C
Cary R.
Guest
While implementing the stochastic (queue) tasks/function for Icarus
Verilog I ran into a number of under specified conditions. I resolved
these in what I consider a sane manner and would like to submit the
changes for consideration/inclusion in the next standard update. The
question is, other them mailing them to the address in the standard,
likely a black hole, what is the best way to get this to the people that
matter?
My normal standard contact person is currently also acting like a black
hole so I'm resorting to more drastic measures.
FYI my implementation should be 100% compatible with the standard while
addressing the limitations and extending the functionality slightly.
It's also interesting to note that every simulator I tested had at least
one bug in the implementation. Modelsim misinterpreted the definition of
average wait time to be the average wait time for item currently in the
queue. VCS incorrectly divided the interarrival time be N instead of
N-1. NCVerilog appears to have the same problem as VCS except a couple
of the values were incorrect. I'm guessing they missed the easy way to
calculate interarrival time and their routine has more than just the N
vs N-1 bug.
Thoughts, including you're crazy to even consider getting involved are
appreciated. If you want to contact me directly poking around the Icarus
Verilog SorceForge pages should be fruitful.
Regards,
Cary
Verilog I ran into a number of under specified conditions. I resolved
these in what I consider a sane manner and would like to submit the
changes for consideration/inclusion in the next standard update. The
question is, other them mailing them to the address in the standard,
likely a black hole, what is the best way to get this to the people that
matter?
My normal standard contact person is currently also acting like a black
hole so I'm resorting to more drastic measures.
FYI my implementation should be 100% compatible with the standard while
addressing the limitations and extending the functionality slightly.
It's also interesting to note that every simulator I tested had at least
one bug in the implementation. Modelsim misinterpreted the definition of
average wait time to be the average wait time for item currently in the
queue. VCS incorrectly divided the interarrival time be N instead of
N-1. NCVerilog appears to have the same problem as VCS except a couple
of the values were incorrect. I'm guessing they missed the easy way to
calculate interarrival time and their routine has more than just the N
vs N-1 bug.
Thoughts, including you're crazy to even consider getting involved are
appreciated. If you want to contact me directly poking around the Icarus
Verilog SorceForge pages should be fruitful.
Regards,
Cary