Wilson Snyder
Wilson Snyder
A crash in V3Dead or V3Broken often indicates an earlier transform messed something up. A pointer of 0x1 is used in an object that was earlier deleted. So the crash...
Yes, it probably got mistracked in scope creation. Interfaces are complicated to handle correctly as they are both data types and modules.
V3Dead should get a correct tree as an input, it shouldn't have to do other resolution work, so it's presumably a problem in V3LinkDot(?)
What matters is the interface level > level of any modules using it. It is a rank of the graph you posted.
I see it keeps calling randomize forever (add some displays), it's not stuck inside the solver. As usual can you please remove all of the unneeded stuff and debug the...
This should have thrown a BLKANDNBLK error, as the scheduler doesn't support mixed array assignments like this. If you use this logic [1:0] pp; it will properly warn. One workaround...
@solomatnikov note this is pending on you, or if you prefer please close, thanks.
@gezalore we improved BLKNBLK to either warn, or properly simulate (or at least so says the docs). This case seems to still neither warn nor give correct results relative to...
As above: >I see it keeps calling randomize forever (add some displays), it's not stuck inside the solver. As usual can you please remove all of the unneeded stuff and...
Great! Just note that Verilator is not at all optimized yet to give great performance with gate level models, especially gate level models with lots of timings. The steps would...