Build error in std::tuple for cuda 11.1 with gcc102/gcc103 using c++17

[valassi]

This is similar to #290.

During a cleanup of Makefiles in #281 I have moved cuda from c++14 to c++17 as was already the case for c++. But then I also have to use cuda11.1 because of performance regression #282. Now that I am using more combinations, I am getting more errors:

ccache /usr/local/cuda-11.1/bin/nvcc  -O3  -lineinfo -std=c++17 -I. -I../../src -I../../../../../tools -I../../../../../tools -I/usr/local/cuda-11.1/include/ -DUSE_NVTX -arch=compute_70 -use_fast_math -ccbin /cvmfs/sft.cern.ch/lcg/releases/gcc/10.3.0-f5826/x86_64-centos7/bin/g++ -DMGONGPU_FPTYPE_DOUBLE -DMGONGPU_CURAND_ONDEVICE -c gcheck_sa.cu -o gcheck_sa.o
/cvmfs/sft.cern.ch/lcg/releases/gcc/10.3.0-f5826/x86_64-centos7/include/c++/10.3.0/tuple(570): error: pack "_UElements" does not have the same number of elements as "_Elements"
          detected during:
            instantiation of "__nv_bool std::tuple<_Elements...>::__nothrow_constructible<_UElements...>() [with _Elements=<const std::string &>, _UElements=<>]" 
/cvmfs/sft.cern.ch/lcg/releases/gcc/10.3.0-f5826/x86_64-centos7/include/c++/10.3.0/bits/stl_map.h(502): here
            instantiation of "std::map<_Key, _Tp, _Compare, _Alloc>::mapped_type &std::map<_Key, _Tp, _Compare, _Alloc>::operator[](const std::map<_Key, _Tp, _Compare, _Alloc>::key_type &) [with _Key=std::string, _Tp=float, _Compare=std::less<std::string>, _Alloc=std::allocator<std::pair<const std::string, float>>]" 
timermap.h(40): here

1 error detected in the compilation of "gcheck_sa.cu".

This looks similar to https://githubmemory.com/repo/alpaka-group/cupla/issues/190

[valassi]

I confirm I do not get this issue with cuda 11.2 and gcc102 (I get other issues with gcc103, related to my buggy gcc103 I think). I will use c++17 only with cuda 11.2 and above.

[valassi]

Fixed in an upcoming poatch for PR #281

[valassi]

Note that because of the issues discussed in #277, #281, #282, I have downgraded back from cuda 11.4 to cuda 11.1 as baseline. Unlike cuda 11.0, this allows using gcc10, but I am using now gcc10.2 (as my cvmfs 10.3 has some issues). I am using c++17 in gcc and c++14 in cuda.