Hi
time of execution using time command is 10s user 4.5 s and sys 5.5s
time of execution using nvvp is 5 sec including host functions
please clarify how to believe on nvvp
You may be misinterpreting the output of the time command. Also, if you have significant CUDA start-up time, the usage of nvvp can cause this to disappear, since nvvp incurs that startup time instead of your application.
time may be difficult to use for comparison in these cases. You might want to instrument your program with host-based timing (e.g. gettimeofday) or cudaEvent based timing, and then compare that with nvvp if you wish.
why it takes long time for cuda start up ??
in simple module it is not taking much start time whereas in current module its taking a lot of time.