i wrote cuda code to caculate many sincosf,when i increase quantity of sincosf,my code was terminated after about 12 seconds and cuda gave error message:“the launch timed out and was to terminated”.i cheched shared memory and register consumed in code, they didn’t exceed my card capability.If anyone has any idea what is causing the crashes when my code needs much time to run on gpu, I would be grateful.
my cuda sdk and toolkit is 2.0, card is 8800 GTS. codes run in windows system. all the codes compile in visual studio 2005.
to avoid the watchdog timer, you either need to boot directly to a Linux console without running X or you need a dedicated compute card.
thank you,i will try linux
thank you,i will try linux