Task hang on CUDA tsak hanging when proecessing CUDA for a long time

Dear all,

I was just wondering if there is anyone suffered this problem like me?

about my CUDA program, using 256 bytes shared memory, and some texture memory.
the program will hang for running about 30 mins...
than i need to kill the process by tsak manager.


Yes, it was when I programmed a nice infinite loop.

maybe the compiler use the local memory to replace the shared memory~
check the ptx code~

dear Kyzhao,

thanks for ur suggestion.

 but, i still have some questions....

why the compiler will use the local memory to replace the shared mem,

 and not always happened. --> maybe running 30mins, it'll hang. maybe not

 even if it was replaced, cuda program will be hangeg...?

Thank you.

i agree with E.D. Riedijk.

if there is an infinite loop, it will happen…


Infinte loop isn’t used in this case.

I’m curious about if continuous fetching texture memory for each thread, anything happen?

For example,

dim3 my_grid(1000,1000);

my_threads(256, 256);

my_kernel<<<my_grid, my_thread>>> …

global void my_kernel()


    int nTmp = tex1Dfetch(TexA, threadIdx) + tex1Dfetch(TexB, nTmp_threadIdx) + tex1Dfetch( TexC, nTmp_threadIdx) + ... ;