Failing GPU tests on CUDA 8

[mattndu]

I'm running on a GTX 1070. This is compiled with CUDA 8.0 release candidate.

The ./test_gpu script fails with the following error:

Running GPU tests
terminate called after throwing an instance of 'std::runtime_error'
  what():  Error: compute_ctc_loss in small_test, stat = execution failed
Aborted (core dumped)

Attaching a debugger, I see:

(gdb) run
Starting program: /warp-ctc/build/test_gpu 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Running GPU tests
[New Thread 0x7fffef84b700 (LWP 10325)]
[New Thread 0x7fffef04a700 (LWP 10326)]
terminate called after throwing an instance of 'std::runtime_error'
  what():  Error: compute_ctc_loss in small_test, stat = execution failed

Program received signal SIGABRT, Aborted.
0x00007ffff6d55c37 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56
56  ../nptl/sysdeps/unix/sysv/linux/raise.c: No such file or directory.
(gdb) bt
#0  0x00007ffff6d55c37 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56
#1  0x00007ffff6d59028 in __GI_abort () at abort.c:89
#2  0x00007ffff7660535 in __gnu_cxx::__verbose_terminate_handler() () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#3  0x00007ffff765e6d6 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#4  0x00007ffff765e703 in std::terminate() () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#5  0x00007ffff765e922 in __cxa_throw () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#6  0x0000000000403e33 in throw_on_error (message=0x4097f8 "Error: compute_ctc_loss in small_test", status=<optimized out>) at /storage/deep_learning/warp-ctc/tests/test.h:11
#7  small_test () at /storage/deep_learning/warp-ctc/tests/test_gpu.cu:63
#8  0x000000000040360f in main () at /storage/deep_learning/warp-ctc/tests/test_gpu.cu:333

[ekelsen]

The default compilation flags don't compile for the Pascal architecture, you'll need to modify CMakelists.txt to include sm_60.

[mattndu]

That worked, thank you! For reference, here is what I added to CMakeLists.txt:
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_60,code=sm_60")

[ahbon123]

I installed cuda10.1, RTX 2080 GPU, but still have this problem.

# need to be at least 30 or __shfl_down in reduce wont compile
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_30,code=sm_30 -O2")
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_35,code=sm_35")

set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_50,code=sm_50")
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_52,code=sm_52")

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")
ENDIF()