mirror of
https://github.com/saymrwulf/pytorch.git
synced 2026-05-14 20:57:59 +00:00
f53010370b
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/15651 Add count_include_pad arg for PoolOpGradient on CPU and fix ARM performance issue. Reviewed By: houseroad Differential Revision: D13564257 fbshipit-source-id: 3a143f1122bc507ccb7827e9b46908d5c7203735
65 lines
1.1 KiB
C++
65 lines
1.1 KiB
C++
#ifndef CAFFE2_OPERATORS_POOL_OP_UTIL_H_
|
|
#define CAFFE2_OPERATORS_POOL_OP_UTIL_H_
|
|
|
|
#include "caffe2/core/types.h"
|
|
#include "caffe2/utils/cpu_neon.h"
|
|
|
|
namespace caffe2 {
|
|
namespace pool_op_util {
|
|
|
|
bool IsNeon4x4p0s0Eligible(
|
|
int input_h,
|
|
int input_w,
|
|
int output_h,
|
|
int output_w,
|
|
int kh,
|
|
int kw,
|
|
int stride_h,
|
|
int stride_w,
|
|
int pad_t,
|
|
int pad_l,
|
|
int pad_b,
|
|
int pad_r,
|
|
int dilation_h,
|
|
int dilation_w,
|
|
const float* X,
|
|
float* Y);
|
|
|
|
bool IsNeon2x2p0s0Eligible(
|
|
int input_h,
|
|
int input_w,
|
|
int output_h,
|
|
int output_w,
|
|
int kh,
|
|
int kw,
|
|
int stride_h,
|
|
int stride_w,
|
|
int pad_t,
|
|
int pad_l,
|
|
int pad_b,
|
|
int pad_r,
|
|
int dilation_h,
|
|
int dilation_w,
|
|
const float* X,
|
|
float* Y);
|
|
|
|
void RunNeonAveragePool4x4p0s0NCHW(
|
|
int N,
|
|
int C,
|
|
int H,
|
|
int W,
|
|
const float* X,
|
|
float* Y);
|
|
|
|
void RunNeonMaxPool2x2p0s0NCHW(
|
|
int N,
|
|
int C,
|
|
int H,
|
|
int W,
|
|
const float* X,
|
|
float* Y);
|
|
|
|
} // namespace pool_op_util
|
|
} // namespace caffe2
|
|
|
|
#endif // CAFFE2_OPERATORS_POOL_OP_UTIL_H_
|