Has anyone compared the lastest version hspice with the older one in terms of speed?
i did a simple simulation using hspice 2007.03 and hspice 2004.03, using a same netlist, same tolerance settings, i didn't see any speed improvement, and the newer one is even 10 seconds slower. the resulting waveform is exactly the same. anyone can explain?
I have compared hspice 2006.09 and hspice 2005.09 for Linux using a very small circuit. With default options (runlvl=3 in 2006.09), the former is a bit slower than the latter. When runlvl=1 in 2006.09, the speed is about 1/3 faster than that of 2005.09 ( this version doesn't support runlvl option).
Thanks for the information.
I think the newer hspice is getting faster speed by adding in an option to adjust the accuracy. that runlvl is a parameter which can scale all the other tolerence parameters at the same time, e.g abstol, vntol, reltol, etc.
it is not getting real faster, just more features.
I was told that Hspice 2007.9 is very very fast. There was this code which took Hspice 2007.3 3 days to simulate it. Apparently HSpice 2007.9 does the simulation in 5 MINUTES !!!!!!!!!!!!!!
It is correct. i just compared the 2007.09 and 2004.03, this time the speed of 2007.09 is 7 times of the other and the resulting waveform matched perfectly. Very very nice for me, now i can finish simulation real fast.
Thanks for the information.
I think the newer hspice is getting faster speed by adding in an option to adjust the accuracy. that runlvl is a parameter which can scale all the other tolerence parameters at the same time, e.g abstol, vntol, reltol, etc.
it is not getting real faster, just more features.