This is part of my CUDA code. But last part of this code says some error message.
unsigned int *mat_count;
off_t *mat_position;
unsigned int *matches_count;
off_
Compile your application with debug flags nvcc -G -g
and try running your application inside cuda-memcheck
or cuda-gdb
. It might give you a hint where the problem might lie.
Just run
cuda-memcheck ./yourApp
An unspecified launch failure is almost always a segfault. You've got an indexing mistake somewhere in your kernel, probably while accessing global memory.
I'd look through your code, but it's mildly incomprehensible...