[petsc-dev] [petsc-maint #88993] Petsc with Cuda 4.0 and Multiple GPUs

[Barry Smith]

  Dang, I had two copies of thrust and cusp and was using the "wrong" one hence everything was working for me. 

   Ok, I'll try to get txpetscgpu updated


    Barry

On Oct 3, 2011, at 10:18 PM, Satish Balay wrote:

>>>> 
> /home/wdn/Projects/Petsc/src/branches/master/petsc-dev/LINUX_GNU_OPTIMIZE_SERIAL_CUDA_40_LITE/include/txpetscgpu/include/csr_spmv_part_vector_gpu.h:23:44: error: thrust/detail/device/cuda/arch.h: No such file or directory
> /usr/bin/ar: aijcusp.o: No such file or directory
> <<<
> 
> This file is at ./thrust/detail/backend/cuda/arch.h in the
> latest thrust.
> 
> Looks like txpetscgpu isn't updated to latest thrust.
> 
> [fixing 'device' -> 'backend' gives me different errors]
> 
> Satish
> 
>>>>>>>>>>>> 
> balay at bb30:~/petsc-dev/src/mat/impls/aij/seq/seqcusp>make
> nvcc -g -arch=sm_13  -c --compiler-options="-Wall -Wwrite-strings -Wno-strict-aliasing -Wno-unknown-pragmas -g3   -I/home/balay/petsc-dev/include -I/home/balay/petsc-dev/arch-cuda-double-tx/include -I/usr/local/cuda/include -I/home/balay/petsc-dev/arch-cuda-double-tx/include/txpetscgpu/include    -D__INSDIR__=src/mat/impls/aij/seq/seqcusp/" aijcusp.cu aijAssemble.cu
> /home/balay/petsc-dev/arch-cuda-double-tx/include/txpetscgpu/include/csr_tri_solve_gpu_level_scheduler.h(312): error: name followed by "::" must be a class or namespace name
>          detected during instantiation of "int csr_tri_solve_level_scheduler(const csr_matrix_data<index_type, value_type> *, index_type, index_type, index_type, const index_type *, const index_type *, const index_type *, value_type *) [with index_type=PetscInt, value_type=PetscScalar]" 
> aijcusp.cu(682): here
> 
> /home/balay/petsc-dev/arch-cuda-double-tx/include/txpetscgpu/include/csr_spmv_part_vector_gpu.h(275): error: name followed by "::" must be a class or namespace name
>          detected during:
>            instantiation of "bool csr_spmv_part_vector_gpu<assign_add_subtract,index_type,value_type>(const csr_matrix_data<index_type, value_type> &, index_type, index_type, index_type, index_type, index_type, const value_type *, value_type *, index_type, index_type, index_type, unsigned int) [with assign_add_subtract=operator_sub, index_type=PetscInt, value_type=PetscScalar]" 
> (412): here
>            instantiation of "bool csr_spmv_sub_part_vector_gpu(const csr_matrix_data<index_type, value_type> &, index_type, index_type, index_type, index_type, index_type, const value_type *, value_type *, index_type, index_type, index_type, unsigned int) [with index_type=PetscInt, value_type=PetscScalar]" 
> /home/balay/petsc-dev/arch-cuda-double-tx/include/txpetscgpu/include/csr_tri_solve_gpu.h(111): here
>            instantiation of "bool csr_tri_solve_gpu_hybrid(const csr_matrix_data<index_type, value_type> &, index_type *, uplo_enum, value_type *, value_type *, const csr_matrix_data<index_type, value_type> &, cuda_memory<value_type> &, cuda_memory<value_type> &, index_type, index_type, unsigned int) [with index_type=PetscInt, value_type=PetscScalar]" 
> /home/balay/petsc-dev/arch-cuda-double-tx/include/txpetscgpu/include/csr_tri_solve_gpu.h(173): here
>            instantiation of "int csr_tri_solve_gpu_hybrid(const csr_matrix_data<index_type, value_type> *, index_type *, uplo_enum, value_type *, value_type *, const csr_matrix_data<index_type, value_type> *, value_type *, value_type *, index_type, index_type, unsigned int) [with index_type=PetscInt, value_type=PetscScalar]" 
> aijcusp.cu(815): here
> <<<<< and more >>>>>>>>>
> 
> On Mon, 3 Oct 2011, Dave Nystrom wrote:
> 
>> Hi Barry,
>> 
>> Just to be safe, I cloned the latest version of petsc-dev.  I then cloned
>> cusp and thrust and installed them in /usr/local/cuda/include.  I then ran my
>> petsc build script and the build failed.  I've included the petsc
>> configure.log and make.log as well as the log from my makePetscDev script.
>> 
>> Thanks,
>> 
>> Dave
>> 
>> 
>> 
>