error: command 'gcc' failed with exit status 1

Question

error: command 'gcc' failed with exit status 1

minygd opened this issue 4 years ago · 7 comments

Hi, I meet an Error while building the Extension in a Server with Titan RTX and OS Ubuntu 18.04(gcc/g++ 5.5.0) machine. My Pytorch vision is 1.0.0 and all dependencies are well prepared.

A lot of message are presented as:

   T *grad_input_ptr           = grad_input.data_ptr<T>();

../Spherical-Package/ext_modules/include/cpp/mapped_avg_pool.h:98:56: error:expected primary-expression before ‘)’ token

So I think the reference code are in this link

And the final report is: error: command 'gcc' failed with exit status 1

Answer 1 · 2020-07-03T13:20:54.000Z

The issue is the pytorch version. I think it was around pytorch 1.3.0 when the C++ API changed for referencing a tensor’s pointer, which is what is now causing your error. Please update pytorch to the current stable version and this should go away.

…

On Jul 3, 2020, at 4:44 AM, minygd ***@***.***> wrote: Hi, I meet an Error while building the Extension in a Server with Titan RTX and OS Ubuntu 18.04(gcc/g++ 5.5.0) machine. My Pytorch vision is 1.0.0 and all dependencies are well prepared. A lot of message are presented as: T *grad_input_ptr = grad_input.data_ptr<T>(); ../Spherical-Package/ext_modules/include/cpp/mapped_avg_pool.h:98:56: error:expected primary-expression before ‘)’ token And the final report is: error: command 'gcc' failed with exit status 1 — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

Answer 2 · 2020-07-03T15:03:33.000Z

I've updated the readme to show that this needs at least pytorch 1.4.0 or higher.

Answer 3 · 2020-07-03T23:56:33.000Z

@meder411 Hi, thank you. After changing to Pytorch 1.5.1, the errors mentioned above have gone. But another error occurs:

../../Spherical-Package/ext_modules/include/core/distortion.h:144:67: error: no matching function for call to ‘host_device_max(double, int)’
                                          abs(rel_step_size * x(1)));

I search for function host_device_max(), and finally get it in this link. So I have tried to paste it in distortion.h, but failed to build with the error and also a note below:(Since I paste it in distortion.hL#40 and build)

../../Spherical-Package/ext_modules/include/core/distortion.h:41:30: note: candidate: template<class T> T spherical::core::host_device_max(T, T)
 __host__ __device__ inline T host_device_max(const T val1, const T val2) {

This is with the Environment(Pytorch 1.5.1 / Cudatoolkit 10.2 / python 3.8), and all requirements are well prepared.

Answer 4 · 2020-07-15T17:48:57.000Z

Sorry for the delay. Let me take a look, and I'll see if I can replicate this.

Answer 5 · 2020-07-15T22:33:35.000Z

It's fine, THX!

Answer 6 · 2020-07-28T01:09:54.000Z

Sorry for the delay. Let me take a look, and I'll see if I can replicate this.

Hi, I still can't implement the Spherical-Package, do you replicate it?

Answer 7 · 2020-07-28T04:01:37.000Z

I just checked everything using the provided Docker setup. There actually was a typo with the Dockerfile, but this was a different error than you are running into. With the changes I just pushed (v.1.0.3), the Docker setup works correctly and all tests pass and examples run.

If you are still running into this issue, I recommend either using the provided Docker setup or simply running the commands from the Dockerfile sequentially to manually install the dependencies. That said, it definitely doesn't look like a dependency issue.

I'm not entirely sure what might be causing your issue. It looks like an include might have been deleted somewhere? You shouldn't have to move that function to distortion.h. Perhaps do a git pull or a fresh git clone?