In version 1.1.6 of the CUDA Visual Profiler, the option for max execution time seems to only affect the percentage bar during the running of executables. It never actually changes the run time of the executable, it times out at 30s regardless of what is entered there.
I have an application which requires a 200mb file load and parsed (into main memory) at app launch which then data is selectively passed into my several kernels, eventually.
Unfortunately, the load and parse takes about a minute, so no kernels ever load so I get an error
Error in reading profiler output.
No data rows in profiler output file.
My application does run and properly. If I circumvent the load/parse I can get the profiler to work mostly, but again only until the 30s time out occurs.
Does anyone else have this problem?
my system is the following
CUDA v2.0
GPU: BFG GTX280
Driver: 181.20
CPU: intel Q9300
RAM: 2GB OCZ DDR2
MOBO: ASUS P5Q Deluxe intel P45 chipset and ICH10R
OS: Win XP
Compilers: MSVS2005 and Intel Compiler v11.0 for MSVS I’ve tried both