Forum Discussion
Altera_Forum
Honored Contributor
11 years agoHi,
I can't tell much about cycle accurate timing simulation, but one sure thing is if the workflow in your kernel is data dependent, then it is very hard to get a good performance measurement in advance. If your kernel is of type NDRange and you activate the resource-driven optimizer, you will get some estimate such as how many work-items are processed per unit of time. All in all, the best performance measurement I have found is by activating the profiling in your OpenCL code and timing the kernel execution. This, along with the Altera profiler. I hope this helps.