wehub-resource-sync
113 lines
3.6 KiB
Plaintext
113 lines
3.6 KiB
Plaintext
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
|
|
//
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
//
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// 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/nonzero_kernel.h"
|
|
#include "paddle/common/ddim.h"
|
|
#include "paddle/phi/core/dense_tensor.h"
|
|
#include "paddle/phi/core/kernel_registry.h"
|
|
#include "paddle/phi/kernels/funcs/cub.h"
|
|
#include "paddle/phi/kernels/funcs/math_function.h"
|
|
#include "paddle/phi/kernels/funcs/select_impl.cu.h"
|
|
|
|
namespace phi {
|
|
template <typename MaskT, typename IndexT, typename OutT>
|
|
struct IndexFunctor {
|
|
IndexT strides[DDim::kMaxRank];
|
|
int rank;
|
|
|
|
explicit IndexFunctor(const DDim &in_dims) {
|
|
rank = in_dims.size();
|
|
// Get strides according to in_dims
|
|
strides[0] = 1;
|
|
for (IndexT i = 1; i < rank; i++) {
|
|
strides[i] = strides[i - 1] * in_dims[rank - i];
|
|
}
|
|
}
|
|
|
|
HOSTDEVICE inline void operator()(OutT *out,
|
|
const MaskT *mask,
|
|
const IndexT *index,
|
|
const int num) {
|
|
int store_fix = 0;
|
|
for (int idx = 0; idx < num; idx++) {
|
|
if (mask[idx]) {
|
|
IndexT data_index = index[idx];
|
|
// get index
|
|
for (int rank_id = rank - 1; rank_id >= 0; --rank_id) {
|
|
out[store_fix] = static_cast<OutT>(data_index / strides[rank_id]);
|
|
data_index = data_index % strides[rank_id];
|
|
store_fix++;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
};
|
|
|
|
template <typename T, typename Context>
|
|
void NonZeroKernel(const Context &dev_ctx,
|
|
const DenseTensor &condition,
|
|
DenseTensor *out) {
|
|
if (condition.numel() == 0) {
|
|
dev_ctx.template Alloc<int64_t>(out);
|
|
return;
|
|
}
|
|
DenseTensor in_data;
|
|
auto dims = condition.dims();
|
|
using Functor = IndexFunctor<T, int64_t, int64_t>;
|
|
Functor index_functor = Functor(dims);
|
|
funcs::SelectKernel<T, T, int64_t, 0, Functor>(
|
|
dev_ctx, condition, in_data, out, index_functor);
|
|
}
|
|
|
|
template <typename T, typename Context>
|
|
void RestrictNonZeroKernel(const Context &dev_ctx,
|
|
const DenseTensor &condition,
|
|
const int64_t total_true_num,
|
|
DenseTensor *out) {
|
|
DenseTensor in_data;
|
|
auto dims = condition.dims();
|
|
if (condition.numel() == 0) {
|
|
dev_ctx.template Alloc<int64_t>(out);
|
|
return;
|
|
}
|
|
|
|
using Functor = IndexFunctor<T, int64_t, int64_t>;
|
|
Functor index_functor{dims};
|
|
|
|
funcs::RestrictSelectKernel<T, T, int64_t, 0, Functor>(
|
|
dev_ctx, condition, in_data, total_true_num, out, index_functor);
|
|
}
|
|
|
|
} // namespace phi
|
|
|
|
PD_REGISTER_KERNEL(nonzero,
|
|
GPU,
|
|
ALL_LAYOUT,
|
|
phi::NonZeroKernel,
|
|
int64_t,
|
|
int,
|
|
int16_t,
|
|
phi::float16,
|
|
phi::bfloat16,
|
|
bool,
|
|
float,
|
|
double,
|
|
phi::complex64,
|
|
phi::complex128) {
|
|
kernel->OutputAt(0).SetDataType(phi::DataType::INT64);
|
|
}
|
|
|
|
PD_REGISTER_KERNEL(
|
|
restrict_nonzero, GPU, ALL_LAYOUT, phi::RestrictNonZeroKernel, bool) {}
|