Support multi-output feature for elementwise (#38410)

* first commit

* pass ctest of  elementwise_div_grad

paddle/fluid/operators/elementwise/elementwise_op_broadcast.cu.h

@@ -162,7 +162,8 @@ struct DimensionsTransform {
   }
 };
 
-template <ElementwiseType ET, typename InT, typename OutT, typename Functor>
+template <ElementwiseType ET, typename InT, typename OutT, typename Functor,
+          int NumOuts = 1>
 void LaunchBroadcastElementwiseCudaKernel(
     const platform::CUDADeviceContext &ctx,
     const std::vector<const framework::Tensor *> &ins,
@@ -190,11 +191,12 @@ void LaunchBroadcastElementwiseCudaKernel(
   for (int i = 0; i < pt_outputs_tmp.size(); i++) {
     pt_outputs.push_back(pt_outputs_tmp[i].get());
   }
-  pten::LaunchBroadcastElementwiseCudaKernel<ET, InT, OutT>(
+  pten::LaunchBroadcastElementwiseCudaKernel<ET, InT, OutT, Functor, NumOuts>(
       ctx, pt_inputs, &pt_outputs, axis, func);
 }
 
-template <ElementwiseType ET, typename InT, typename OutT, typename Functor>
+template <ElementwiseType ET, typename InT, typename OutT, typename Functor,
+          int NumOuts = 1>
 void LaunchElementwiseCudaKernel(
     const platform::CUDADeviceContext &cuda_ctx,
     const std::vector<const framework::Tensor *> &ins,
@@ -222,8 +224,8 @@ void LaunchElementwiseCudaKernel(
   for (int i = 0; i < pt_outputs_tmp.size(); i++) {
     pt_outputs.push_back(pt_outputs_tmp[i].get());
   }
-  pten::LaunchElementwiseCudaKernel<ET, InT, OutT>(cuda_ctx, pt_inputs,
-                                                   &pt_outputs, axis, func);
+  pten::LaunchElementwiseCudaKernel<ET, InT, OutT, Functor, NumOuts>(
+      cuda_ctx, pt_inputs, &pt_outputs, axis, func);
 }
 
 }  // namespace operators

paddle/fluid/operators/elementwise/elementwise_op_impl.cu.h

@@ -38,7 +38,8 @@ namespace kps = paddle::operators::kernel_primitives;
 
 using ElementwiseType = pten::ElementwiseType;
 
-template <ElementwiseType ET, typename InT, typename OutT, typename Functor>
+template <ElementwiseType ET, typename InT, typename OutT, typename Functor,
+          int NumOuts = 1>
 void LaunchSameDimsElementwiseCudaKernel(
     const platform::CUDADeviceContext &ctx,
     const std::vector<const framework::Tensor *> &ins,
@@ -66,8 +67,8 @@ void LaunchSameDimsElementwiseCudaKernel(
   for (int i = 0; i < pt_outputs_tmp.size(); i++) {
     pt_outputs.push_back(pt_outputs_tmp[i].get());
   }
-  pten::LaunchSameDimsElementwiseCudaKernel<ET, InT, OutT>(ctx, pt_inputs,
-                                                           &pt_outputs, func);
+  pten::LaunchSameDimsElementwiseCudaKernel<ET, InT, OutT, Functor, NumOuts>(
+      ctx, pt_inputs, &pt_outputs, func);
 }
 
 }  // namespace operators

paddle/pten/kernels/hybird/cuda/elementwise/elementwise.h

@@ -19,7 +19,11 @@ limitations under the License. */
 
 namespace pten {
 
-template <ElementwiseType ET, typename InT, typename OutT, typename Functor>
+template <ElementwiseType ET,
+          typename InT,
+          typename OutT,
+          typename Functor,
+          int NumOuts = 1>
 void LaunchElementwiseCudaKernel(
     const paddle::platform::CUDADeviceContext &cuda_ctx,
     const std::vector<const DenseTensor *> &ins,
@@ -33,14 +37,14 @@ void LaunchElementwiseCudaKernel(
     dims_size.emplace_back(in->dims().size());
   }
   if (no_broadcast_flag) {
-    LaunchSameDimsElementwiseCudaKernel<ET, InT, OutT>(
+    LaunchSameDimsElementwiseCudaKernel<ET, InT, OutT, Functor, NumOuts>(
         cuda_ctx, ins, outs, func);
   } else {
     axis = axis == -1
                ? *std::max_element(dims_size.begin(), dims_size.end()) -
                      *std::min_element(dims_size.begin(), dims_size.end())
                : axis;
-    LaunchBroadcastElementwiseCudaKernel<ET, InT, OutT>(
+    LaunchBroadcastElementwiseCudaKernel<ET, InT, OutT, Functor, NumOuts>(
         cuda_ctx, ins, outs, axis, func);
   }
 }

paddle/pten/kernels/hybird/cuda/elementwise/elementwise_broadcast.cu.h

@@ -208,7 +208,7 @@ __device__ void ElementwiseBroadcastKernelImpl(
     int block_offset,
     Functor func) {
   InT args[Arity][VecSize];
-  OutType<OutT, NumOuts> result[VecSize];
+  ConditionalT<OutT, NumOuts> result[VecSize];
 
 #pragma unroll
   for (int i = 0; i < Arity; i++) {
@@ -224,7 +224,7 @@ __device__ void ElementwiseBroadcastKernelImpl(
   constexpr bool kCallElementwiseAny =
       paddle::platform::FunctionTraits<Functor>::has_pointer_args;
   ElementwisePrimitiveCaller<InT,
-                             OutType<OutT, NumOuts>,
+                             ConditionalT<OutT, NumOuts>,
                              VecSize,
                              Functor,
                              Arity,
@@ -455,18 +455,17 @@ void LaunchBroadcastElementwiseCudaKernel(
                         "is %d, the arity of functor is %d.",
                         ins.size(),
                         kArity));
-  PADDLE_ENFORCE_EQ(kArity,
-                    2,
+  PADDLE_ENFORCE_LE(kArity,
+                    ElementwiseType::kTernary,
                     paddle::platform::errors::InvalidArgument(
-                        "Currently only broadcast of binary is supported and "
-                        "verified, but received %d.",
+                        "Currently only broadcast of ternary is supported "
+                        "and verified, but received %d.",
                         kArity));
-  PADDLE_ENFORCE_EQ(
-      outs->size(),
+  PADDLE_ENFORCE_EQ(outs->size(),
                     NumOuts,
                     paddle::platform::errors::InvalidArgument(
                         "Number of outputs shall equal to number of functions, "
-          "but number of outputs is %d, number of functions is %d.",
+                        "but number of outputs is %d, of functions is %d.",
                         outs->size(),
                         NumOuts));
   int in_vec_size = 4;

paddle/pten/kernels/hybird/cuda/elementwise/elementwise_common.cu.h

@@ -27,7 +27,7 @@ enum ElementwiseType { kUnary = 1, kBinary = 2, kTernary = 3, kAny = -1 };
 /* Packing scalar type T(float, int etc.) into Array<T, NumOuts> type
    for supporting multiple-output feature in elementwise system.*/
 template <class T, int Num>
-using OutType =
+using ConditionalT =
     typename std::conditional_t<Num == 1, T, paddle::framework::Array<T, Num>>;
 
 template <typename InT,
@@ -86,7 +86,7 @@ template <typename OutT, int VecSize, bool IsBoundary, int NumOuts>
 struct ElementwiseWriteDataCaller {
   __device__ __forceinline__ void operator()(
       paddle::framework::Array<OutT *, NumOuts> outs,
-      OutType<OutT, NumOuts> src[VecSize],
+      ConditionalT<OutT, NumOuts> src[VecSize],
       int block_offset,
       int num) {
     OutT dst[NumOuts][VecSize];

paddle/pten/kernels/hybird/cuda/elementwise/elementwise_no_broadcast.cu.h

@@ -55,16 +55,17 @@ template <typename InT,
           typename OutT,
           typename Functor,
           int Arity,
+          int NumOuts,
           int VecSize,
           bool IsBoundary>
 __device__ void VectorizedElementwiseKernelImpl(
     const paddle::framework::Array<const InT *__restrict__, Arity> &in,
-    OutT *out,
+    paddle::framework::Array<OutT *, NumOuts> outs,
     int num,
     int data_offset,
     Functor func) {
   InT args[Arity][VecSize];
-  OutT result[VecSize];
+  ConditionalT<OutT, NumOuts> result[VecSize];
 
 #pragma unroll
   for (int i = 0; i < Arity; i++) {
@@ -73,36 +74,53 @@ __device__ void VectorizedElementwiseKernelImpl(
         args[i], in[i] + data_offset, num);
   }
 
-  const bool kCallElementwiseAny =
+  constexpr bool kCallElementwiseAny =
       paddle::platform::FunctionTraits<Functor>::has_pointer_args;
   ElementwisePrimitiveCaller<InT,
-                             OutT,
+                             ConditionalT<OutT, NumOuts>,
                              VecSize,
                              Functor,
                              Arity,
                              kCallElementwiseAny>()(func, args, result);
-  kps::WriteData<OutT, VecSize, 1, 1, IsBoundary>(
-      out + data_offset, result, num);
+
+  ElementwiseWriteDataCaller<OutT, VecSize, IsBoundary, NumOuts>()(
+      outs, result, data_offset, num);
 }
 
-template <typename InT, typename OutT, typename Functor, int Arity, int VecSize>
+template <typename InT,
+          typename OutT,
+          typename Functor,
+          int Arity,
+          int NumOuts,
+          int VecSize>
 __global__ void VectorizedElementwiseKernel(
     paddle::framework::Array<const InT *__restrict__, Arity> ins,
-    OutT *out,
+    paddle::framework::Array<OutT *, NumOuts> outs,
     int size,
     int main_offset,
     Functor func) {
   int data_offset = BLOCK_ID_X * BLOCK_NUM_X * VecSize;
   int stride = BLOCK_NUM_X * GRID_NUM_X * VecSize;
   for (; data_offset < main_offset; data_offset += stride) {
-    VectorizedElementwiseKernelImpl<InT, OutT, Functor, Arity, VecSize, false>(
-        ins, out, VecSize * BLOCK_NUM_X, data_offset, func);
+    VectorizedElementwiseKernelImpl<InT,
+                                    OutT,
+                                    Functor,
+                                    Arity,
+                                    NumOuts,
+                                    VecSize,
+                                    false>(
+        ins, outs, VecSize * BLOCK_NUM_X, data_offset, func);
   }
 
   int num = size - data_offset;
   if (num > 0) {
-    VectorizedElementwiseKernelImpl<InT, OutT, Functor, Arity, VecSize, true>(
-        ins, out, num, data_offset, func);
+    VectorizedElementwiseKernelImpl<InT,
+                                    OutT,
+                                    Functor,
+                                    Arity,
+                                    NumOuts,
+                                    VecSize,
+                                    true>(ins, outs, num, data_offset, func);
   }
 }
 
@@ -121,7 +139,12 @@ int GetVectorizedSizeForTensors(const std::vector<const DenseTensor *> &ins,
   return vec_size;
 }
 
-template <typename InT, typename OutT, typename Functor, int Arity, int VecSize>
+template <typename InT,
+          typename OutT,
+          typename Functor,
+          int Arity,
+          int NumOuts,
+          int VecSize>
 void ElementwiseCudaKernel(const paddle::platform::CUDADeviceContext &ctx,
                            const std::vector<const DenseTensor *> &ins,
                            std::vector<DenseTensor *> *outs,
@@ -131,11 +154,15 @@ void ElementwiseCudaKernel(const paddle::platform::CUDADeviceContext &ctx,
   int grid_size =
       ((numel + VecSize - 1) / VecSize + block_size - 1) / block_size;
   auto stream = ctx.stream();
-  OutT *out_data = (*outs)[0]->mutable_data<OutT>();
   paddle::framework::Array<const InT *__restrict__, Arity> ins_data;
-  for (int i = 0; i < Arity; i++) {
+  paddle::framework::Array<OutT *, NumOuts> outs_data;
+
+  for (int i = 0; i < Arity; ++i) {
     ins_data[i] = ins[i]->data<InT>();
   }
+  for (int i = 0; i < NumOuts; ++i) {
+    outs_data[i] = (*outs)[i]->mutable_data<OutT>();
+  }
 #ifdef PADDLE_WITH_XPU2
   block_size = 128;
   grid_size = 8;
@@ -144,20 +171,26 @@ void ElementwiseCudaKernel(const paddle::platform::CUDADeviceContext &ctx,
                               OutT,
                               Functor,
                               Arity,
+                              NumOuts,
                               VecSize><<<grid_size, block_size, 0, stream>>>(
-      ins_data, out_data, numel, main_offset, func);
+      ins_data, outs_data, numel, main_offset, func);
 #else
   int main_offset = (numel / (VecSize * block_size)) * VecSize * block_size;
   VectorizedElementwiseKernel<InT,
                               OutT,
                               Functor,
                               Arity,
+                              NumOuts,
                               VecSize><<<grid_size, block_size, 0, stream>>>(
-      ins_data, out_data, numel, main_offset, func);
+      ins_data, outs_data, numel, main_offset, func);
 #endif
 }
 
-template <ElementwiseType ET, typename InT, typename OutT, typename Functor>
+template <ElementwiseType ET,
+          typename InT,
+          typename OutT,
+          typename Functor,
+          int NumOuts = 1>
 void LaunchSameDimsElementwiseCudaKernel(
     const paddle::platform::CUDADeviceContext &ctx,
     const std::vector<const DenseTensor *> &ins,
@@ -174,19 +207,39 @@ void LaunchSameDimsElementwiseCudaKernel(
                         "is %d, the arity of functor is %d.",
                         ins.size(),
                         kArity));
+  PADDLE_ENFORCE_EQ(outs->size(),
+                    NumOuts,
+                    paddle::platform::errors::InvalidArgument(
+                        "Number of outputs shall equal to number of functions, "
+                        "but number of outputs is %d, of functions is %d.",
+                        outs->size(),
+                        NumOuts));
+
+  if (NumOuts > 1) {
+    for (int i = 1; i < NumOuts; ++i) {
+      PADDLE_ENFORCE_EQ(
+          (*outs)[i]->dims(),
+          (*outs)[0]->dims(),
+          paddle::platform::errors::InvalidArgument(
+              "The shape of each output tensor shall be identical yet, "
+              "but %dth output tensor`s shape is not.",
+              i));
+    }
+  }
+
   // calculate the max vec_size for all ins and outs
   int vec_size = GetVectorizedSizeForTensors<InT, OutT>(ins, *outs);
   switch (vec_size) {
     case 4:
-      ElementwiseCudaKernel<InT, OutT, Functor, kArity, 4>(
+      ElementwiseCudaKernel<InT, OutT, Functor, kArity, NumOuts, 4>(
           ctx, ins, outs, func);
       break;
     case 2:
-      ElementwiseCudaKernel<InT, OutT, Functor, kArity, 2>(
+      ElementwiseCudaKernel<InT, OutT, Functor, kArity, NumOuts, 2>(
           ctx, ins, outs, func);
       break;
     case 1:
-      ElementwiseCudaKernel<InT, OutT, Functor, kArity, 1>(
+      ElementwiseCudaKernel<InT, OutT, Functor, kArity, NumOuts, 1>(
           ctx, ins, outs, func);
       break;
     default: {