marsaev

Tracking internally: AMGX-57

No, it's CUDA only at the moment. You might try https://github.com/ROCm-Developer-Tools/HIPIFY , but i haven't tried it myself on AMGX.

Hi @ztdepztdep , Mixed precision solver is enabled through different value of API parameter. You can see here: https://github.com/NVIDIA/AMGX/blob/main/examples/amgx_capi.c#L266 how `mode` value is set and this value is used in...

> if we store the matrix in float, then at last how to iterative refinement its final solution to double precision which is true meaning of "mixed precision" Restriction/prolongation matrices...

> It seems the second D or F specifies the matrix precision. I tried to set dFDI, but AMGX feedback error "no this kind of mode ". It's other way...

@Soujanyajanga, let me know if we answered your question or you need more details

Which MPI do you use? it seems that compiler error points out to the MPI_Allgather function. Can you check that signature of MPI_Allgather() in your MPI matches with https://www.mpich.org/static/docs/latest/www3/MPI_Allgather.html ?

I'm checking the build with CUDA 11.6, GCC 7.5 and some OpenMpi: ``` # CUDACXX=/usr/local/cuda/bin/nvcc /opt/cmake/bin/cmake -DCMAKE_CUDA_ARCHITECTURES="60;70;80" ../ -- The C compiler identification is GNU 7.5.0 -- The CXX compiler...

@ma-sadeghi That is a tricky one. Problem is that there are dozens of configurations possible: builds for each AMGX version - for each platform - for each CUDA version -...

Hi @luisaforozco , Can you share you `cmake` command line and it's output and what error specifically do you get? Thanks,