I have a group of PVT simulation data (~50 sets) and have been unsuccesful in creating a plot with ViVA. This problem can be attributed to the fact that I have limited memory installed on my PC and my attempts had resulted in memory allocation error.
At the moment, I have an ocean script that utilizes for-loop to go through each simulation data and plot the desired waveforms. What I noticed during the script execuation is that the child processes, java and virtuoso, will consume ever increasing memory with each plot command until all available memory are used up and the program will exit with the memory allocation error.
Is it possible get java and virtuoso to release up the memory before OCEAN opens the next set of simulation data? What I am thinking is something like this: Open results -> plot results -> freeze graphc -> clear memory -> open next results and etc?
Also, is it also possible to enable Fast Viewing Support in OCEAN with the hope that this step will result in lower memory requirement?
I have IC18.104.22.1680.12
Your best bet would be to move to IC615, because that has much improved memory management for waveform data (as well as the new graph implementation). You could run in 64 bit mode to help though in IC614 (run "virtuoso -64").
In reply to Andrew Beckett:
I would try moving to IC615 to see if that version will help to alleviate the memory management.
Is there a way to make sure that plotting with fastview is enable in OCEAN? I did envGetVal("spectre.outputs" "fastViewOption") and the system returned "t", however, I still find the plotting to be very slow.Thanks!
In reply to slim5:
First of all, the fastViewOption will only have an effect if RTSF output is turned on for the simulation (and that will only do anything if the output format is psfbin or psfxl). Secondly, it only helps for signals directly plotted in the waveform tool, and if you build a signal and plot it using plot(), you will most likely not get the speed benefit. That's because it has to load in the entire waveform, and the optimisation is lost.
psfxl output format should help significantly in terms of data load time, so use that if you can (this is the default from IC615 ISR4 onwards).
And IC615 will help in terms of load and draw performance anyway because in IC614 the graph is a separate (Java) process, and so the data is loaded into the ocean/virtuoso process, and then may be sent across to the java process using IPC; in IC615 it is a single process using Qt graphics, and so there's no IPC needed. The graphics performance should (in general) be better.
If after doing this you still have performance issues, please contact customer support - this is something that has had a lot of focus on, especially in IC615.