问题描述
您好,我正在尝试构建一个自定义 python 包,其中包括 cuda 文件和相应的 cpp 和头文件。 我的目录结构是这样的
对下面有点长的问题表示歉意。我不太确定错误的确切位置,因此我提供了大部分相关内容。
头文件如下-
utils.h
#pragma once
#include <ATen/cuda/CUDAContext.h>
#include <torch/extension.h>
#define CHECK_CUDA(x) \
do \
{ \
AT_ASSERT(x.is_cuda(),#x " must be a CUDA tensor"); \
} while (0)
cuda_utils.h
#ifndef _CUDA_UTILS_H
#define _CUDA_UTILS_H
#include <ATen/ATen.h>
#include <ATen/cuda/CUDAContext.h>
#include <cmath>
#include <cuda.h>
#include <cuda_runtime.h>
#include <vector>
#define TOTAL_THREADS 512
inline int opt_n_threads(int work_size)
{
const int pow_2 = std::log(static_cast<double>(work_size)) / std::log(2.0);
return max(min(1 << pow_2,TOTAL_THREADS),1);
}
#define CUDA_CHECK_ERRORS() \
do \
{ \
cudaError_t err = cudaGetLastError(); \
if (cudaSuccess != err) \
{ \
fprintf(stderr,"CUDA kernel Failed : %s\n%s at L:%d in %s\n",\
cudaGetErrorString(err),__PRETTY_FUNCTION__,__LINE__,\
__FILE__); \
exit(-1); \
} \
} while (0)
#endif
cuda_test.h
#pragma once
#include <torch/extension.h>
at::Tensor cuda_test(
at::Tensor verts_list,at::Tensor verts_list_length,const float radius,const int nsample,const int current_nearest_vertex_index);
CUDA 文件
cuda_test.cu
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include "cuda_utils.h"
__global__ void query(int b,float radius,int nsample,int current_nearest_vertex_index,const float *__restrict__ verts_list,const int *__restrict__ verts_list_length
int *__restrict__ query_idx)
{
int batch_idx = blockIdx.x;
<some implementation>
}
void query_wrapper(int b,const float *verts_list,const int *verts_list_length,int *query_idx
)
{
cudaStream_t stream = at::cuda::getCurrentCUDAStream();
query<<<b,opt_n_threads(1),stream>>>(b,radius,nsample,current_nearest_vertex_index,verts_list,verts_list_length,query_idx);
CUDA_CHECK_ERRORS();
}
#include "cuda_test.h"
#include "utils.h"
void query_wrapper(int b,int *query_idx);
at::Tensor cuda_test(
at::Tensor verts_list,const int current_nearest_vertex_index)
{
CHECK_CONTIGUOUS(verts_list);
CHECK_IS_FLOAT(verts_list);
CHECK_CONTIGUOUS(verts_list_length);
CHECK_IS_INT(verts_list_length);
at::Tensor query_idx =
torch::zeros({verts_list.size(0),nsample},at::device(verts_list.device()).dtype(at::ScalarType::Int));
if (verts_list.is_cuda())
{
query_wrapper(verts_list.size(0),verts_list.data_ptr<float>(),verts_list_length.data_ptr<int>(),query_idx.data_ptr<int>());
}
else
{
AT_ASSERT(false,"cpu not supported");
}
return idx;
}
bindings.cpp
#include "cuda_test.h"
PYBIND11_MODULE(TORCH_EXTENSION_NAME,m)
{
m.def("cuda_test",&cuda_test);
}
我尝试构建并安装它,但出现以下错误
build/temp.linux-x86_64-3.7/test_ops/_ext-src/src/cuda_test.o: In function `__device_stub__Z5queryifiiPKfPKiPi(int,float,int,float const*,int const*,int*)':
tmpxft_000001a4_00000000-6_cuda_test.compute_75.cudafe1.cpp:(.text+0x10): multiple deFinition of `__device_stub__Z5queryifiiPKfPKiPi(int,int*)'
build/temp.linux-x86_64-3.7/test_ops/_ext-src/src/cuda_test.o:tmpxft_000001a4_00000000-6_cuda_test.compute_75.cudafe1.cpp:(.text+0x10): first defined here
build/temp.linux-x86_64-3.7/test_ops/_ext-src/src/cuda_test.o: In function `query(int,int*)':
tmpxft_000001a4_00000000-6_cuda_test.compute_75.cudafe1.cpp:(.text+0x180): multiple deFinition of `query(int,int*)'
build/temp.linux-x86_64-3.7/test_ops/_ext-src/src/cuda_test.o:tmpxft_000001a4_00000000-6_cuda_test.compute_75.cudafe1.cpp:(.text+0x180): first defined here
build/temp.linux-x86_64-3.7/test_ops/_ext-src/src/cuda_test.o: In function `query_wrapper(int,int*)':
tmpxft_000001a4_00000000-6_cuda_test.compute_75.cudafe1.cpp:(.text+0x190): multiple deFinition of `query_wrapper(int,int*)'
build/temp.linux-x86_64-3.7/test_ops/_ext-src/src/cuda_test.o:tmpxft_000001a4_00000000-6_cuda_test.compute_75.cudafe1.cpp:(.text+0x190): first defined here
collect2: error: ld returned 1 exit status
我遵循了 this link 中的确切结构,该结构完全正确。我想我已经尽我所能做对了。
我在构建和安装过程中是否犯了任何错误? 提前谢谢你!!
解决方法
暂无找到可以解决该程序问题的有效方法,小编努力寻找整理中!
如果你已经找到好的解决方法,欢迎将解决方案带上本链接一起发送给小编。
小编邮箱:dio#foxmail.com (将#修改为@)