nvcc out of memory ptxas consumes all the system memory

[betete]

Hi, I’ve been porting a dispersion model code to run on the GPU. Something is not working and I want to do some debbuging. But, when I compile with the -G option, ptxas starts filling all the RAM and the swap memory until it is KILLED by the operating system. Moreover, it takes 5 to 6 times the tame it takes to compile when not using the -G option.

I’m working on Linux with CUDA toolkit 3.0.

Can anybody help me?

Thank you.

[ardisschool10]

Hi, I’ve been porting a dispersion model code to run on the GPU. Something is not working and I want to do some debbuging. But, when I compile with the -G option, ptxas starts filling all the RAM and the swap memory until it is KILLED by the operating system. Moreover, it takes 5 to 6 times the tame it takes to compile when not using the -G option.

I’m working on Linux with CUDA toolkit 3.0.

Can anybody help me?

Thank you.

Hi Betete,

did you solve the problem? I have a similar problem with one of my cuda programs. When I compile and execute with the -g -G option it is ok, but when I try to debug the program with the cuda-gdb, it crashes at the first cuda-malloc.

[ardisschool10]

Hi, I’ve been porting a dispersion model code to run on the GPU. Something is not working and I want to do some debbuging. But, when I compile with the -G option, ptxas starts filling all the RAM and the swap memory until it is KILLED by the operating system. Moreover, it takes 5 to 6 times the tame it takes to compile when not using the -G option.

I’m working on Linux with CUDA toolkit 3.0.

Can anybody help me?

Thank you.

Hi Betete,

did you solve the problem? I have a similar problem with one of my cuda programs. When I compile and execute with the -g -G option it is ok, but when I try to debug the program with the cuda-gdb, it crashes at the first cuda-malloc.