@@ -34,7 +34,7 @@ void EmptyLikeKernel(const Context& dev_ctx, DenseTensor* out) {
 
 }  // namespace pten
 
-PT_REGISTER_CTX_KERNEL(empty,
+PT_REGISTER_KERNEL(empty,
                    CPU,
                    ALL_LAYOUT,
                    pten::EmptyKernel,
@@ -50,7 +50,7 @@ PT_REGISTER_CTX_KERNEL(empty,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(empty_like,
+PT_REGISTER_KERNEL(empty_like,
                    CPU,
                    ALL_LAYOUT,
                    pten::EmptyLikeKernel,
@@ -67,7 +67,7 @@ PT_REGISTER_CTX_KERNEL(empty_like,
                    paddle::platform::complex<double>) {}
 
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
-PT_REGISTER_CTX_KERNEL(empty,
+PT_REGISTER_KERNEL(empty,
                    GPU,
                    ALL_LAYOUT,
                    pten::EmptyKernel,
@@ -82,7 +82,7 @@ PT_REGISTER_CTX_KERNEL(empty,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(empty_like,
+PT_REGISTER_KERNEL(empty_like,
                    GPU,
                    ALL_LAYOUT,
                    pten::EmptyLikeKernel,

paddle/pten/kernels/flatten_grad_kernel.cc

@@ -33,7 +33,7 @@ void FlattenGradKernel(const Context& dev_ctx,
 
 }  // namespace pten
 
-PT_REGISTER_CTX_KERNEL(flatten_grad,
+PT_REGISTER_KERNEL(flatten_grad,
                    CPU,
                    ALL_LAYOUT,
                    pten::FlattenGradKernel,
@@ -45,7 +45,7 @@ PT_REGISTER_CTX_KERNEL(flatten_grad,
                    int64_t) {}
 
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
-PT_REGISTER_CTX_KERNEL(flatten_grad,
+PT_REGISTER_KERNEL(flatten_grad,
                    GPU,
                    ALL_LAYOUT,
                    pten::FlattenGradKernel,
@@ -60,7 +60,7 @@ PT_REGISTER_CTX_KERNEL(flatten_grad,
 #endif
 
 #ifdef PADDLE_WITH_XPU
-PT_REGISTER_CTX_KERNEL(flatten_grad,
+PT_REGISTER_KERNEL(flatten_grad,
                    XPU,
                    ALL_LAYOUT,
                    pten::FlattenGradKernel,

paddle/pten/kernels/flatten_kernel.cc

@@ -48,7 +48,7 @@ void FlattenWithXShape(const Context& dev_ctx,
 
 }  // namespace pten
 
-PT_REGISTER_CTX_KERNEL(flatten,
+PT_REGISTER_KERNEL(flatten,
                    CPU,
                    ALL_LAYOUT,
                    pten::FlattenKernel,
@@ -59,7 +59,7 @@ PT_REGISTER_CTX_KERNEL(flatten,
                    int,
                    int64_t) {}
 
-PT_REGISTER_CTX_KERNEL(flatten_with_xshape,
+PT_REGISTER_KERNEL(flatten_with_xshape,
                    CPU,
                    ALL_LAYOUT,
                    pten::FlattenWithXShape,
@@ -71,7 +71,7 @@ PT_REGISTER_CTX_KERNEL(flatten_with_xshape,
                    int64_t) {}
 
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
-PT_REGISTER_CTX_KERNEL(flatten,
+PT_REGISTER_KERNEL(flatten,
                    GPU,
                    ALL_LAYOUT,
                    pten::FlattenKernel,
@@ -83,7 +83,7 @@ PT_REGISTER_CTX_KERNEL(flatten,
                    int,
                    int64_t) {}
 
-PT_REGISTER_CTX_KERNEL(flatten_with_xshape,
+PT_REGISTER_KERNEL(flatten_with_xshape,
                    GPU,
                    ALL_LAYOUT,
                    pten::FlattenWithXShape,
@@ -97,7 +97,7 @@ PT_REGISTER_CTX_KERNEL(flatten_with_xshape,
 #endif
 
 #ifdef PADDLE_WITH_XPU
-PT_REGISTER_CTX_KERNEL(flatten,
+PT_REGISTER_KERNEL(flatten,
                    XPU,
                    ALL_LAYOUT,
                    pten::FlattenKernel,
@@ -107,7 +107,7 @@ PT_REGISTER_CTX_KERNEL(flatten,
                    int,
                    int64_t) {}
 
-PT_REGISTER_CTX_KERNEL(flatten_with_xshape,
+PT_REGISTER_KERNEL(flatten_with_xshape,
                    XPU,
                    ALL_LAYOUT,
                    pten::FlattenWithXShape,

paddle/pten/kernels/gpu/cast_kernel.cu

@@ -61,7 +61,7 @@ void CastKernel(const Context& dev_ctx,
 }  // namespace pten
 
 #define PTEN_REGISTER_CAST_CUDA_BASE_TYPE(op_name, ...) \
-  PT_REGISTER_CTX_KERNEL(cast,                              \
+  PT_REGISTER_KERNEL(cast,                              \
                      GPU,                               \
                      ALL_LAYOUT,                        \
                      pten::CastKernel,                  \

paddle/pten/kernels/gpu/complex_kernel.cu

@@ -21,7 +21,7 @@
 // See Note [ Why still include the fluid headers? ]
 #include "paddle/fluid/platform/complex.h"
 
-PT_REGISTER_CTX_KERNEL(conj,
+PT_REGISTER_KERNEL(conj,
                    GPU,
                    ALL_LAYOUT,
                    pten::ConjKernel,

paddle/pten/kernels/gpu/dot_grad_kernel.cu

@@ -20,7 +20,7 @@ limitations under the License. */
 
 #include "paddle/fluid/platform/complex.h"
 
-PT_REGISTER_CTX_KERNEL(dot_grad,
+PT_REGISTER_KERNEL(dot_grad,
                    GPU,
                    ALL_LAYOUT,
                    pten::DotGradKernel,

paddle/pten/kernels/gpu/dot_kernel.cu

@@ -52,7 +52,7 @@ void DotKernel(const Context& dev_ctx,
 using complex64 = ::paddle::platform::complex<float>;
 using complex128 = ::paddle::platform::complex<double>;
 
-PT_REGISTER_CTX_KERNEL(dot,
+PT_REGISTER_KERNEL(dot,
                    GPU,
                    ALL_LAYOUT,
                    pten::DotKernel,

paddle/pten/kernels/gpu/full_kernel.cu

@@ -18,7 +18,7 @@ limitations under the License. */
 #include "paddle/pten/core/kernel_registry.h"
 #include "paddle/pten/kernels/impl/full_kernel_impl.h"
 
-PT_REGISTER_CTX_KERNEL(full,
+PT_REGISTER_KERNEL(full,
                    GPU,
                    ALL_LAYOUT,
                    pten::FullKernel,
@@ -33,7 +33,7 @@ PT_REGISTER_CTX_KERNEL(full,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(full_like,
+PT_REGISTER_KERNEL(full_like,
                    GPU,
                    ALL_LAYOUT,
                    pten::FullLikeKernel,

paddle/pten/kernels/gpu/math_kernel.cu

@@ -110,9 +110,9 @@ using float16 = paddle::platform::float16;
 using complex64 = ::paddle::platform::complex<float>;
 using complex128 = ::paddle::platform::complex<double>;
 
-PT_REGISTER_CTX_KERNEL(
+PT_REGISTER_KERNEL(
     mean, GPU, ALL_LAYOUT, pten::MeanKernel, float, double, bool, float16) {}
-PT_REGISTER_CTX_KERNEL(add,
+PT_REGISTER_KERNEL(add,
                    GPU,
                    ALL_LAYOUT,
                    pten::AddKernel,
@@ -123,7 +123,7 @@ PT_REGISTER_CTX_KERNEL(add,
                    float16,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(subtract,
+PT_REGISTER_KERNEL(subtract,
                    GPU,
                    ALL_LAYOUT,
                    pten::SubtractKernel,
@@ -134,7 +134,7 @@ PT_REGISTER_CTX_KERNEL(subtract,
                    float16,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(divide,
+PT_REGISTER_KERNEL(divide,
                    GPU,
                    ALL_LAYOUT,
                    pten::DivideKernel,
@@ -145,7 +145,7 @@ PT_REGISTER_CTX_KERNEL(divide,
                    float16,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(multiply,
+PT_REGISTER_KERNEL(multiply,
                    GPU,
                    ALL_LAYOUT,
                    pten::MultiplyKernel,
@@ -157,7 +157,7 @@ PT_REGISTER_CTX_KERNEL(multiply,
                    float16,
                    complex64,
                    complex128) {}
-PT_REGISTER_CTX_KERNEL(sum,
+PT_REGISTER_KERNEL(sum,
                    GPU,
                    ALL_LAYOUT,
                    pten::SumKernel,

paddle/pten/kernels/gpu/matmul_grad_kernel.cu

@@ -19,7 +19,7 @@ limitations under the License. */
 
 #include "paddle/pten/kernels/impl/matmul_grad_kernel_impl.h"
 
-PT_REGISTER_CTX_KERNEL(matmul_grad,
+PT_REGISTER_KERNEL(matmul_grad,
                    GPU,
                    ALL_LAYOUT,
                    pten::MatmulGradKernel,
@@ -29,7 +29,7 @@ PT_REGISTER_CTX_KERNEL(matmul_grad,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(matmul_double_grad,
+PT_REGISTER_KERNEL(matmul_double_grad,
                    GPU,
                    ALL_LAYOUT,
                    pten::MatmulDoubleGradKernel,
@@ -39,7 +39,7 @@ PT_REGISTER_CTX_KERNEL(matmul_double_grad,
                    paddle::platform::complex<float>,
                    paddle::platform::complex<double>) {}
 
-PT_REGISTER_CTX_KERNEL(matmul_triple_grad,
+PT_REGISTER_KERNEL(matmul_triple_grad,
                    GPU,
                    ALL_LAYOUT,
                    pten::MatmulTripleGradKernel,

paddle/pten/kernels/gpu/matmul_kernel.cu

@@ -20,7 +20,7 @@ limitations under the License. */
 #include "paddle/fluid/platform/complex.h"
 #include "paddle/pten/kernels/impl/matmul_kernel_impl.h"
 
-PT_REGISTER_CTX_KERNEL(matmul,
+PT_REGISTER_KERNEL(matmul,
                    GPU,
                    ALL_LAYOUT,
                    pten::MatmulKernel,

paddle/pten/kernels/gpu/scale_kernel.cu

@@ -64,7 +64,7 @@ void ScaleKernel(const ContextT& dev_ctx,
 
 }  // namespace pten
 
-PT_REGISTER_CTX_KERNEL(scale,
+PT_REGISTER_KERNEL(scale,
                    GPU,
                    ALL_LAYOUT,
                    pten::ScaleKernel,

paddle/pten/kernels/gpu/sign_kernel.cu

@@ -23,5 +23,5 @@ limitations under the License. */
 
 using float16 = paddle::platform::float16;
 
-PT_REGISTER_CTX_KERNEL(
+PT_REGISTER_KERNEL(
     sign, GPU, ALL_LAYOUT, pten::SignKernel, float, double, float16) {}