Error: command 'gcc' failed with exit status 1

[munib94]

Hi,

I tried running python setup.py install but I experienced the following error:

`
src/ball_query.cpp:10:23: error: ‘AT_CHECK’ was not declared in this scope; did you mean ‘DCHECK’?
10 | #define CHECK_CUDA(x) AT_CHECK(x.type().is_cuda(), #x, " must be a CUDAtensor ")
| ^~~~~~~~
src/ball_query.cpp:10:23: note: in definition of macro ‘CHECK_CUDA’
10 | #define CHECK_CUDA(x) AT_CHECK(x.type().is_cuda(), #x, " must be a CUDAtensor ")
| ^~~~~~~~
src/ball_query.cpp:16:5: note: in expansion of macro ‘CHECK_INPUT’
16 | CHECK_INPUT(new_xyz_tensor);
| ^~~~~~~~~~~
src/ball_query.cpp:10:39: warning: ‘at::DeprecatedTypeProperties& at::Tensor::type() const’ is deprecated: Tensor.type() is deprecated. Instead use Tensor.options(), which in many cases (e.g. in a constructor) is a drop-in replacement. If you were using data from type(), that is now available from Tensor itself, so instead of tensor.type().scalar_type(), use tensor.scalar_type() instead and instead of tensor.type().backend() use tensor.device(). [-Wdeprecated-declarations]
10 | #define CHECK_CUDA(x) AT_CHECK(x.type().is_cuda(), #x, " must be a CUDAtensor ")
| ^
src/ball_query.cpp:12:24: note: in expansion of macro ‘CHECK_CUDA’
12 | #define CHECK_INPUT(x) CHECK_CUDA(x);CHECK_CONTIGUOUS(x)
| ^~~~~~~~~~
src/ball_query.cpp:17:5: note: in expansion of macro ‘CHECK_INPUT’
17 | CHECK_INPUT(xyz_tensor);
| ^~~~~~~~~~~
^
.
.
.
from src/ball_query.cpp:1:
/home/munib/.pyenv/versions/miniconda3-latest/envs/flownet3d_pytorch/lib/python3.5/site-packages/torch/include/ATen/core/TensorBody.h:341:7: note: declared here
341 | T * data() const {
| ^~~~
src/ball_query.cpp:22:27: error: ‘THCState_getCurrentStream’ was not declared in this scope
22 | cudaStream_t stream = THCState_getCurrentStream(state);
| ^~~~~~~~~~~~~~~~~~~~~~~~~
error: command 'gcc' failed with exit status 1

`
I can't find any solutions. Do you know how to fix this issue?

System:
OS: Ubuntu 20.04 LTS
Python version: 3.5
PyTorch version: 1.5.1
Torchvision version: 0.6.1
CUDA/cuDNN version: 10.1/7.6.4
GPU: Nvidia Titan RTX

[munib94]

After further investigation, I came upon this: pytorch/pytorch#14500

It seems like it is an issue with PyTorch 1.5. I replaced THCState_getCurrentStream with
at::cuda::getCurrentCUDAStream() in src/ball_query.cpp and it fixed that issue. Can you confirm this is an issue with PyTorch 1.5?

I re-ran python setup.py install but I still ran into an error. Do you know what the issue could be? I attached the full output on the terminal here:
setup_install_output.txt

[hyangwinter]

Your problem is more like this: https://stackoverflow.com/questions/61559479/cannot-compile-nms-cuda-in-mmdetection-using-colaboratory
open-mmlab/mmdetection#2513

Maybe you should replace AT_CHECK with TORCH_CHECK to support pytorch 1.5 or downgrade your pytorch version to 1.4 .

[munib94]

I replaced all mentions of AT_CHECK and THCState_getCurrentStream(state) with TORCH_CHECK and at::cuda::getCurrentCUDAStream(), respectively. I was then able to install setup.py without errors. Thank you for your help!

[QLuanWilliamed]

I replaced all mentions of AT_CHECK and THCState_getCurrentStream(state) with TORCH_CHECK and at::cuda::getCurrentCUDAStream(), respectively. I was then able to install setup.py without errors. Thank you for your help!

Hi, could you share your revised code here? because I am not familiar with C++ languages. Thanks!

[sunyanzeng]

Why did I replace these two functions with the latest one, and still have an error

[ChidanandKumarKS]

worked well in my case by replacing THCState_getCurrentStream with at::cuda::getCurrentCUDAStream()