upgrade to cuda9(Volta) GPU arch

[dzhwinter]

Currently, I fix the cuda9 error and some issues in https://github.com/baidu-research/warp-ctc/pull/117. However, our branch has diverged with original repo, this PR cherry-picked from my repo, now it can be safely merged. Please review this PR in your spare time, thanks!

[t-vi]

@dzhwinter @wangkuiyi Say, would you be able to shed some light on the mask (0 if I see this correctly) you are using? in most other venues, it seems to be that shfl is replaced by shfl sync with all mask bits set. I would be most grateful!

[Diamondfan]

Seems not work for me with cuda9.2 and V100. With "./test_cpu: error while loading shared libraries: libwarpctc.so: cannot open shared object file: No such file or directory". Any suggestion?

[ThomasDelteil]

can this PR be merged? WarpCTC is failing on V100 otherwise.

edit: With this branch I am getting:

./test_gpu
Running GPU tests
Some or all tests fail

[aonotas]

@Diamondfan maybe you need to set

$ cd build
$ export LD_LIBRARY_PATH=./:$LD_LIBRARY_PATH
$ ./test_gpu

I also get same error as @ThomasDelteil.

$ ./test_gpu
Running GPU tests
Some or all tests fail

I use cuda-9.0 and Volta-100.

[PapaMadeleine2022]

use cuda10, and modify the code in CMakeLists.txt:

IF (CUDA_VERSION GREATER 7.6)
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_60,code=sm_60")
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_61,code=sm_61")
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_62,code=sm_62")
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_70,code=sm_70")
ENDIF()

it works.