error: parameter packs not expanded with ‘...’

[duckworthd]

I suspect this issue is not directly related to instant-ngp's code, but I'll post it here all the same in case anyone else sees a similar issue.

TL;DR there's an issue compiling std_function.h, a dependency of tiny-cuda-nn.

> cmake --build build --config RelWithDebInfo

Consolidate compiler generated dependencies of target tiny-cuda-nn
[  1%] Building CUDA object dependencies/tiny-cuda-nn/src/CMakeFiles/tiny-cuda-nn.dir/common.cu.o
/usr/include/c++/11/bits/std_function.h:435:145: error: parameter packs not expanded with ‘...’:
  435 |         function(_Functor&& __f)
      |                                                                                                                                                 ^
/usr/include/c++/11/bits/std_function.h:435:145: note:         ‘_ArgTypes’
/usr/include/c++/11/bits/std_function.h:530:146: error: parameter packs not expanded with ‘...’:
  530 |         operator=(_Functor&& __f)
      |                                                                                                                                                  ^
/usr/include/c++/11/bits/std_function.h:530:146: note:         ‘_ArgTypes’
gmake[2]: *** [dependencies/tiny-cuda-nn/src/CMakeFiles/tiny-cuda-nn.dir/build.make:76: dependencies/tiny-cuda-nn/src/CMakeFiles/tiny-cuda-nn.dir/common.cu.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:306: dependencies/tiny-cuda-nn/src/CMakeFiles/tiny-cuda-nn.dir/all] Error 2
gmake: *** [Makefile:91: all] Error 2

Here's a bit more info on my setup. I'm using CUDA v11.4.4 and GCC v.11.2.0 on Debian. I saw an identical issue when compiling with CUDA v11.6.0.

> nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Mon_Oct_11_21:27:02_PDT_2021
Cuda compilation tools, release 11.4, V11.4.152
Build cuda_11.4.r11.4/compiler.30521435_0

> gcc --version
gcc (Debian 11.2.0-13) 11.2.0
Copyright (C) 2021 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Any tips appreciated!

[krasin]

Just as a data point, I don't have this problem and my compiler versions are:

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Fri_Dec_17_18:16:03_PST_2021
Cuda compilation tools, release 11.6, V11.6.55
Build cuda_11.6.r11.6/compiler.30794723_0

$ gcc --version
gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Note that my GCC version is much older (since I test it on Ubuntu 20.04), while nvcc is from 11.6. Not sure, if that's relevant.

[krasin]

I got curious and tried it with gcc-11/g++-11:

$ mkdir build
$ cd build
$ CC=gcc-11 CXX=g++-11 cmake -GNinja -DCMAKE_BUILD_TYPE=RelWithDebInfo  ..
-- The CXX compiler identification is GNU 11.1.0                                                                                                                                                           
-- The CUDA compiler identification is NVIDIA 11.6.55
...
$ ninja
...

It succeeded.

There must be something else in that Debian environment that makes it wrong.

[duckworthd]

A colleague identified the issue as originating from gcc. Using an older build of GCC (GCC v10.3.1, to be precise) solved my issue.

$ sudo apt install gcc-10 g++-10
$ export CC=/usr/bin/gcc-10
$ export CXX=/usr/bin/g++-10
$ export CUDA_ROOT=/usr/local/cuda
$ ln -s /usr/bin/gcc-10 $CUDA_ROOT/bin/gcc
$ ln -s /usr/bin/g++-10 $CUDA_ROOT/bin/g++
(Build Instant-NGP as described)

[geoeo]

This has reappeared on ubuntu jammy. Default (Aptitude) gcc is 11.2 and cuda version is 11.6.

This is also not fixed with downgrading to gcc-10

The problem is that the aptitude version is cuda toolkit version is 11.5.
When 11.6 is used then this is fixed!
Currently you have to download the toolkit via the ubuntu 20.04 runfile
Then install via sudo sh cuda_11.6.2_510.47.03_linux.run --toolkit --silent --override
Also set your environment flags appropriately.

For me this fixed the issue. I guess the aptitude version should be upgraded!

[sherwoac]

A colleague identified the issue as originating from gcc. Using an older build of GCC (GCC v10.3.1, to be precise) solved my issue.

$ sudo apt install gcc-10 g++-10
$ export CC=/usr/bin/gcc-10
$ export CXX=/usr/bin/g++-10
$ export CUDA_ROOT=/usr/local/cuda
$ ln -s /usr/bin/gcc-10 $CUDA_ROOT/bin/gcc
$ ln -s /usr/bin/g++-10 $CUDA_ROOT/bin/g++
(Build Instant-NGP as described)

Switching to gcc/g++-10 worked for me too. The healthy way to do this is to use update-alternatives and set the default to gcc/++-10 for the build.

[vikasTmz]

A colleague identified the issue as originating from gcc. Using an older build of GCC (GCC v10.3.1, to be precise) solved my issue.

$ sudo apt install gcc-10 g++-10
$ export CC=/usr/bin/gcc-10
$ export CXX=/usr/bin/g++-10
$ export CUDA_ROOT=/usr/local/cuda
$ ln -s /usr/bin/gcc-10 $CUDA_ROOT/bin/gcc
$ ln -s /usr/bin/g++-10 $CUDA_ROOT/bin/g++
(Build Instant-NGP as described)

This worked for me as well.

Specs
Ubuntu: 22.04.1 LTS
GPU: Quadro P6000
CUDA version: 11.6

original GCC/G++ version == 11.2 did not work. Downgraded to 10 and it worked.

[lucasjinreal]

Why the solution is change gcc not change code?
change g++10 not work for me, but the code can be changed to support more decent gccs