Home machine:
GTX760, linux mint
time to create a context: milliseconds
Work machine (no monitor, ssh-ed into):
3x K20, RHEL 6
time to create a context: 5-10 seconds
Any idea why the speed of creating a context should be so different for the same code?
While using all three devices?
Does it change when you pick only one of the K20 as device?
First start or also on consecutive starts of the program?
Can I assume that by “creating a context” you’re referring to the time taken to compile the context, and not the time taken to generate the context object? I experienced something similar: 1-2 seconds to compile on my Quadro K4000 vs 5-10 seconds on two Tesla K40 accelerators.
I did some profiling and determined that for the machine with two Teslas, OptiX goes through the process of compiling twice - once for each card, even though they’re identical cards. Also, if your remote machine has to read your PTX files over a network (like mine does), that could be slowing it down as well.
This is strictly the time taken to execute the line
Context context = Context::create();
which takes over 10 secs on the work computer.
It takes the same amount of time on each run, not just the first run.
I’m not specifying a number of devices to use…not sure whether it is defaulting to 1 or all 3.