CUDA-GDB: the GPU does not have hardware debugging support on Tesla S870

Accelerated Computing CUDA CUDA Programming and Performance
1

Configuration:
CUDA version 3.0, OS Ubuntu 10.4 64bit,

CUDA-gdb version:
NVIDIA ® CUDA Debugger
BETA release
GNU gdb 6.6
This GDB was configured as “x86_64-unknown-linux-gnu”

GPU: Tesla S870

The error message is:

1437 cudaMalloc((void **)&d_para_v0_ptr, para_size);
(cuda-gdb)
Warning: a GPU was made unavailable to the application due to debugging
constraints. This may change the application behaviour!
CUDA-GDB: the GPU does not have hardware debugging support

Which is the first line that invokes cuda code.

The compilation cmd parameters I use are:

nvcc -g -G xxx.cu -o xxx_dbg -I /usr/local/cuda/include -I /var/lib/nvidia/shared/inc -I /var/lib/nvidia/C/common/inc --opencc-options -OPT:Olimit=0 --verbose --ptxas-options=-v

The CUDA-GDB manual does not list Tesla S870 as one of the exceptions that do not support hardware debugging. Is there any problem of my compilation, or is the hardware debugging is not supported by default, and i need to do something to activate it? Any suggestion?

2

It isn’t supported, and the cuda-gdb manual (or at least the copy I am looking at) say so on p16:

3

It isn’t supported, and the cuda-gdb manual (or at least the copy I am looking at) say so on p16: