From 0f3b1ad6391c9d1c7e3ef563b6b78b9cf26eae93 Mon Sep 17 00:00:00 2001
From: ronnywang <ronny1996@163.com>
Date: Fri, 2 Dec 2022 14:11:00 +0800
Subject: [PATCH] fix phi capi kernel registration macro error (#48616)

* fix capi kernel registration macro error

* update
---
 paddle/phi/capi/include/kernel_registry.h |  4 +++-
 paddle/phi/capi/include/kernel_utils.h    | 16 +++++++++++-----
 paddle/phi/capi/lib/c_kernel_context.cc   | 12 ++++++++++--
 3 files changed, 24 insertions(+), 8 deletions(-)

diff --git a/paddle/phi/capi/include/kernel_registry.h b/paddle/phi/capi/include/kernel_registry.h
index 47ddc0bf5be..73318561dd9 100644
--- a/paddle/phi/capi/include/kernel_registry.h
+++ b/paddle/phi/capi/include/kernel_registry.h
@@ -167,6 +167,7 @@ inline std::vector<phi::capi::DenseTensor> PD_MultiInputAt(
   for (size_t i = 0; i < list.size; ++i) {
     ret.emplace_back(data[i]);
   }
+  PD_DeletePointerList(list);
   return ret;
 }
 
@@ -182,13 +183,14 @@ inline std::vector<phi::capi::DenseTensor> PD_MultiOutputAt(
   for (size_t i = 0; i < list.size; ++i) {
     ret.emplace_back(data[i]);
   }
+  PD_DeletePointerList(list);
   return ret;
 }
 
 template <typename T>
 inline std::vector<T *> PD_GetPointerVector(std::vector<T> *vec) {
   std::vector<T *> ret;
-  for (auto &item : vec) {
+  for (auto &item : *vec) {
     ret.push_back(&item);
   }
   return ret;
diff --git a/paddle/phi/capi/include/kernel_utils.h b/paddle/phi/capi/include/kernel_utils.h
index 6c1d3f3c0ee..d92a9e22052 100644
--- a/paddle/phi/capi/include/kernel_utils.h
+++ b/paddle/phi/capi/include/kernel_utils.h
@@ -564,18 +564,24 @@ namespace capi {
       static_assert(out_idx == 0,                                            \
                     "Kernel's Input should appear before Outputs.");         \
       auto arg = PD_MultiInputAt(ctx, in_idx);                               \
-      auto arg_wrapper = PD_GetPointerVector(&arg);                          \
+      std::vector<const tensor_type *> tensor_ptr_vec;                       \
+      for (auto &tensor : arg) {                                             \
+        tensor_ptr_vec.push_back(tensor.raw_data() ? &tensor : nullptr);     \
+      }                                                                      \
       CustomKernelCallHelper<Tail...>::                                      \
           template Compute<dev_ctx_idx, in_idx + 1, attr_idx, out_idx>(      \
-              ctx, pargs..., arg_wrapper);                                   \
+              ctx, pargs..., tensor_ptr_vec);                                \
     }                                                                        \
     template <int idx, typename... PreviousArgs>                             \
     static void VariadicCompute(const std::tuple<DevCtx, Args &...> &ctx,    \
                                 PreviousArgs &...pargs) {                    \
       auto &arg = std::get<idx>(ctx);                                        \
-      auto tensor = PD_TensorVector(reinterpret_cast<PD_Tensor *>(           \
+      auto tensor_vec = PD_TensorVector(reinterpret_cast<PD_Tensor *>(       \
           const_cast<std::vector<const tensor_type *> *>(&arg)));            \
-      auto tensor_ptr_vec = PD_GetPointerVector(&arg);                       \
+      std::vector<const tensor_type *> tensor_ptr_vec;                       \
+      for (auto &tensor : tensor_vec) {                                      \
+        tensor_ptr_vec.push_back(tensor.raw_data() ? &tensor : nullptr);     \
+      }                                                                      \
       return CustomKernelCallHelper<Tail...>::template VariadicCompute<idx + \
                                                                        1>(   \
           ctx, pargs..., tensor_ptr_vec);                                    \
@@ -681,7 +687,7 @@ namespace capi {
         tensor_ptr_vec.push_back(tensor.raw_data() ? &tensor : nullptr);     \
       }                                                                      \
       CustomKernelCallHelper<Tail...>::                                      \
-          template Compute<dev_ctx_idx, in_idx + 1, attr_idx, out_idx>(      \
+          template Compute<dev_ctx_idx, in_idx, attr_idx, out_idx + 1>(      \
               ctx, pargs..., tensor_ptr_vec);                                \
     }                                                                        \
     template <int idx, typename... PreviousArgs>                             \
diff --git a/paddle/phi/capi/lib/c_kernel_context.cc b/paddle/phi/capi/lib/c_kernel_context.cc
index d38a19038e3..e9fe2aada1f 100644
--- a/paddle/phi/capi/lib/c_kernel_context.cc
+++ b/paddle/phi/capi/lib/c_kernel_context.cc
@@ -60,7 +60,11 @@ PD_List PD_KernelContextMultiInputAt(PD_KernelContext* ctx, size_t index) {
       range.first, range.second);
   PD_List list;
   list.size = tensor_vec.size();
-  list.data = tensor_vec.data();
+  list.data = new void*[list.size];
+  for (size_t i = 0; i < list.size; ++i) {
+    (reinterpret_cast<void**>(list.data))[i] =
+        reinterpret_cast<void*>(const_cast<phi::DenseTensor*>(tensor_vec[i]));
+  }
   return list;
 }
 
@@ -78,7 +82,11 @@ PD_List PD_KernelContextMultiOutputAt(PD_KernelContext* ctx, size_t index) {
       range.first, range.second);
   PD_List list;
   list.size = tensor_vec.size();
-  list.data = tensor_vec.data();
+  list.data = new void*[list.size];
+  for (size_t i = 0; i < list.size; ++i) {
+    (reinterpret_cast<void**>(list.data))[i] =
+        reinterpret_cast<void*>(tensor_vec[i]);
+  }
   return list;
 }
 
-- 
GitLab