You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Paddle/paddle/fluid/operators/elementwise/elementwise_floordiv_op.h

92 lines
2.8 KiB

/* Copyright (c) 2019 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. */
#pragma once
#include "paddle/fluid/framework/eigen.h"
#include "paddle/fluid/operators/elementwise/elementwise_op.h"
#include "paddle/fluid/operators/elementwise/elementwise_op_function.h"
#include "paddle/fluid/operators/math/blas.h"
namespace paddle {
namespace operators {
template <typename T>
struct FloorDivFunctor {
inline HOSTDEVICE T operator()(T a, T b) const {
#ifdef __CUDA_ARCH__
if (b == 0) {
printf("Error: Divide by zero encounter in floor_divide\n");
asm("trap;");
}
#else
if (b == 0)
PADDLE_THROW(platform::errors::InvalidArgument(
"Divide by zero encounter in floor_divide"));
#endif
return static_cast<T>(std::trunc(a / b));
}
};
template <typename T>
struct InverseFloorDivFunctor {
inline HOSTDEVICE T operator()(T a, T b) const {
#ifdef __CUDA_ARCH__
if (a == 0) {
printf("Error: Divide by zero encounter in floor_divide\n");
asm("trap;");
}
#else
if (a == 0)
PADDLE_THROW(platform::errors::InvalidArgument(
"Divide by zero encounter in floor_divide"));
#endif
return static_cast<T>(std::trunc(b / a));
}
};
template <typename DeviceContext, typename T>
void elementwise_floor_div(const framework::ExecutionContext &ctx,
const framework::Tensor *x,
const framework::Tensor *y, framework::Tensor *z) {
int axis = ctx.Attr<int>("axis");
auto x_dims = x->dims();
auto y_dims = y->dims();
if (x_dims.size() >= y_dims.size()) {
ElementwiseComputeEx<FloorDivFunctor<T>, DeviceContext, T>(
ctx, x, y, axis, FloorDivFunctor<T>(), z);
} else {
ElementwiseComputeEx<InverseFloorDivFunctor<T>, DeviceContext, T>(
ctx, x, y, axis, InverseFloorDivFunctor<T>(), z);
}
}
template <typename DeviceContext, typename T>
class ElementwiseFloorDivKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext &ctx) const override {
auto *x = ctx.Input<framework::LoDTensor>("X");
auto *y = ctx.Input<framework::LoDTensor>("Y");
auto *z = ctx.Output<framework::LoDTensor>("Out");
z->mutable_data<T>(ctx.GetPlace());
// dtype of x and y is int64 or int32
elementwise_floor_div<DeviceContext, T>(ctx, x, y, z);
}
};
} // namespace operators
} // namespace paddle